Ch< 1000^03

Ia

INIS-mf—12693

— r~"~^T"™~~T"

<5 3 I • ' | —

II Lrin-. Paul Scherrer Institut

PSI Nuclear Energy ResearchProgress Report 1989

Paul Scherrer Institut Telefon 056 / 99 2111Wurenllngen und Vllllgen Telex 82 7414 psi chCH-5232 Vllllgen PSI Telefax 056 / 98 23 27

ON THE COVER:

Front view of a hot cell, equipped for non-destructive testing of light water reactor(LWR) fuel pins.A process controller is moving the pins through the measuring heads for pro-filomctry, eddy current defect and oxide thickness testing, cold gap and gamma-spcclromctry.

LrtH Paul Scherrer Institut

Annual Report 1989

ANNEX IV

PSI Nuclear Energy Research

Progress Report 1989

Edited by: Department Nuclear EnergyEditing: H. P. Alder, K. W. Wiedemann

Mrs. S. Maurer

Table of Contents

Introduction 1H. P. Alder

Overview of Nuclear Energy Research 31 Reactor Physics and Systems 3

2 Thermal-hydraulics 73 Materials Technology and Nuclear Processes 104 Waste Management Program 13

5 LWR Safely Program 16

Best-estimate Analysis of Control Rod Ejection Accidents in the Beznau nPressurized Water Reactor 19E, Knoglinger, L. A. Belblidia, M. A. Zimmermann et al.

Thermomechanical Behaviour of a Subassembly Hexcan under total Instan-taneous Blockage Conditions 27R. Attinger, W. Heer, P. Wydler, D. Desprez, J. Louvet, A. Zucchini

Boil-off Experiments with the PSI-NEPTUN-Facility: Analysis and CodeAssessment Overview Report 32S. N. Aksan, F. Stierli, G. Th. Analytis

Effects of Vapour/Aerosol and Pool Formation on Rupture of Vessels ContainingSuperheated Liquid 38J. Schmidli, S. Banerjee, G. Yadigaroglu

High Temperature Materials 47G. Ullrich, K. Krompholz

Experience with the VS-Decontamination Process Developed at PSI 52

E. Schenker, D. Buckley, H. P. Alder, W. Francioni

The Effects of Non-linear Sorption on Radionuclide Transport 55

P. A. Smith, A. Jakob

Complexation of Cu2+, Ni-+ and UO2-+ by Radiolytic Degradation Productsof Bitumen 59Z. Kopajtic, L. R. Van Loon

Local Monitoring Technique in Acoustic Emission at PSI 65

P. Auinger, B. Tirbonod, L. Hanacek, E. Schneider, K. Wulle, M. Emmenegger

Residual Stresses Due to Multipass Welding in Radioactive Waste Overpacks 71

R. Attinger

Teaching Activities and Lectures 77University Level Teaching 77Teaching at other Schools and Colleges 77Lectures 77

Scientific Publications 79PSI-Reports 79Common Reports 80

Publications in Scientific and Technical Journals and Conference Reportsas well as other Scientific Reports 80

Introduction

H. P. Alder

Head of the Department Nuclear Energy, ad interim

The Department Nuclear Energy is concerned with re-search, training and providing scientific-technical servicesin the field of nuclear technology. Apart from some activ-ities at the Swiss Federal Institutes of Technology (ETH),this department is the only organisation in Switzerland ableto handle basic problems of a high technical and scien-tific level which exceed the capacities of the authorities andthe nuclear power plants. To be successful, this requires acostly infrastructure, collaboration with partners and thirdparty financing.

The activities in 1989, the second year of PSI, werequite successful. The basic functions mentioned at the startwere not only carried forward but strengthened. Severalfactors contributed to this result:

The concern of the personnel arising from the neworganisation could be gradually reduced. The numberof people who left the nuclear field for other activitieswas relatively small.

The involvement of professors of the ETH Zuerichand EPF Lausanne in the activities of the departmentwas found to be of benefit to both sides.

The allocation of the PSI funds more or less met ourrequirements. Together with third party financing thefunding not only allowed the purchase of equipmentbut also improvements in the infrastructure.

- About one quarter of the operating budget and man-power was covered by third party financing from thelicensing authorities, the NAGRA, the nuclear powerplants and industrial firms. It is gratifying that thenuclear community in Switzerland expressed their in-terest in PSI research not only in words but also incash.

DEPARTMENT NUCLEAR ENERGY

PROGRAMS

WASTE MANAGEMENT

LWR-SAFETY

Aerosols andFission Products

Structural Mechanics

LABORATORIES

REACTOR PHYSICSAND SYSTEMS

Reactor Physics andNumerical Methods

System Engineering

Research Reactors

Reactor School

THERMAL-HYDRAULICS

Heat Transfer andFluid Dynamics

Thermal-Hydraulicslor LWR-Salety

MATERIALS TECHNOLOGYAND NUCLEAR PROCESSES

Structural andFracture Mechanics

Process Technology

Hot Laboratory

PROJECTS

HIGH TEMPERATUREREACTOR

ADVANCED PWR

SIMULATION OFTRANSIENTS STARS

THERMAL-HYDRAULICS IOF LIQUID METALS I

NUCLEAR FUEL

LWR CONTAMINATIONCONTROL

- The five-year plan 1988 to 1993 calls for a reductionin PSI staff in the nuclear area of some 20 %. In 1988and 1989 the step-wise reduction could be balancedby third party financing. It is hoped that this will alsobe the case in the coming years.

- In selected areas an effort was initiated to transferthe nuclear know-how to non-nuclear research prob-lems, e. g. in thermal-hydraulics for heat exchangeoperations in the industrial area or, based on nuclearfuel technology, for the development of high-qualityceramics.

The activity of the Nuclear Energy Department is based onthe following laboratories:

- Laboratory for Reactor Physics and Systems (hightemperature reactor project, advanced pressurised wa-ter reactor project, project for the simulation of tran-sient behaviour of Swiss reactors).

- Laboratory for Thermal-Hydraulics (thermal-hydrau-lics of liquid metals project)

- Laboratory for Materials Technology and Nuclear Pro-cesses (nuclear fuels project, project for LWR con-tamination control)

The following programs are part of the matrix structure:

- LWR safety program

- Waste management program

Overview of Nuclear Energy Research 1989

1 Reactor Physics and Systems

The Laboratory for Reactor Physics and Systems (LRS) isinvolved with the study of advanced reactors and their to-tal systems, with systems analysis of current reactor plants,with Ihe modelling of physical-chemical processes in con-nection with nuclear waste management, with the operationof research reactors and with the training of specialists inthe area of reactor technology.

The LRS activities are moving increasingly in the di-rection of basic research which contributes to the betterunderstanding of the connections between systems technol-ogy and safety in the sense of "keeping the nuclear energyoption open". To this belongs the successful PWR physicsexperiment on the tight pitch lattice LWR (APWR). Thiswill make it possible to actively engage on new "still safer,still more economic" reactor concepts in the future.

In this sense also our HTR-Proteus experiment in thepreparation stage deals with basic questions of reactor phy-sics which can contribute to improvements in a future hightemperature reactor. Similarly the calculation of presentreactors in the framework of the STARS project help tovalidate and improve LWR calculation models for designand licensing of small heating reactors for example.

Fast reactors also have a role to play in the long term useof nuclear energy. In order to follow the further technical-scientific developments the LRS carries out selected basicresearch on these reactors in an international framework.The experience thus gained was of special value in 1989 asthe LRS scientists became involved in deep discussions onthe subject of breeder reactor safety, and were thereby ableto factually and competently assist the federal authorities intheir dealings with the opponents and their experts.

1.1 New and Safer Reactor Types

In the framework of studies aimed at choosing suitable newreactor concepts having a sufficiently large research poten-tial, the LRS is often confronted with what could be thelikely future development of any given reactor type. Studyof the available literature shows that a direct comparison isimpossible in that data or comparable approaches are miss-ing. For this reason the LRS was a co-iniuator of a Swissproposal for an international study in this area.

During 1989 the LRS was actively engaged in the prepa-ration of this study. In mid 1989 work was started under theleadership of the laboratory on the needs for the develop-ment and the future potential of small and medium reactors

with the Nuclear Energy Agency of the OECD. In particu-lar, the topic of nuclear heat generation was examined andan attempt was made to evaluate the conditions for a moreintensive growth of nuclear energy in the heating market.For while the heating reactor activities at PSI are practicallyat a low level it seems that the heating market could be animportant beneficiary of nuclear energy in the future. Thisview is receiving increasing recognition internationally.

1.2 The STARS Project

STARS is a research activity which is being carried out inconjunction with the HSK with participation of the sub-committee for nuclear power of the Swiss Utility Associa-tion. In the framework of this activity which began in 1988and is planned to run for six years, mathematical modelsare to be developed and verified which cover the simulationof stable and transient behaviour of the core, fuel and bal-ance of plant in each of the nuclear power reactors in thiscountry. The simulations arc to deal with all the thermal-hydraulics and reactor induced disturbances and accidents,including loss of coolant accidents which might occur inboiling and pressurised water reactors. For this the insti-tute has access to computer programs both from its ownresearch programs, from international agreements or fromparticipation in international research programs.

The development and verification of plant and transient-specific simulation models and the updating of the data col-lections required for analysing off-normal events, makesit possible to have a rapid access to a realistic accidentanalysis in the case of an actual event at any of the plants.

In addition to the development and testing of (lie modelsand the continuous updating of the data collection whichform the main pan of the research it is intended to developalso a computer supported data management system whichwill compile and store not only the plant descriptions anddata but also analysis results. In this connection the use ofexpert systems is also foreseen.

This general aim was extended shortly after starting theproject at the wish of the HSK to include the analysis of aconcrete beyond-design- basis control rod transient for thenuclear reactors Bcznau (KKB-I1) and Muehlebcrg begin-ning with a problem of a control rod ejection in KKB-II.

Whereas in the first year of the project, after developingand testing a suitable point kinetic model for core and plantof the reactors the work was concentrated on parameterstudies and sensitivity analyses for the design-basis scenar-ios, the work in the year now under review was concerned

with the analyses of these accidents under beyond-design-basis conditions. This together with the commissioning ofthe computer programs RELAP5-MOD2 and TRAC/BF1needed for the LOCA calculations formed the main part ofthe work in 1989.LOCA analysis

In the area of LOCA analysis which is carried out incollaboration with the Laboratory for Thermal-hydraulics(LTH), a program for analysing specific breaks in the Muchlc-berg and Beznau nuclear plants was formulated and the im-plementation of the best estimate codes RELAP5-MOD2and TRAC/BF1 which are needed for this analysis wasstarted. Both programs are installed on the CRAY/X-MP.RELAP5 is ready for operation and is to be used, after be-ing successfully tested, for the analysis of the double endedguillotine break in the cold leg of KKB-II. Analyses for de-termining the most important factors affecting this accidenthave been started.

The control rod ejection in KKB-II as beyond-design-basis reactivity accident:

In order to limit the result of a control rod ejection inwhich it is assumed that a sudden fracture of the controlrod drive support on the reactor pressure vessel head causesthe control rod to be violently shot out, suitable measuresboth in the design and fabrication stages as well as duringoperation are taken to ensure that approved limiting valuesfor system pressure, fuel and cladding temperatures duringand as a result of the accident arc not exceeded.

The underlying beyond-design-basis scenarios for ana-lysing the control rod ejection assume that one or severalrods are outside the permitted range of travel before theaccident is initiated, in other words arc driven further intothe core than allowed. Two types of bcyond-design- basiscases are considered:

- The rod ejection in a hot shutdown reactor (zeropower) with fully inserted rods of the two controlbanks A and B.

- The rod ejection at design power with the banks Aand B on the insertion limits. The highest worth rodseparates from its coupling just before the accidentinitiation and falls into the core.

Since all the nuclear parameters which are influencedby the accident such as reactivity worth of the control rods,initial power density distributions, reactivity coefficients forfuel and moderator temperature and not least the kineticparameters are dependent on the bum-up state of the core,both the start as well as the end of the 16th operating cycleare analysed.

The listed parameters were determined using the PSIcode system ELCOS and where available compared withthe operating parameters of the nuclear plant.

The transient analyses were carried out with RETRANfor point kinetics and using QUABOX/CUBOX-HYCA,(QCH) for space and time dependent calculations whichsolved in three dimensions the transient neutron diffusionequations in two energy groups taking into account thethermal-hydraulic feed back effects.

Powerdensity

t = 0

0.15s

0.8 s

Fig. 1: Radial power density distribution in the upppercore region in the case of an ejection of the most reactivecontrol element from the shutdown reactor at the end of the16th operation cycle of KKB-II. The ejection lasts 0.1 s (t:time from accident initiation)

The cross section libraries for the transient QCH anal-yses were established using the two dimensional PSI pro-gram BOXER which produces for each element in the core2-group constants dependent on burn up, control status,moderator temperature and density, fuel temperature andboric acid concentration in coolant based on the multi-grouptransport theory.

The analyses confirmed that even under the extremelyunlikely assumptions on which the selected scenarios werebased, (simultaneous fracture of the support and incorrectinsertion of the control rods) no design criteria are exceededrelated to the safety analysis of the plant. The design of theplant therefore contains the selected "beyond design basisaccident" in a conservative manner, even though in the caseof a rod ejection from the hot shutdown reactor which leadsto a super-prompt criticality the reactor power rises for abrief period to around 50'000 MW.

The power excursion is completely restrained by thereactivity effects in the fuel (dominant) and coolant wherebythe spacial distribution of the power density (Fig. 1) andthe neutron flux in reactor effectively support this feedbackeffect.

The most important results of the three dimensional ac-cident analysis were presented at the second technologicalcommittee meeting of the IAEA on "Safety aspects of re-activity initiated accidents in November 1989 in Vienna.

1.3 Research Reactor SAPHIR

The research reactor SAPHIR was operated in the past yearin a four week cycle (three weeks of three shift full poweroperation and one week low power operation) wilh an avail-ability of 95 % of the planned 6053 operating hours. Dueto the shortage of licensed operators an additional weekshutdown occurred in the autumn.

Continuous 10 MW full power operation was providedfor the isotope production, for the beam tube operation forIhc spectrometers of the Laboratory for Neutron Scattering,for diploma studies for the university of Fribourg, aerosolchemistry (gas-jet), numerous activation analyses, the dos-ing of 158 kg of semi-conductor silica for ABB as wellas various irradiations for other research groups such asST1LO. For this a total of 238 isotopic clement movementswith a total irradiation lime of 12072 hours were carriedout. The chemistry laboratory carried oul 580 irradiationswith the pneumatic lube. On average Ihc reactor in fulloperation undertook some six experiments simultaneously.

The low power operalion served to irain ihc SAPHIRoperators, ihc provision of practical reactor training for ofIhc reactor school for power plant operators, for the fed-eral technical colleges and HTL students as well as forthe necessary maintenance, repair, core loading and opera-tional measurements. In December the burn up of 62 fuelelements was determined by reactivity measurements. Inorder to verify the results of this costly measuring methodburn up measurements arc planned in 1990 using gammaspcclromctry.

The licensed operating team were engaged for 80 % inshift work and training new personnel. During the year iwoof the staff were successful in becoming operators, threeshift leader and two reactor chief I after having completedtheir licensing examinations.

The change over from highly enriched to low enrichedfuel was largely completed on SAPHIR. After starling up anuclidc specific activity measuring cell developed in-house(high resolution gamma spcctromeiry) it is now possible todetect the type and amount of nuclides directly in the poolwater circuit. An aerosol measuring unit with a fixed fillerfor checking the activity in the hall with specific nuclidcdetection is being prepared.

A large part of ihc existing nculronic instrumentationof SAPHIR has been in continuous operation since the sev-enties. Therefore in the last years (especially in the yearnow reported) a review has been under way for replace-ment instrumentation from well known European suppliersin readiness for placing orders.

In parallel to replacing the increasingly unreliable ncu-lronic instrumentation with interface compatible units, adigital, decentralised control system based on a redundantbus system (PROCONTROL or ABB) is to be installed slcpby step alongside ihc existing system, first in order to coverthe process data collection, the alarms, and the recording,later Ihc reactor control and finally Ihc rcaclor protection.

The bus technology permits access to all of the signalsof the plant for the central process data collection. Viaa PROCONTROL-PC interface (PIF) now in developmentthe processes can be visualised in a real-time data bank ofa man-machine communication system network. Participa-tion in the development of the interface being carried oulby ABB Baden is a demanding and motivating la.sk for thePSI personnel working in die area of process information(Fig. 2).

The process visualisation on colour monitors reducesthe need for conventional inslrumcnuition in the control

Fig. 2: Signal flow from a process (research reaclor SAPHIR)for the planned man-machine communications system.

room and limits the need for frequent system changes, whicharc typical for research reactors, to modifications to thesoftware. Networking with other work stations via Eth-ernet opens further possibilities such as fuel management,management of irradiations and archiving. In addition theintroduction of the software interface now in developmentand the further development of an expert system being in-stalled in SAPHIR will help in supporting the operatingteam. The shift teams will find simplified operating proce-dures in the control room and be relieved from monotonousmonitoring and manual recording duties.

In the reporting year a PC operator work station withfunction keyboard and a dynamic representation of theSAPHIR system was provided for the operators for the pur-pose of testing and optimisation of the operating interface,alarms and recording, to demonstrate part simulations quali-tatively in real lime. In this way the needs and the operatingexperience in designing the system could be continuouslychecked and an understanding and acceptance of the newtechnology could be gained.

1.4 Project Advanced Pressurised Water Re-actor (APWR)

An improved fuel utilisation is the main objective of ahigh conversion so-called advanced pressurised water re-actor whose core consists of a hexagonal PuO-..AJOL. fuellattice. The investigation of the reactor physics and emer-gency cooling thermal-hydraulics lor such reactors (see alsoch. 3) have been followed for some years at PSI in closecollaboration with the Karlsruhe Nuclear Research Center(KIX) and Sicmcns/KWU.

Fig. 3: Using the rod changer designed at PSI the relativefission product activity of up to 15 single PuO^/UO.., fuelrods irradiated in PROTEUS can be measured. Such mea-surements contribute to a data base for the verification ofcalculated power density distributions in tight test latticesof the APWR.

The studies using the zero power reactor PROTEUS atPSI provide a broad d«ta base for the experimental verifica-tion of the APWR physics calculations. Experiments withand without void simulation in reference lattices with mod-erator to fuel ratios M/F of 0.5 and 1.0 were completed in1989. The neutron balance in pure PuOs/UOo test latticesand in B4C poisoned lattices were studied (Fig. 3). ThePROTEUS results have shown that from a safety point ofview the void reactivity coefficient for a not too light APWRlattice (M/F approx 1.0) is acceptable. Although the agree-ment between the calculated and the measured reactivityeffects appear to be sufficiently close different compensat-ing errors in calculating single neutron balance componentshave been found. Therefore it is necessary to make furtherimprovements to the calculation methods and data.

1.5 PROTEUS-LEU-HTR Experiments

Gas cooled high temperature reactors (HTR) are a valuableoption for the future development of nuclear technology.Their inherent safety characteristics and unique design suchas fission product barriers and structural integrity up to veryhigh temperatures, high heal capacity in core etc. makethem especially suitable for producing nuclear electricityand process heat on sites close to areas of high populationdensity.

HTRs have been well studied for a long lime, howeverthe change to a lower enrichment, the abandonment of theuranium/thorium cycle as well as the introduction of newcore materials (for example Hafnium as burnable absorber)has shown thai there is a lack of experimental data for val-idating design and safety analyses. Further in small HTRswhich arc attracting increasing interest, some less well stud-ied effects such as reactivity increase as a result of wateringress arc particularly important.

In order in these areas to provide experimental datato reduce the uncertainties in the design of helium cooledreactors with low and medium power and with low enricheduranium (LEU), thus simplifying the licensing, a range ofcritical experiments is planned in the zero power reactorPROTEUS.

The main aim of the new experiment is for the first timeto obtain measured data of high quality in relation to a) thecriticality of simple and easily interpretable HTR systemswith a single core region and LEU fuel for different mod-erator ratios and several lattice geometries; b) the changesin reactivity, neutron balance components and the controlrod worths as a result of water ingress; and c) the effect ofboron and/or hafnium absorbers which are used to modifythe reactivity and power distribution in typical HTR sys-tems.

Permission has been received to conduct these new ex-periments and work on the design and approval of the mod-ified PROTEUS facility (Fig. 4) is at the moment underway. The Swiss contribution to the international LEU-HTRexperimental program consists carrying the costs of design,licensing and operation of the facility. PSI is in additionproviding a large part of the scientific personnel. The agree-ment of the FRG to supply the spherical LEU fuel compactsfor the initial experiments as well as key scientific supportwas vital in planning these experiments. The necessary in-ternational contractual and safety agreements for the trans-port of the fuel to PSI are about to be realised. The start ofthe critical experiments on the HTR is planned for 1991.

The experiments have been accepted as part of the co-ordinated research program of the IAEA under the title of"Validation of the safety related reactor-physics calculations

Fig. 4: Critical LEU-HTR experiment in the PROTEUS re-actor: arrangement of the pebble bed reactor in the graphitereflector.

for the low enriched HTR" in the framework of the workinggroup "Gas cooled reactors" of the Vienna agency. In ad-dition to the main cooperation Switzcrland-FRG, the USA,USSR, Japan and the Peoples Republic of China have de-cided to participate. The partners in this international col-laboration met at PSI in October 1989 in order to define thefirst details of the initial phase of the experimental program.

As a result of this meeting and based on detailed PSIcalculations on various design configurations for the facil-ity, it was decided to begin the first phase of the tests usinga pebble bed core with the lowest moderator ratio of 1:2(moderator to fuel pebbles, atomic ratio C:U approx 950:1)with a diameter of 125 cm.The experiments will continuewith other moderator ratios and deterministic as well as sto-ichiastic pebble loading geometry so that an adequate andcomplete data bank will exist for the validation of designcalculations for pebble bed as well as prismatic-LEU HTRconcepts. A new safety report is being written for ihe zeropower reactor PROTEUS.

1.6 Reactor School

The reactor school provides a federally recognised tech-nical college specialising in power reactor technology fornew operators and offers courses for power reactor staff re-quiring licenses as well as practical reactor experience forstudents studying reactor theory.

The technical apprenticeship course T-22 begun in 1988continued in the reporting year and ended in April 1989.The lraining consisted of two semesters of theory (44 weeks)and a practical course on a power plant (13 weeks). The twosemesters contained general training (German, informatics,working practices, English, history of nuclear energy), ba-sic topics (mathematics, physics, nuclear physics, radiation)and topics for the specialist training (radiation protection,reactor kinetics, reactor dynamics, thcrmo-dynamics, me-chanical engineering, electrical engineering, nuclear plantcontrol, nuclear plant safety and nuclear plant systems).Thirteen participants from Swiss nuclear plants success-fully completed the technical apprenticeship and receivedthe "technician TS" diploma specialising in nuclear powerplant technology. This is recognised by the HSK whenlicensing reactor operators as proof of a sufficient basictheoretical training. The official recognition of the techni-cal college by the federal cabinet occurred on 18 December1989. In future the apprenticeships will be carried out inan overlapping annual cycle.

In the reporting year 21 refresher courses were carriedout. These included 4 three day thermodynamic courses fora total of 30 participants from KKL, 4 three day radiationprotection courses for 37 participants from KKG, 9 twoday PWR simulator courses for 46 participants from KKG,1 four day and 2 two day BWR simulator courses and 1three day control system course for 24 participant fromKKM. Further 5 participants of KKB and 4 from SAPHIRtook part in a five day shift manager course in radiationprotection. In addition 9 HSK staff took part in 2 five daysimulator courses.

In 1989 5 reactor practical took place with 95 stu-

dents. The reactor practicals included in addition to nuclearphysics and radiation protection experiments also exerciseson the SAPHIR research reactor on reactor statics and ki-netics and demonstrations on the reactor school simulatoron the dynamic behaviour of a pressurised and a boilingwater reactor.

2 Thermal-hydraulics

The name thermal-hydraulics includes the combined spe-cialities of thermodynamics, heat transfer and fluid flow.Thermal-hydraulics is used in almost all areas of industry,but plays a particularly important role in energy productionand process technologies and therefore in the design andsafety analysis of nuclear power plants.

The research activities of the Laboratory for Thermal-hydraulics (LTH) was concentrated in the year under reviewmainly on ihe continuation of the safely analyses of today'slight water reactors (LWR) as well as studies on the coolingof heat exchanging equipment containing liquid metals.

Only the newer activities, eg. selected thermal-hydraulicproblems concerning new reactor concepts, release of super-heated liquids and the liquid metal cooling of the SINQtargets are described in more detail below.

2.1 LWR-Thermal-hydraulics

Two phase flow and heat-transfer problems important forthe accident and safety analysis of light water reactors wereexperimentally and theoretically studied.

A long term research program is aimed at building up"know how" in this area in order to competently advise theSwiss licensing authorities.

Parts of this research program were carried out in theframework of agreements between the American regulatoryauthorities (NRC) and the federal office of energy (BEW).The main aim of this work is the realistic modelling andcalculation of the on-going physical phenomena in lightwater reactors under transient and accident conditions.

Participation in the International Code Assessment pro-gram (ICAP) on the use, evaluation, and improvement ofparticular calculation programs such as RELAP5 and TRAC,form the basis for studying loss of coolant accidents inSwiss nuclear power plants. In this year the data collection,calculations and analyses for the plants KKB and KKM be-gan.

The experimental work (NEPTUN) on simulating LWRemergency cooling situations in an advanced pressurisedwater reactor (APWR) was carried out. The experimentdata obtained previously was analysed and influence of thedifferent lest parameters was studied. In the frameworkof the tripartite APWR studies (Siemcns-KWU, KfK andPSI) a workshop was organised in which, using data fromNEPTUN, Ihe still existing deficiencies in the calculationmethods and the computer models could be demonstrated.

A

Fig. 5: Outline (left) and isometric view of the DY-NAFLASH experiment.

ABCDE

Core simulatorPlate type heat exchangerPrimary circuit - diffuserPrimary condenserSecondary condenser

GHIKL

Interm. hold upCold poolWarm poolCooling circuitMixing basin

F Cleaning circuit - diffuser

2.2 DYNAFLASH Experiment

In the reporting year the DYNAFLASH Experiment wasbuilt (Fig. 5). The basis for this test facility was the 10MW Geyser-Heating reactor which was scaled to a 30 kWfreon facility. The experimental aims of the DYNAFLASHfacility are:

- Studies of the dynamics of the natural circulation cir-cuit which are driven by the so-called flashing. Flash-ing means the evaporation of a liquid with a drop inpressure without the addition of heat.

- Simulation of the dynamics of the reactor for typicalload changes of a district heating system (continuousor sudden increase or decrease in the heat demand).

Comparison of experimental and calculated results.

The DYNAFLASH experiment contains all the relevant cir-cuits. The reactor core is simulated using electrically heatedelements. The power regulation using the boron content ofthe water is replaced by an electronic controller having asimilar time constant. The plant is operated with freon(R113) in the primary and secondary circuits. The tertiarysystem which simulates Ihe district heating network useswater. A large part of the facility is made of glass compo-nents. In this way different phenomena of two phase flows(liquid and gas) and the liquid levels can be observed.

Freon has, in comparison to water, various advantagesand is therefore often used in lest facilities. It is thus pos-sible to study the behaviour of the fluid at pressures anatemperatures a factor of ten below as that of water itself.In this way also the heat losses are reduced. A compari-son of the data of a 10 MW reactor and the DYNAFLASHExperiment shows this in Table 1. In addition the heightdifferences needed to drive the natural circulation are re-duced by a third, since the density of R 113 compared towater is about one and a half times larger.

The experiments will be carried out in the Laboratoryfor Thermal-hydraulics in collaboration with the Labora-tory for Nuclear Technology at the ETH Zuerich in 1990.

Table 1Comparison of the 10 MW reactor with theDYNAFLASH

Primary circuitPowerPressureTemperatureFlow rateMain circuitFlow rateCleaning circuitSecondary circuitPressureTemperatureFlow rateTertiary circuitSupply temperatureReturn temperatureFlow rate

10 MW reactor10 MW2.7 bar129.9 C

488.3 kg/s

160 kg/s

2.1 bar121.8 C468.6 kg/s

120 C70 C47.6 kg/s

DYNAFLASH30 kW1.4 bar57.5

4.9 kg/s

1.1 kg/s

1.05 bar48.4 C14.3 kg/s

30 C27 C2.4 kg/s

2.3 Release of Super-Heated LiquidsReactive and poisonous substances arc often handled andstored as liquids under pressure. In the case of a suddenrelease a part of the liquid evaporates since the evaporationtemperature under room pressure is well below the roomtemperature. In this case a cloud is formed of gas anddroplets which are usually heavier than air which causes itto sink and mix with the air. Examples of liquids storedunder pressure are propane, butane, ammonia and chlorine.In recent years a number of catastrophes have occurred asa result of such unexpected releases.

In order to estimate the risks in handling materials whichare under pressure, it is necessary to determine how the

cloud which forms on release disperses in lime and space.For calculating the dispersion computer models are alreadyavailable whose data are based on analysis of accidents.One of the problems of these models is that the. startingconditions are insufficiently known. The reason for this isthe poor understanding of the physical mechanisms. In therelease of overheated liquids three mechanisms play a role:

1 Expansion and evaporation of the liquid on suddenpressure reduction

2 Sinking of the aerosol cloud due to the higher densitycompared to the air

3 Turbulent mixing with the surrounding air

The first mechanism was studied in the frame of thisproject. The liquid R-l 14 was chosen for the first test whichhas similar properties to butane but which is not flammable.

On release, a part of the liquid immediately evaporates.A further amount is carried as droplets by the gas, alsopartly evaporates and the remainder drops to the groundand forms a pool. This forms a secondary source for furtheremission of gas. The size of the cloud is of interest as wellas its expansion velocity, the size and speed of the dropletsand finally the expansion, mass and temperature of the pool.

After preliminary tests with different containers, spher-ical vessels of 50 to 100 cm 3 were chosen which could bebroken with a hammer. Since in this case the glass explodedquicker than the liquid evaporated this gave a reproducibletest arrangement. Fig. 6 shows the test matrix of the firsttest series. The following results were obtained:

1 After a doubling of the initial volume the liquid dis-persed into droplets

2 The droplets moved radially and the droplet velocitywas independent of the size

50 or 100 ml flask

Hai

Thermocouple

Fig. 6: This apparatus is used for tests on the expan-sion and evaporation of liquids in the case of sudden dc-pressurization. The container is broken with a hammer.The evaporating liquid is visible due to a thin curtain ofsmoke. The pool of liquid is caught by the cooled dish of abalance and weighed. Quick acting thermocouples measurethe temperature of the pool.

3 During this initial expansion the gas front hardly mo-ved

4 The droplets moved through the curtain of smokewithout influencing it

5 After a certain time and after the droplets had passedthrough, the smoke curtain began to move. This wasdue to the sinking of the cloud formed from evapo-rated droplets

6 The pool formed immediately after release as the partof the liquid collected.

2.4 Thermal-hydraulics of Liquid MetalsAfter finishing the tests on natural convection in a sodiumcooled pin bundle at medium coolant temperatures (SO-NALCO-lExperimem), the NALO loop was rebuilt in 1989for higher coolant temperatures and flows. In a new re-search program the results ani experience obtained mainlyon liquid metals were transferred to the problems of theSpallation Neutron Source (SIN-Q) Project. Using ana-lytica: and experimental methods a contribution was madeto the thermodynamic design of the lead-bismuth targetsand the behaviour of various components of the spallationsource was studied. The results were also used as a contri-bution to the safety report.

Physical mechanisms were numerically simulated forthe SIN-Q work. The ASTEC code was used to modelthe relatively complex geometry and flow conditions. Thiscode was originally developed by the AEA in Dounrcayfor the analysis of natural convection in sodium cooled fastreactors but could be adapted for a wider use. The re-sults of the ASTEC calculations, under natural convectionconditions had already been validated with the help of theSONACO measurements. For calculating the cooling ofthe SIN-Q targets (lead-bismuth natural convection circuit)a first model was created which could take account of thecomplex window geometry, the proton beam heal source inthe target liquid and structures and the immersion cooler.The first results gave the expected flow profiles of a naturalconvection loop.

In order to reduce the uncertainties in the ASTEC cal-culations of the local temperature distributions an experi-mental verification was started using the TACOS (TargetCOoling Simulation) experiment. The first pan of the testprogram dealt with the cooling of the window (Fig. 7).Here it was possible to use the existing infrastructure of theNALO loop with sodium as coolant for the experiment. Forhealing the full scale window an electrical heating was usedinstead of lhe volumetric heating of the proton beam. Thetemperature distribution of the window and the flow stabil-ity was tested under symmetric and asymmetric convectioncooling by adjusting the coolant flow distributions.

At the same time as the thermal-hydraulic layout ofthe target, work was carried out on the DL,O tanks belowthe target. To study the influence of the thermal loads thetemperature distributions in the structure (walls) had to becalculated. The modelling with ASTEC is particularly com-

Fig. 7: The lower part of ihe SINQ target showing differentwindow geometries. The arrows show the flow directions.The coolabilily of the window is studied both analyticaly(ASTEC caclulations) and experimentally (TACOS experi-ment)

plcx since the lank is fitted with double walls and the cool-ing is achieved mainly with separated circuits outside andwithin the walls, and also since, in order to allow for thedetailed form of the neutron beam tube a complicated threedimensional structure is needed.

3 Materials Technology and NuclearProcesses

The Laboratory for Materials Technology and Nuclear Pro-cesses (LWV) covers the behaviour and the endurance ofconstruction materials in commercial light water reactorsand studies materials for (heir suitability for advanced reac-tor systems. In the fuel cycle the behaviour of standard fuelis examined and new forms of fuel and also the disposal ofradioactive materials is handled.

Special emphasis in the work of LWV in the programsand projects of the Department Nuclear Energy is given to

- Waste Management Program: leaching of waste ma-terial; interaction between rocks and radionuclidcs;the behaviour of colloids

- Program LWR Safely: Fracture mechanical studies ofsamples from reactor pressure vessels after neutron ir-radiation; fission product distributions in containmentin the case of a beyond-design-basis accident.

- Project High Temperature Reactor: Gas/metal inter-actions and endurance of high temperature materials

In the reporting year the following activities were of im-portance:

- Expansion and modernisation of the analytical equip-ment, in particular for analysis of water (anion chro-matography, organic carbon, suspended particles)

- Use of the experience gained on the side of nucleartechnology for non-nuclear needs in particular thecreation of a ceramic laboratory

- Specific use of third party funding by making re-search agreements with other organisations under thenew rules for research contracts of the board of (hetechnical colleges with HSK, NAGRA, the nuclearpower plants and industry

3.1 Structures and Fracture MechanicsThe spectrum of the materials studied ranges from ferriticand austcnitic steels for light water reactors (LWR) to highstrength nickel based alloys for the high temperature reactor(HTR).

In the framework of the Swiss contribution to the de-velopment of the HTR and the build up of knowledge onhigh temperature materials in general, the initiation of crackgrowth in a cracked impact test piece of IN800H under dy-namic loads (5m/s) was studied. Also tubes with transversecracks give problems of crack initiation. For the first time,in 1989, graphite having a coating of silicon carbide wastested in a helium circuit at 900 °Cin order to determine thestability of the material as well as its retention characteris-tics for radionuclides in impure helium. In the same circuitferritic steel material at a lower temperature of 550 °C wasexposed to a gas atmosphere with carburising properties.

Under a research contract the third irradiation series ofsamples from KKB II for the irradiation induced embrittle-ment of the reactor pressure vessel was evaluated.

3.2 Ceramics

In fabricating high performance ceramics the preparationof the powders plays a vital role. In contrast to the usualprocesses in which dry powders are mechanically mixedand milled, wet chemical processes have an advantage inthat they allow an even distribution of the components at anatomic level and the grain size is even and to some extentcontrollable. It was therefore appropriate to transfer theexperience gained from the sol-gel processes used in ma-king experimental nuclear fuel to the preparation of organicpowders generally.

In 1989, in the specially fitted out laboratory the fol-lowing subjects were handled:

- Microspheres of aluminium oxide and yttrium oxideof 30 - 100 //m diameter were prepared using thesol-gel method for the special task "Therapy" of theradio-pharmaceutical laboratory. These microsphereswere then activated by irradiation and at present eval-uations are proceeding for their possible use in cancertherapy.

- Tests to fabricate titanium oxide spheres gave porous

10

material but which for use as an ion exchanger un-der LWR conditions were insufficiently resistant toabbrasion.

- An extensive literature study on the manufacture ofpowder for super conductive materials showed thatin spite of a large number of cations such asBi2Ca :SrCu20 r it should be possible to make thestarting powders by a chemical route

3.3 Hot Laboratory

The Hot Laboratory is an integrated facility designed withhot cells, workshops, laboratories and analytical units forhandling highly concentrated radioactive materials. In fivehot cells activities of the order of 1015 Bq (Co-60) canbe handled, in particular activated steel samples, irradiatedfuel rods from reactors and highly active waste. Smallsamples of these materials with reduced activity can befurther analysed with modem shielded instruments. In aspecial part of the building, the pluionium wing, kilogramquantities of plutonium containing reactor fuel can be made.The research capacity serves not only the programs andprojects of the Department Nuclear Energy but also the PSIresearch as a whole:

- Fusion: Gamma-spectrometry and micro-fractrometryon irradiated samples from the PIREX project (Proton-Irradiation-Experiment)

- Spallation Neutron Source SINQ: collecting the firstdata on mercury evaporation from the lead-bismuthtarget

- Defect physics: determination of the impact tough-ness of samples from the STILO experiment (SteelIrradiation Loop)

- Combustion technology: Determination of the sizedistribution of aluminium oxide and soot particles;anion concentrations measurements on boiler conden-sate samples

- Electrochemistry: Characterisation of the structureof polymer coated electrodes using scanning electronmicroscopy

- Radio-pharmaceutical development: Analysis of elu-tions from strontium-rubidium generators; trace anal-ysis in iodine solutions

- Chemistry of cement: carbon determinations on ce-mented ion exchange resin wastes

Research agreements with industrial collaborators and theHSK were established. The most important tasks in 1989were:

- Post irradiation examination on fuel rods from thecladding corrosion program KKG

3000

2600 -

£ 2200

|<5

1800 -

1400

1000

L2 + cubic

cubic solid(U,Zr)O2

f cubic (U. Zr)O2

tetragonal (U. Zr)O2

cubic (U, Zr)O2

monoclinic (U, Zr)O2

. . . I . . . i 1 , , 1

_ _ ^

L2 + cubic

J

/ I tetra-/ gonal

tetragonal\

monoclinicf

monoclinic1 7*

0

UO,

20 40

mol-%

60

i

80 100

ZrO2

Fig. 8: Material from the damaged core of the reactor TMI-2 is characterised in the framework of an OECD-NEA pro-gram. The UO 2 fuel and the cladding material Zircaloy-4have taken up oxygen and inter reacted. Possibly the reac-tion products were molten. Following the phase diagram inquasi-binary system UO2-Zr02 a transformation occurs un-der 2600 K. Macroscopically a cubic, uranium-rich, mothermatrix and a tetragonal distorted zirconium phase is formedfrom the cubic (U, Zr)O2. The composition of these phaseswas determined with the shielded microprobe by micro-analysis of uranium, zirconium and oxygen and set in thephase diagram (o). From this it is shown that the thermody-namic equilibrium conditions at around 17S0 K must havebeen frozen in place by rapid cooling to room temperature.

Post irradiation examination on mixed oxide and ga-dolinium doped fuel rods from the international fuelsprograms TRIBULATION and GAIN.

Post irradiation examination on fuel rods with spherepac fuel

- Materials testing of pressure vessel samples and struc-tural materials from Swiss nuclear power plants

• Characterisation of samples from the core of a da-maged US pressurised water reactor (TM1-2) (Fig.8).

11

3.4 Nuclear Fuel Project

For many years the project has been involved with develop-ing a new form of nuclear fuel in the form of microsphcresproduced by wet gelation techniques and to test these inreactors in suitably designed fuel rods. The cycle of fueldevelopment, includes a scries of activities and can extendover several years ie.:

Evaluation and studies - development work - man-ufacture of fuel rods - irradiation in reactor - postirradiation examination - evaluation of the results.

In 1989 work covered most of these areas of activity, mainlyfor the light water reactor (LWR) but in parallel also for theFast Breeder Reactor (FBR).Work for the fast breeder reactor:

- In the frame of a collaboration USA-PSI, 25 fuel pinswith uranium-plutonium carbide microsphcre fuel havebeen irradiated since 1986 in the Fast Flux Test Facil-ity (FFTF) in Richland. In October 1988 the target of620 full power days (85 000 MWd/t) was achieved.In 1989 the planning of the post irradiation examina-tion (PIE) with three national laboratories in the USAwas negotiated.

- A collaboration with the CEA-Cadarache aims at de-veloping uranium plutonium nitride fuel. In PSI twotypes of uranium nitride particles were fabricated;dense microspheres for direct use in fuel pins andporous microsphcres which are to be developed aspellets and pressed by the CEA. For both productsthe process is still being optimised.

The preparations for the use of uranium plutonium oxide(MOX) in light water reactors were continued:

- Four fuel rods each with three segments containinguranium dioxide sphere-pac fuel have been under ir-radiation since 1986. The non-destructive examina-tion of segments after an irradiation of from one tothree years show only slight differences to segmentscontaining pellet fuels. The cross section of a sphcrc-pac segment after one cycle of irradiation, 14 GWd/tburn up shows little restructuring (Fig. 9). A fourthand last cycle is now under way.

- Over several years the pressing of uranium oxide mi-crosphcrcs to pellets has been studied with Sicmcns-KWU. The procurement of a pellet press in 1989 andits commissioning in 1990 will allow PSI to furtherdevelop on its own an important concept of producingMOX pellets from sol-gel particles.

- In 1989 three size fractions of fuel were manufacturedand the end- plug welding procedures introduced inpreparation for a joint MOX irradiation experimentwith Bclgonuclcaire.

These developments and future needs required that a rangeof improvements be made or developed for equipment andprocesses, for example:

Fig. 9: Cross section of a segment with UO2 microsphercsafter one irradiation cycle in KKG. In the center a part ofthe fine fraction has lightly sintered together.

- In using sphere-pac fuel the question invariably arises-can in the case of a defective fuel pin, fuel paniclesbe lost to the coolant? Laboratory experiments undersimulated conditions of the start up phase of an ir-radiation (beginning of life) with unsintercd particlesshow that a partly wetted zone forms just above thecladding crack and that less than 4 g of material islost. It is expected that this loss will be even smallerfor partly sintered material in reactor.

For the use of pellet or sphere-pac fuel in a reactor theprediction of the fuel behaviour is necessary. The code 1N-TERPIN/FRPS for pellet fuels was obtained from Studsvik,validated and used first for an evaluation of results from theinternational fuels study program "High Burn-up Effects".

To cover the needs of PSI, the utilities, and the safetyauthorities when modelling fuel behaviour a comparisonwas made in the search for a suitable code between four in-ternationally known codes (ENIGMA from the UK, STAV-Swcdcn, TRANSURANUS-BRD/EC and COMETHE-Bcl-gium). The capabilities and limitations of these codes wereevaluated at PSI and comparative calculations are now un-der way with the aim of selecting the most suitable codeearly in 1990.

3.5 LWR Contamination Control Project

Reactor circuits can become contaminated due to the de-position of radioactive material, mainly cobalt 60. In par-ticular in the case of Boiling Water Reactors (BWR) with

12

1986 1990

Fig. 10: Activity deposition in the new circulating pipeworkof the MUhlebcrg power plant.1 Inlet 2 Outlet

external circulation pipework this leads to an increased ra-diation exposure of the operating staff during maintenancework. The transport of activity and the deposition is de-pendent on several factors, especially choice of material,water chemistry and method of reactor operation. In orderto study this problem with the aim of uncovering the basicmechanisms involved and to make recommendations for aspecific plant a comprehensive evaluation is necessary. Theprogram of work therefore is divided into several interact-ing parts:

- To follow the international scientific and technicaldevelopments the problem of the BWR is being in-dependently studied in the framework of the IAEAprogram WACOLFN1 in which 15 countries arc tak-ing part.

- The on site nuclide specific measurements on thecirculation systems of the Miihlcbcrg (KKM) andLeibstadt (KKL) reactors was continued in 1989 incollaboration with Studsvik Encrgitcknik AB, Swe-den. It is interesting to note that the activity increasein the circulating pipework of KKM which was re-placed in 1985 has now stabilised (Fig. 10).

- Laboratory tests in autoclaves under static BWRconditions serve to provide basic data for the mathe-matical modelling. In 1989, the exchange capacity oftitanium oxide as an inorganic ion exchanger whichis stable under BWR conditions was studied. Decon-tamination factors of more than 1 000 were measuredfor Co-58, Zn-65 and Sb-124, whereas for Sr-85 andAg-110 values under 10 were obtained.

- The pilot circuit for simulating the dynamic condi-tions in the BWR circulating systems operated underdesign conditions for 5 months (90 bar, 290 f'C). Sev-eral weak points (seals, flow meters, oxygen analysis,thermocouples) had to be repaired. It was found that

the use of absorblion and ultraviolet irradiation wasnot sufficient to remove organic impurities from thefeed water whereas ionisalion was more successful.

- A large effort was devoted to advance methods ofwater analysis:

- Measurement of dissolved organic carbon (DOC)in water with different oxidation methods.

Anion chromalography for determining the sili-cates, chlorides, carbonates and sulphates in wa-ter.

- In order to determine the suspended particle sizeand distribution by reflected light a measuringprobe with a saphir window was built and ameasuring circuit with three filters. Both instru-ments arc to be used under reactor conditions.

- The decontamination process VS (Very Soft) devel-oped in PS1 using an oxidation phase with chromicacid and an reduction stage with oxalic acid was suc-cessfully used by ABB to decontaminate three pri-mary pumps in the Miihlhcim-Kiirlich nuclear plantand two pumps in the Biblis plant. Here decontami-nation factors of over 100 were measured.

In view of the importance of this project especially forthe radiation protection, the HSK stated that again in thecoming year it is ready to provide financial support TheBWR operators also provide active support for the on-sitemeasurements.

1WACOUN: Investigations on Water Chemistry Control and CoolantInteraction with Fuel and Primary Circuit Materials in Water Coold PowerReactors

4 Waste Management Program

The safe disposal of radioactive waste from reactors, medi-cine, industry and research in suitable final repositories isworld-wide an environmental protection task of the highestpriority. In order to prove the long term safely of such fi-nal rcpostitorics models are needed describing the relevantmechanisms and processes as well as reliable data lo allowthe likely behaviour of the repository in a quantitative man-ner. PSI in close collaboration with NAGRA carries outresearch on selected topics and thus contributes to solvingsome of the questions of waste management. Decisive forthis in view of the range and the multi-disciplinary natureof the problems is the involvement in national and interna-tional activities. Thus close contacts were maintained withnational and foreign institutes and intensive collaborationcontinued in several international studies.

The work at PSI covers a large part or the chain of nu-clidc release, transport through the technical barriers of therepository and transport through the gcosphcrc and the bio-sphere leading to a dose calculation. The main efron lies indeepening the understanding of the mechanisms involved,the collection of high quality data and, based on carefullyselected experiments, the validation of models.

13

4.1 Characterisation of Waste BodiesDuring 1989 the properties of the three waste body types,glass, bitumen and cement were studied.

The tests on inhibiting glass corrosion with lead orcadmium ions continued. In a short term experiment (28days) the corrosion rate of the French glass SON-68 wasfound lo be ten times smaller as without inhibitor addition.The same effect was found in the more reactive Britishglass MW. The decisive question whether with inhibitors asimilar reduction of long term corrosion rales can also beobtained is to be answered with experiments with a runningtime of a year.

For the safety analysis of a final repository a modelis needed, which describes the time dependent nuclide re-lease from the glass. The attempt to use an existing modelto describe corrosion experiments showed some basic de-ficiencies in the conceptual approach. A related documen-tation is to be used as the starting point for developing anadvanced model which takes into account the relationshipbetween the corrosion rate and the removal of the corrosionproducts.

Medium active waste from reprocessing, and also a partof the waste from reactor operation will be fixed in bitu-men. Last year the main interest was devoted lo the ques-tion whether the radiological decay products of the bodycould influence the transport of nuclides by the formationof stable complexes. The degradation products are formedfrom the irradiation by external -/-irradiation. The analyti-cal characterisation of these and the study of the behaviourof complex formation compared to Cu3+, Ni2 + and UO2+2

lead to the conclusion that only oxalate must be consid-ered as a possible ligand (Fig. 11). Spoliation calculationsshowed that the influence of oxalate under the conditionsin the repository can be neglected. This finding is of majorimportance for a safety analysis also because it simplifiesthe sysicm to be studied. This work is now completed.

The experimental investigations of the reverse absorp-tion released nuclidcs on the bitumen body is hindered by

Experimentscalculated

No complexation5ppm OxaEate

, < • ? • •

3.0 3.2 3.4 3.6 3.8 4.0 4.2 4.4 4.6 4.8 5 0

- log [Cu] ,„,.,,

Fig. 11: Complexing of copper by decay products of bitu-men . The Pff of the solution is 6 and ionic strength 0.1mol/L. The titrauon curve is also shown with an oxalate (Sppm) and a calculated curve for determining the complexformation constants.

source reactions. The surface concentration of the absorb-ing groups could be determined as 2 to 3 /imol/m2 on fineparticles of a bitumen suspension. The stability of selectednuclides of the surface complexes for selected nuclides isnow to be determined.

The study of the cement bodies of operating wastes in-cludes the continuation of the long term tests on the leach-ing of plates (tiles). The build up of an apparatus for con-tinuous cement mixing was completed. The method for de-termining safety related parameters was further developed.In future these will continue under contract to the powerplants not as a part of a research program.

4.2 Behaviour of a Final Repository

The aim of this study is to evaluate the long term behaviourof the repository components and to describe the releaseof nuclides into the geosphere. In 1989 theoretical andexperimental work on cement was in the forefront. Cementis used in part in stores for highly active wastes and in storesfor medium and low level waste it forms a major technicalbarrier against release.

The experimental program conceived in 1988 to studythe aging of cement and the transport of radio-nuclidesthrough cement barriers was started. The erection of theinfrastructure (boxes giving a controlled gas atmosphere,diffusion and degradation cells) and the reproducible pro-duction of samples with sufficiently high water permeabilityrequired more lime than foreseen. In the meantime mea-surements of the hydraulic conductivity in relation to thehardening lime have been obtained and also the first dif-fusion measurements wiih halogens (36C1, I25I). Diffusionmeasurements with further nuclides, accompanied by sorp-tion measurements on broken cement can now go ahead.These investigations will extend over a long period sincethe influence of the artificial aging of cement (eg. the in-creasing leach out) on its transport properties is of centralinterest.

The earlier developed cement degradation model wasused in 1989. Of the problems examined, the water chem-istry of the pores is more dependent on the ground waterchemistry than on the initial cement composition. A goodexample is the time dependent change in the solubility ofuranium in the water-filled pores of the cement. This isstrongly dependent on the choice of the solubility limitingsolid phase, in which the difference in solubility can be aslarge as ten orders of magnitude. The uncertainly in thethermodynamic data is often very large. For this reasonlarge efforts have been undertaken to evaluate the complexforming constants and to obtain a consistent basis for thesafety analysis. Specifically the data on selenium, nickeland paladium as well as the solid phase silicon oxide werecritically examined.

In the repository surroundings strongly changing con-centrations of materials in solution depending on time andlocation can occur. For this reason a reliable descriptionneeds a coupling between the physical and chemical spe-ciation. Instead of the formulation of complex non-lineardifferential equations one can apply the local conditions

and using the approach of veclorisation one can solve theproblem of coupled processes with the aid of the methodof cellular automates. A first study served to formulate theproblem in detail and to show the potential of this mod-em method of problem solving. Further the collaborationcontinued in the international program CHEMVAL.

4.3 Work in Connection with the RepositorySurroundings

The distant surroundings of the final repository include thegcosphcre and the biosphere. The problem can be definedas describing the transport behaviour of radionuclidesfrom the repository to the population including the doseexposure. Laboratory and field experiments and the relatedmodelling form the main thrust of the program: informationon the transport mechanisms is obtained by experimentsover small and large distances on various natural systemsand model concepts tested whose influence on the safetyanalysis can then be investigated.

In the framework of the migration tests in the Grim-sel rock laboratory of NAGRA the first tests covered theabsorbing tracer -MNa and helium in a dipolc experiment.Helium behaves as the non absorbing tracers fi-Br and flu-orosccne. -'1Na was used at concentrations far under thenatural sodium concentration and thus pure isotopic ex-change was expected as the sortition mechanism. It wasfound that the retardation is much smaller than would beexpected from the laboratory tests. The discrepancy is at-tributed to the high water velocities and diffusion kinetics.This interpretation must still be checked with model calcu-lations. A special problem is the influence of the chemistryof the injection water. Under these conditions an injectionwater must be used which is slightly different to the waterin the clufts. Then the breakthrough of the material in thewater can be measured. A first such experiment has pro-vided information over the transport behaviour of fluoride,chloride, sodium, calcium, strontium and magnesium.

The field trials in the rock laboratory were accompaniedby a comprehensive laboratory program. The opportunityto measure sorption under realistic conditions greatly wasextended in 1989 with the installation of box lines with acontrolled gas atmosphere. The oxygen and carbon dioxidecontent can be selected for the specific field conditions. Thework on measuring the cation exchange capacity of Grim-sel mylonite and the sorption of sodium ("Na), strontium(85Sr), and caesium (137Cs) are now completed and inter-preted with an ion exchange model. Also the dependence ofthe equilibrium distribution constants on the particle size ofthe mylonite was investigated. The ™Na and 85Sr in a con-centration below that of the natural Grimsel water showedthe expected linear isotherm, whereas the sorplion of 137Cscould be described by a Freundlich-isothcrm.

In dynamic drill core infiltration tests using materialfrom the migration fault the transport of 24Na and 82Brwas studied and a first modelling of the breakout curve wascarried out. The results obtained are in agreement withthe values obtained from other independent experiments.The design of the pressure infiltration cell when working

with broken material required some modification in orderto allow a proper modelling.

Concerning a final repository for low and medium levelwaste the main effort moved to the sorption of crystallinerock on clay. Before carrying out sorplion and diffusionmeasurements the composition of a potential stable claywater must be determined. The first results have alreadybeen obtained.

In the frame of a partial program on colloidal chem-istry the possible influence of natural colloids on nuclidesmust be evaluated. For this a knowledge of typical col-loidal concentrations in deep water is necessary. In orderto estimate the nuclide absorption the colloid material mustalso be identified.

The studies now completed on the colloids in the waterof a migration fault in the Grimsel laboratory had as oneaim the testing of methods for sampling and characterisingthe colloids. In order now to study their sorption behaviouradditional electro-analytical methods with higher sensitivityare being introduced. In this way the phase separation canbe avoided which causes problems in the case of smallcolloid concentrations.

101b5 to" io'

Time, aFig. 12: Nuclide flow for 135Cs at a migration distanceof 500 m normalised to the water flow through a finalrepository as a function of time after a failure of a con-tainer. The influence of the natural caesium concentrationin ground water Cmin [mol/L] for the calculation usingthe Freundlich-sorplion isotherm is shown. Also shown forcomparison is a calculation with a linear isotherm (Kd=0.03m3/kg).

15

A drilling into the mineral source Eglisau offered theopportunity to isolate the colloids from the water of thelower fresh water mollases(a possible Wirtgcstcin for a fi-nal repository for high active wastes). In the range of 40to 450 nm the water contained around 0.3 mg of colloidalsubstances per liter. The cation exchange capacity of thecolloidal fraction in the water was 10-6 mol/1. Here the con-centration of colloids is ten limes larger than in the Grimselwater.

For the modelling of the nuclide absorption on colloidsexisting models can be used. There arc however still uncer-tainties concerning the heavy metal absorption on layeredsilicates. Basic studies to identify the different absorptionmechanisms arc being started.

In the area of modelling the geosphere transport thework was completed on the influence of a non linear ab-sorption isotherm for transport in a medium with a dualporosity with a sensitivity study. Here it was found that itis very important to know the natural concentration of thestable isotope of an element for the transport of the radio-isotope (Fig. 12). For the water carrying zone a gap andan artery geometry was investigated. It was found that theresults for the gap geometry is more sensitive to parametervariations. A variation of the model with pure diffusionwas used lo interpret English diffusion measurements insandstone. Further, work was done in the international IN-TRAVAL study. The participants again found great interestin the infiltration experiments mentioned above.

The work on the biosphere transport was mainly a partof the now completed international BIOMOV study. Theinvestigation of the transport from saturated ground in a un-saturatcd root zone lying above was completed. A transferto the biosphere region of northern Switzerland showed lhatthe doses from the food chains are heavily reduced whenthe capilliary transport arc taken into account.

5 LWR Safety Program

Every country which operates nuclear facilities is forcedto become actively engaged in the international research onthe real existing safely margins of its plants to maintain andexpand its scientific and technical knowledge in this areaand to contribute to raising the actual safety of its reactors.For this reason safety research was identified by the SwissFederal Cabinet, in the terms of reference for PSI, along-side waste disposal, as one of the main topics in the areaof'Nuclear Energy Research" . Also Switzerland has to bein a position to judge questions of safety arising from theoperation of its own power stations using its own expertise.Taking account of its size and effort available, Switzerlandmust be an equal partner with other international researchinstitutes in all important areas. PSI is the only institute inSwitzerland which carries out reactor safely research.

A major part of the work is the analysis of possible evo-lution of beyond design basis accidents having a very lowfrequency of initiation. Here the calculation of the releaseof radioactive material (source term) to the surroundingsof the reactor experiencing this extremely unlikely accident

stands in the forefront. From the knowledge thus gained ad-ministrative and technical measures can be prepared whichcan reduce the impact of a reactor core melting or which canbring the accident under control before significant releaseoccurs.

An important aim of the LWR Safety Program at PSI isto carry out own research on the source term determina-tion and thus where relevant, to gain support from foreignwork. Here accident evolution in existing plants must bedescribed dcterministically from the chemical, mechanical,thermodynamic, metallic and acrophysical points of view.This work is accompanied by experiments which are to helpto understand questions which arc still open.

Basic studies in thermodynamics and hydrogen distribu-tions and combustion in the plant buildings of a power reac-tor also belongs to the program item "source term" as wellas studies on the behaviour of the melting reactor core andthe interaction of the core melt with the concrete founda-tions. The behaviour of iodine, gases and aerosol particlesare particularly important. Experiments have been carriedout to answer in particular (he following questions to assistin the use of computer codes: How much elementary io-dine and how many soluble and insoluble aerosol particlesare retained when carried in a flow of steam through a wa-ter filled vessel? How many aerosol particles deposited onwalls can become resuspended? How do iodine and otherfission products behave in the water of building sumps asthese dry out? How do airborne fission products behaveduring a hydrogen fire?

Finally the LWR Safely Program participates in jointinternational experiments which serve the source term anal-ysis. The research on the behaviour of iodine during severeaccidents must be particularly emphasised. In 1989 the pro-gram IMPAIR2/M was developed and opened for criticismamongst international experts. Iodine is so important inthe source term analysis because its isotope 1-131 formsthe largest part of the risk potential external to the nuclearplant in a core melt accident and because iodine is veryreactive and can occur not only in gaseous but also in par-ticle and in soluble form. The results of the source termanalyses were systematically applied to one of the Swissreactors in 1989.

The aim of the program point component safety is theimprovement of the knowledge on the remaining enduranceof components of the Swiss reactors which arc importantfor safety and which can only be replaced with enormouseffort and cost. The current view is lhat these are the re-actor pressure vessel (RPV) the directly connected pipingwith their supports and the internals of the pressure vessel(Fig. 13). A central task is to collect data and improvemethods in order to predict whether and for how long thesecomponents can remain in operation with sufficient confi-dence perhaps exceeding the planned operation life. Thesafety margin of a component is estimated from the dif-ference between the maximum duty (appropriate loadingconditions for the component) and the carrying capacity ofthe component for the given type of loading (Fig. 14). Ifa component is starting operation one assumes lhat it pos-sesses a certain operational reserve capacity for the normal

16

Internals

Fig. 13: Components which are important for the safety ofa nuclear power plant and which can only be replaced withenormous technical and financial effort.

r—Starting operation

II

III IV

• Normal duty

i i i i ii i i i iLTJJJ

Loading Load capacity

Fig. 14: Reduction in the safely margins of a componentwith time as a result of increased operational demands andmaterial quality reducing influences.

duty (safety margin I) as well as for certain off normaloperational conditions or events (safety margin II). Withincreasing service life the material properties of the com-

ponent change and in general the capacity decreases. Bydefinition the safety margin becomes smaller (safety marginIII). With time the operational conditions or (he demandson the component can change thus reducing the safety mar-gin further (reduction of the safety margin IV and/or V).For example the loads and the environmental influencescan depart from those originally assumed or additional andnot previously accounted for transients can lead to higherstresses than assumed in the licensing procedures.

All activities are part of a worldwide program of workon questions of neutron embriulement and damage to LWRmaterials, crack growth and detection under given condi-tions. In particular pressure vessels, supports and pipingcontaining cracks are analysed under static, steadily in-creasing, cyclic and also dynamic mechanical and thermalloads. Further experimental cracks are locally studied us-ing an acoustic emmission instrument developed in house.Preliminary studies arc being made on possible neutron ir-radiation experiments.

The Program LWR-Safety is participating in investiga-tions being carried out on a prc-dcfcctcd pipe containingan internal circumferential crack, under blow-down condi-tions. The geometry, material as well as the orientationand the size of the defect for the pipe to be tested wasestablished. An internal circumferential crack with a cir-cumferential length equivalent to 60° and a crack depth ofa third of the wall thickness produced by an initial artifi-cial notch followed by cyclic loading, will now be studied.The circumferential notch which has been chosen is espe-cially interesting since it is relatively short: short cracks canfirstly be easily missed in routine examinations of operat-ing piping, and piping having short circumferential crackscan build up a large amount of deformation energy just be-fore failure, increasing the likelihood of a pipe fail' re. If apipe break is to be allowed for reasons of obtaining valu-able results on the leak-before-break behaviour, additionalmeasures have to be taken for the building safety.

In 1989 the numerical tool was prepared in order todetermine stresses in components during welding or othercases were high temperatures can occur. The calculationswere carried out separately e.g. there were no interac-tions assumed between the calculations of the heat source,the temperature and the stress/strains. The main part ofthe work consisted of the implementation of an extendedthermoplastic material model in the finite element programSOLV1A, which, also coupled with a temperature rise (forsteel from 800 °C) accounts for the rapid loss of strengthand therefore allows the input of deformations approach-ing the melting point. This material model was validatedby verifying the results of an experiment in which a fuelelement channel under constant inner pressure and steadilyincreasing temperature was brought to the point of collapse

As the main effort on the surveillance of componentsin 1989, the final report on the acoustic emmision mea-surements was published; here an artificial crack was intro-duced in a pressure vessel experiencing various operationalconditions and observed. The quantitative analysis of themeasurements was based on a source model and a litera-ture study. A single event of a crack development and the

17

repeating events in the unloading phase during the cyclicloading could be clearly attributed to a friction process,possibly ratcheting; friction processes also appear to be thecause of repeating events during the loading phase.

Since a LWR is a thermal machine having water andsteam circuits, applied thermal-hydraulics and the studyof single phenomena in multi-phase flows with differingheat transfer conditions plays a leading role. In addition tointroducing young scientific staff to the large existing com-puter codes, which is impossible without in-house modeldevelopment backed up by the related experiments, theLWR Safety Program intends more and more to replacethe large existing codes systems having conservative sim-plified models with more smaller best estimate codes for

special problems. Further (hermal-hydraulic methods willnot only be applied to loss of coolant accidents and tran-sient accidents but also to beyond-design-basis accidents.Thermal-hydraulics has to provide decisive contributionsto the measures for the management of severe accidentsmentioned above. Thermal-hydraulic studies needs, along-side a good theoretical basis, a sound understanding of thespecific plant. For this reason a direct collaboration withthe plant operators as in the case of the source term analysisis very important. In the foreground is the development offurther models in computer codes in order to calculate thecomplex flow patterns in plant pipework in loss of coolantaccidents. This topic is also dealt with in international col-laborations.

18

Best-estimate Analysis of Beyond Design-Basis Control Rod Ejection Accidents inthe Beznau II Pressurized Water Reactor

E. Knoglinger, L.A. Belblidia, P. Grimm, M.A. Zimmermann

G. Abu-Zaicd, A. Galpcrin, Ph. Jacqucmoud, J.M. Kallfclz, R. Mylonas, J.M. Paratte, D. Saphier, W. Schippers,G.M. Sigut, G. Skoff, P. Zehnder

Laboratory for Reactor Physics and Systems Technology

Abstract

This report presents results of a three-dimensional analysisof beyond-design basis rod-ejection accidents in the Bez-nau II PWR obtained within the framework of the STARSproject. The analysis is performed with the three dimen-sional code QUABOX/CUBBOX-HYCA (QCH) which com-bines a coarse-mesh neutron flux expansion method with aparallel-channel model for the thermal-hydraulics. A seriesof static calculations is performed to locate the most re-active rod and to qualify QCH by comparing calculationalresults with measured data. The study presents a case athot zero power and analyzes in detail the spatial powerand temperature distributions in the core. A hot-channelanalysis with FRAP-T and a pressure increase calculationwith RETRAN show that the parameters of concern in thistransient are well below the safety limits.

1.1 Description of the Accident

The design-basis REA is defined as the mechanical failureof a control rod drive mechanism pressure housing such thatthe reactor coolant system pressure would eject the con-trol rod mechanism and drive shaft to its fully withdrawnposition. This failure results in a rapid reactivity insertion(within approximately 100 ms) with large local power peak-ings. The associated power excursion is limited by Dopplcrreactivity feedback due to increased fuel temperatures, andthe transient is ultimately terminated by the reactor protec-tion system. It must be shown under these circumstancesthat the energy released in this accident does not lead tocatastrophic fuel damage and that limits specified by thesafety authorities are not violated.

The probability of this accident is closely lied to theprobability of failure of a control rod drive mechanism

1 Introduction

In July 1988 the Paul Scherrer Institute and Uic SwissFederal Nuclear Safety Inspectorate (HSK) entered into anagreement on the development of simulation models for thetransient analysis of the reactors in Switzerland (STARS).Under this program PSI is presently engaged in generatingsimulation models for all Swiss nuclear power plants, andin compiling, developing, and continuously updating thenuclear and plant-specific data bases required for steady-state, transient, and accident analyses of these plants [1],The first phase of this project includes the analysis of hy-pothetical "rod-ejection accidents" (REA) in the Beznau IIpressurized water reactor, and falls within the framework ofthe increased attention since Tchernobyl given to reactivityinitiated events. Fig. I shows a block diagram of the codepackage used for this work.

OIMBOXICUBBOX

THERMIT

FF1«PT« STr»n»l« at

—

' 1WssssCorvVil* |RETRANRELAP

TRAC-BF

Fig. 1: STARS transient analysis code package

19

housing or nozzle [2]. This probability is low (between10~4 and 10"6 per reactor-year) [3], but the potentially se-vere consequences of the accident lias prompted the US Nu-clear Regulatory Commission to issue a Regulatory Guide[41 specifying the analysis requirements.

The accidents investigated in this study have an evenlower probability of occurence, since they also assume vi-olation of the Technical Spcciricaiions due to an operatorerror or malfunction of the safely system, leading to starlingconditions beyond those addressed in the safety analysis re-port. This results in reactivity insertions in excess of thoseassumed in the design-basis event. The accidents analyzedarc therefore beyond-design basis accidents..

1.2 Protection Against the Accident

Mechanical and nuclear features preclude the possibility ofoccurrence of such an accident and limit its consequencesif it were lo occur. These include conservative mechanicaldesign, proper material selection, fabrication, and qualitycontrol, and a nuclear design which minimizes the ejectedreactivity worth. Each housing of the control rod drivemechanism is part of the reactor coolant system and is de-signed and manufactured to criteria in Section HI of theASME Boiler and Pressure Vessel Code for Class 1 com-ponents, and is tested before and after attachment to thereactor vessel head at pressures in excess of the maximumdesign pressure.

If the accident were to occur, its severity is limited bythe nuclear design which controls the worth and positionof the control rods. During normal operation at full power,only one bank is inserted but restricted to the top of thecore, thereby minimizing the ejected rod worth. Reactivitychanges due to xenon transients, fuel depletion, and temper-ature changes are compensated by borated water. At lowerpower, larger control rod insertions are permitted because(for a given reactivity insertion) the REA is less severe atlow powers.

The Technical Specifications contain power-dependentrod insertion limits based on the REA that guarantee com-pliance with local power density limits, shutdown marginrequirements, and limits on consequences contained in theSafety Analysis Report. The position of each control rod iscontinously displayed in the control room and an alarm willgo off if a control bank approaches its insertion limit or ifa rod is misaligned. Operating procedures require borationof the core when the insertion limit is reached. Ultimately,period, high neutron flux, or pressure signals cause the re-actor protection system to initiate a scram to shut down thereactor.

1.3 Limit Criteria

There are three types of criteria associated with the REAcontained in the regulations. The purpose of the first typeis to prevent catastrophic fuel failure. Several studies of thethreshold of fuel failure and conversion of thermal energyto mechanical energy have been conducted in the US andin Japan [5,6]. The threshold for conversion of thermal en-

ergy to mechanical energy which could lead to catastrophicfailure (large fuel dispersal and pressure rise) was reportedat 300 cal/g . It was also shown that clad failure can occurin unirradiated fuel at 210-220 cal/g due lo brittle fracturecaused by severe oxidation. This limit is even lower forpreprcssurized irradiated fuel or if the fuel has been sub-jected to excessive cycling or has become waterlogged.

As a consequence of the above experiments, the reactorvendor has adopted the following conservative criteria loensure that fuel dispersal in the coolant and gross latticedistortion will not occur [7.81:

• Average fuel pellet enthalpy at hot spot below 225cal/g for unirradiated fuel and 200 cal/g for irradiatedfuel.

• Peak clad temperature below ~ 1500 °C.

• Fuel melting less than 10% of the fuel volume at thehot spot.

In order to show compliance with these criteria, fuel en-thalpy and temperature distributions must be determined.The calculations involve use of computer programs thatmodel the reactor physics, thermal hydraulics, and fuel be-havior in the core, and should be performed for differentpower and burnup conditions.

The second type of limits insures the integrity of the re-actor coolant system (RCS) boundary by limiting the max-imum reactor pressure during any portion of the assumedtransient to less than the value that will cause stresses toexceed the Emergency Condition stress limits as definedin Section III of the ASME Boiler and Pressure VesselCode. Calculation of the RCS pressure must therefore beperformed with a code that simulates the whole plant, andthe peak RCS pressure must be shown, for the case con-sidered in this study, to never surpass 22 MPa (FSAR limitfor Beznau II [8]).

Finally, the last limit deals with the radiological conse-quences if fuel damage occurs. Table 1 shows NRC lim-its and Westinghouse design criteria. The latter are usedthroughout this study.

Table 1: NRC and Westinghouse Limit Criteria

Avg. Fuel EnthalpyClad TemperatureMelt VolumeNo. of rods at DNBSystem PressureOffsite Dose

LimitNRC

280 cal/g--

ASME Code10CFR100

ValueW

200-225 cal/g1480 °C< 10%< 10%22 MPa

-

2 Goal of the Analysis

The purpose of this study is lo investigate potential coredamage as a result of a beyond design-basis control rod

20

ejection accident starting from initial conditions not ex-plicilcly addressed in the safety analysis report of the plantand necessitating violation of the Technical Specificationsand operator/system malfunction during power operation.Such conditions are:

• All control rods except the safety rods beyond theircorresponding insertion limits fully inserted in thecore at hot-zcro-power conditions.

• The most reactive rod completely inserted into thecore prior to us ejection at full power conditions.

The analysis was done for hot-zero power (HZP) con-ditions at beginning-of-cycle 16 (BOC) and end-of-cyele16 (EOC), and for full-power conditions (HFP) at BOC.However, due to lack of space, only HZP, BOC dynamicsresults will be reported here. More detail is given in Ref.[9].

This study does not address the radiological consequencesof the accident. Only fuel damage and pressure responseare investigated.

3 Description of the Reactor and Cy-cle Analyzed

The Bcznau II reactor is a pressurized water reactor with arated thermal power of 1130 MW. The core consists of 121fuel assemblies each consisting of a 14 X 14 array. The ini-tial heavy metal mass is approximately 39 t, and the activecore height is 3.05 m. Due to the relatively low specificpower, the fuel assemblies are normally irradiated duringfour one-year cycles in contrast lo other PWRs in which thefuel resides for three years. Core reactivity control is pro-vided by a combination of neutron absorbing control rodsand a soluble chemical shim (boric acid). The control rodsconsist of 25 silver-indium-cadmium control rod clusters,and are divided into two control banks (A and B) and asafety bank (S). The fuel cycle analyzed lasted approxi-mately 11 months and accumulated a core-average burnupof 9159 MWd/t. Initial enrichments ranged between 2.8and 3.6%.

4 Calculational Methods

The standard tool for the REA analysis is point kinetics suchas in the RETRAN [10] code. However, it is well knownthat three-dimensional (3D) spatial effects in the core can bequite significant in the analysis of reactor transients ([II]-[13]). The two key parameters in the analysis of a REAwhich determine the magnitude of the peak power and fueltemperature arc the rod worth and the Dopplcr feedback.Since the largest fuel temperature rise occurs in the core re-gion with the highest power, Dopplcr feedback is generallyunderestimated if the power shape distortions arc not takeninto account. This is the case in standard point-kinetics cal-culations which results in conservatively high rises in powerand temperatures. Proper accounting of three-dimensional

spatial effects is, therefore, very important for best-estimatecalculations. In the present study the 3D transient multi-group diffusion codeQUABOX/CUBBOX-HYCA [14] wasused in combination with the transient fuel pin code FRAP-T6 [15], to predict fuel temperature, fuel enthalpy, and crit-ical heat flux ratio in the core.

Both point-kinetics and 3D dynamics require physicsdata that have lo be generated in steady-state physics cal-culations. In the case of QCH, the input data are fuel assem-bly averaged 2-group constants and the macroscopic burnupdistributions in the core at the beginning of the transient.The group constants are functions of bumup, fuel and mod-erator temperatures, boron concentration in the. coolant, andthe presence of control for each fuel assembly. In the caseof RETRAN, radial peaking factors, control rod worths,and reactivity coefficients for each accident scenario haveto be calculated as a function of burnup, power, and controlrod positions. The corresponding calculations were per-formed using the EIR LWR Code System (ELCOS) de-veloped at PSI. The LWR code system ELCOS consistsof four codes: ETOBOX, BOXER, CORCOD, and SIL-WER. ETOBOX processes cross section data in ENDF/Bformat and produces a cross section library for BOXER.BOXER performs cell and two-dimensional transport anddepletion calculations. CORCOD computes fits to the ho-mogenized few group cross sections from BOXER for thethree dimensional calculations. S1LWER carries out staticthree-dimensional neutronics and thermal-hydraulics calcu-lations. A more detailed description of this system is givenin Refs. [16] through [18].

QCH combines a coarsc-mesh neutron flux expansionmethod with a parallel channel model for the core thermalhydraulics. The basis of the coarse-mesh method is thespatially integrated mulligroup neutron diffusion equation

dt Jv, Jv,M'dV,j = , M (1)

where 4< = W>,C)T, and <f> and C represent the neutron fluxesand delayed neutron precursor concentrations, respectively.The domains of integration Vj correspond lo rectangularnode volumes. These equations are supplemented by conti-nuity conditions across interfaces between adjacent boxes,and by relations describing the dependence of the nuclearcross sections on node conditions. Spatial coupling is re-stricted to nearest neighbors. A frequency transformationoption, which factors oul the flux into a strongly time-dependent amplitude function and a slowly lime-dependentshape function, reduces truncation errors and allows use ofreasonably large time steps.

From a thermal-hydraulic point of view, the core is de-scribed by a set of parallel channels, and in each channelthe heat transfer to the coolant is represented by an aver-age fuel pin model. The coolant model consists of mass,energy, and momemtum equations in one-dimensional ax-ial geometry. The Tuel heat conduction model allows anynumber of radial meshes in the pin and one radial zone forthe cladding. Fuel conductivity is temperature dependent,and the gap and clad conductivities are assumed to be con-stant. Axial heat conduction is neglected. Details of ihc

21

theory can be found in References [19] through [21].The pressure rise during the transient was calculated us-

ing RETRAN with a multichannel core model, developed•U PSI, in order to account for the contribution of potentialvoid formation in the hot regions of the core to total pres-sure buildup during the transient [22]. The multichannelcore model subdivides the core into three unequal radial re-gions. The first region is the core subasscmbly from whichthe control rod is ejected. The second core region includesthe eight subassemblics which surround the assembly fromwhich the control rod is ejected. In this region the powerexcursion is also significantly above the core average powerexcursion and consequently its thermal-hydraulic behaviorwill be significantly different from the rest of the core. Thethird region includes the rest of the core subassemblies andis represented by some average core characteristics.

To accomodate the multichannel model, RETRAN hadto be modified to accept a time-dependent power split func-tion and time-dependent regional reactivity feedback weight-ing factors. Approximations of these time-dependent fac-tors were produced from a generic Wcstinghouse report [7]and the plant FSAR [81.

5 Static CalculationsA detailed 3D QCH model of the core was set up and staticcalculations were performed using the tableset generatedby CORCOD. The emphasis in this phase was to qualifyQCH by comparing calculated results with measured plantdata.such as critical boron concentrations, rod worths, andpower distributions, and wilh the static 3D code S1LWERquarter-core calculations. The core was represented by 122hydraulic channels (one channel per assembly and one chan-nel type for ihe reflector assemblies) and 20 axial layers (16of them for the active core).

Table 2 summarizes (he critical boron concentrations fordifferent control rod patterns measured at startup as well asthe calculated values from SILWER and QCH. This tablealso shows comparisons of reactivity worths of differentcontrol groups at HZP, BOC conditions. The critical boronconcentrations are ovcrpredictcd by 46 to 90 ppm, corre-sponding to keff overestimates of up to 0.7%. The differ-ences between calculated and measured control rod worthsare within the generally accepted margin of ±10%.

The measured and calculated radial power distributionsat hot-zero power, with control bank B inserted, are shownin Fig. 2. Figure 3 shows the measured and calculated coreaverage axial power distributions for the same conditions.

The maximum ejected rod worth was determined bychecking the worth of every rod in one core octant with allother rods from both banks A and B fully inserted. Since noexperimental data arc available for checking the worth ofsingle ejet'.ed rods, the QCH results can only be comparedwilh the corresponding SILWER results. The maximumejected rod worths and static peaking factors for HZP andHFP, calculated by QCH and SILWER, are given in Tables3 and 4 and show good agreement.

Table 2: Results of startup measurements and calcu-lated values at beginning of cycle

Quantity

Critical Boron atHZP (ppm):

All Rods OutBank B InBanks A + B In

Reactivity Worths

Bank BBanks A + BBanks A + B + S

Measured

150112711025

1.773.90

-

SILWER

159113541071

1.834.166.17

QCH

157813441050

1.804.026.14

A

B

C

D

E

F

G

II

I

J

0.6B9(1714+3.7

1.349

'3?

K

°$i

$10.9300,842

•9.4

L

M

1.554

I!

1.1381.167+2.6

1.448

'11

1.13S10X7•4,1

I.S351.467•4.4

If?

1.4711.480+0.6

*1.8

0.4780.459

-3.9

0,31)

"SI

'$1

1.4871.466-1.4

0.7180.714

.0.6

1.1161.168+4.6

1.4671.467+0 0

UG8-1.0

6.361)0.368

0.9951.042+4.H

if

0.2780.283+2.6

0.4440.460+3.7

10 11 12 13

Bcznau IIZP, BOC16 • Bank B Inserted

(QCHDifference {*)

Oon l roll cilHitman(l l .nl II)

Fig. 2: Radial power distribution for HZP, BOC case

The above results confirm that SILWER and QCH arccapable to predict with acceptable accuracy critical boronconcentrations, rod worths, and power distributions in aPWR core.

1.4-

1.2-

„ 0 . 8 -

i 0 . 6 -X

0.4 -

0.2 •

0.0-

— MoosuredCalculated

100 150 200 250

Com Height, cm

Fig. 3: Axial power shape for HZP, BOC case (bank Binserted)

Table 3: Worth of Most Reactive Rod (No Xenon)

CoreCondition

HZP, BOCHZP, EOCHFP, BOC

RodPosition

C9C9E9

Maximum Rod Worth(% Ap)

QCH

0.830.880.26

SILWER

0.810.860.24

Table 4: Stationary Peaking Factors

CoreCondition

HZP, BOC

HZP, EOC

HFP, BOC

PositionEject. Rod

C9InC9Out

C9InC9Oul

E9In

Peaking FactorFry

1.99 (2.05)4.24 (4.35)

2.09(2.11)4.50 (4.62)

1.46(1.48)1.36 (1.37)

F,

2.84 (2.92)6.06 (6.20)

4.00 (3.99)8.57 (8.74)

1.79 (1.81)1.66 (1.67)

Note: the values in parentheses are results of SILWERcalculations.

6 Dynamic Calculations

Static results have shown that the most reactive rod is inposition C9 in the core and has a worth of 0.83% or aboutSI.45. Ejection of this rod will therefore result in a super-prompt critical transient. The transient was simulated withthe power in the core initially at 1% of nominal power bypulling the most reactive rod completely out of the core in100 msec.

Because of the high reactivity insertion and the delayin the fuel temperature rise that stops (he power surge, themaximum power reached is very large. The average corepower increases by a factor of about 3000 from its initialvalue (or 30 limes the nominal power), and the maximumis reached about 146 msec after the start of the rod ejection.Also, the radial power is very skewed after the rod is out,resulting in a maximum hoi node factor of 6.2 and a nodalpower of 2810 kW/m. Later in the transient, the higherfuel temperatures and the negative Doppler effect reducethe peaking factor to a value of about 4.0. Figure 4 showsthe progression of the radial power shape at a specific axiallevel during the transient. Total peaking factor and axialpower shapes in the hoi bundle arc given in Figs. 5 and6. These figures show that the rod ejection results in veryasymmetric power distributions, particularly in the x-y di-rection, and demonstrate the importance of spatial effects[23].

Figure 7 shows the 3D results with the correspondingresults from a point-kinetics calculation with RETRAN (us-ing a Doppler weighting factor of 1.0 and a constant totalpeaking factor of 6.2). The differences found in the time de-

t = 0

0.15s

t = 0.8s

Fig. 4: Progression of the radial power distribution at planeof maximum power generation (HZP, BOC)

Fig. 5: Peaking factors variation during a REA at HZP,BOC.

23

Fig. 6: Axial power distribution in channel B7 duringcontrol rod ejection at HZP, BOC

6 0 -

! 50-

s 4D"» 3 D -

1 "•10-

: /

Q/C-H

- - • REtRAN

\i

Fig. 7: Core average power during a REA at HZP, BOC

pendcnce of core power and fuel temperature clearly showthe importance of spatial temperature-weighting in the caseof strongly asymmetric power shapes and confirm the as-sumption that point kinetics is conservative without appli-cation of spatial Doppler weighting.

7 Hot-Channel Analysis

A FRAP-T calculation was performed for the purpose ofdetermining temperature distributions, fuel enthalpy, andcritical heat flux at the hot spot during the REA. The powerhistory for the hottest channel (channel B7), as well as radialand axial power shapes, were taken from the correspondingQCH calculation. Local effects were accounted for by usinga local peaking factor of 1.15.

The radial averaged fuel enthalpy reaches the maximumof 55 cal/g shortly after the power peak and declines after-wards (see Fig. 8). Figure 9 gives the temperatures in thefuel rod as a function of time. The fuel centcrline temper-ature does not reach the maximum during the simulationtime, but the rise is tapering off f. 0.8 sec and the tem-perature stays well below the melting point. The claddingtemperature reaches its maximum value of 488°C shortlyafter the power peak and declines afterwards. The CHFratio (W-3 correlation) is shown in Fig. 10 and has a min-

imum of 2.36 which is well above the allowable minimumCHF ratio.

These results are well within the safety limits and provethat the integrity of the fuel is never in jeopardy during thistransient.

Fig. 8: Enthalpy at hot spot during a REA at HZP, BOC

•oo.a _

4D0.O _

j

«li« in

•Ua aui

••a loot

•ooo .oooo

Fig. 9: Temperatures at hot spot during a REA at HZP,BOC

Fig. 10: CHF ratio at hot spot during a REA at HZP, BOC

24

8 Pressure Calculations

The calculation of the pressure rise following a rod-ejectionaccident was performed using RETRAN and a multichannelcore model, and included models for the cold legs with re-circulation pumps, hot legs, steam generators with constantconditions on uie secondary side, and the pressurizer withsurgeline. The analysis was done for end-of-life conditionsas documented in the FSAR. These conditions have beenshown to be conservative with regard to ejected rod worthand reactivity feedback.

In order to bracket possible failure modes, four caseswere considered:

• 100% break (45 cm2) with escaping subcooled fluid,

• 10% break with escaping subcooI.U fluid,

• 10% break with escaping two-phase fluid,

• no break and no pressure relief from opening of safetyvalves.

The results indicate that the reactor power excursionis not significantly affected by the behavior of the systempressure since the power excursion (peak amplitude andhalf width) is only determined by the reactivity worth ofthe ejected control rod cluster and ihe Doppler feedback.Thus, the peak fuel enthalpy is also, to a large degree, notinfluenced by the system pressure.

The influence of the break size on the maximum pres-sure is as follows (see Fig. 11):

• The pressure drops immediately after the rod is ejectedin the case of the full-sized break located at the upperclosure head,

• For a break size reduced to 10 %, the pressure firstincreases about 0.5 MPa above the initial pressure of15.7 MPa before dropping. Also, for lhe case wheretwo-phase fluid is released through the 10% break, thepressure first increases about 0.5 MPa before drop-ping.

• If the break size is zero and only the steam generatorsserve as heat sinks, the maximum system pressurereached is 17.6 MPa.

No brtok- - 100S brtoU. tubcooUd liquid

- • 10* braak. fubcooUtf liquidI0X brach. Iwo-phati flow

Fig. 11: System pressure variation for a REA at HZP, EOL

All these pressure rises are well below the limit of 22.0MPa. This demonstrates that the integrity of the primarysystem is not endangered following a rod-ejection accidenteven under the conservative assumptions used in these cal-culations.

9 Conclusions

The transient, presented here, is characterized by a rela-tively high rod worth (in excess of one dollar), and thereforecauses a superprompt critical power excursion and produceslarge power shape distortions. The results of the analysisshow that none of the safety criteria is violated and confirmthe conservatism used in the design of the KKW Beznau II.Even for beyond-design basis starting conditions (resultingin excessive reactivity worths), the integrity of the fuel andthe safety of the plant arc never threatened during the REA.

The study showed that core parameters, such as criticalboron concentrations, reactivity worths, and power distri-butions can be determined with good accuracy by SILWERand QCH. It also proved the importance of spatial effectsin (his type of transient.

Acknowledgments

This work would not have been possible without thecooperation and assistance of the plant management in pro-viding the necessary information. Support from the SwissFederal Nuclear Inspectorate (HSK) and the UnterausschussKerncncrgie dcr Uberlandwerke (UAK) is gratefully ac-knowledged.

References

[1] E. KNOGLINGER, "Simulationsmodelle zur Transien-ten-Analyse der Reaktoren in der Schweiz," Energie-Forschung 1988, Jahresberichte der Beaufiragten, Sek-tion Encrgieforschung BEW, Bern (1988).

[2] D.J. DIAMOND, CJ. HSU, R. FITZPATRICK, "Reac-tivity accidents - A reassessment of the design-basisevents,"NUREG/CR-5368, BNL-NUREG-52198, BrookhavenNational Laboratory, January 1990.

[3] "Nuclear Safety Criteria for the Design of StationaryPressurized Water Reactor Plants," ANSI/ANS-51.1-1983, American Nuclear Society (1983).

[4] "Assumptions Used for Evaluating a Control Rod Ejec-tion Accident for Pressurized Water Reactors," Regu-latory Guide 1.77, U.S. Atomic Energy Commission(1974).

25

[5] R. K. McCARDELL, D.I. HERBORN, J.E. HOUGH-TALING, "Reactivity accident test results and analysesfor the SPERT III core - A small oxide fueled pres-surized water reactor," IDO-17281, Phillips PetroleumCo. (1969).

[6] T. FUJISHIRO, "Current status of research and regula-tion relevant to reactivity initiated accidents in Japan,"in Proceedings of the Technical Committee Meeting onReactivity Transient Accidents, IAEA-TC-610, Interna-tional Atomic Energy Agency, November 1987.

[7] D. H. RISHER, Jr., "An Evaluation of the Rod EjectionAccident in Westinghouse Pressurized Water ReactorsUsing Spatial Kinetics Methods," WCAP-7588, Revi-sion 1A, Westinghouse Co. (1975).

[81 Westinghouse Electric Corporation, "Final Facility De-scription and Safety Analysis Report for the Nor-dostschweizerische Kraftwerke A.G. Beznau NuclearPower Plant Unit 1," Volumes A - C (1970).

[9] E. KNOGLINGER, L.A. BELBLIDIA, G. SKOFF,J.M. KALLFELZ, "Three dimensional space-time de-pendent analysis of a rod ejection accident in a PWR,"PSI Bericht No. 52, 1989.

[10] J. H. McFADDEN, RE. NARUM, C.E. PETER-SON, C.NOBLE, R.F. FARMAN, J.A.McCLURE,M.P. PAULSEN, K.D. RICHERT, E.D. HUGHES, G.C.GOSE, L.V. ELLIS, NX. ELLIS, K.R. KATSMA, C.G.MOTLOCH, G.C. RICE, D.S. TROTT, "RETRAN02 -A program for transient thermal-hydraulic analysis ofcomplex fluid flow systems," Vol I of EPRI-NP-1850-CCMA, Electric Power Research Institute, June 1987.

[11] D. J. DIAMOND, "The control rod ejection ac-cident," in "The Reactor Analysis Support Package(RASP)Volume 3: PWR Event Analysis Guidelines,"EPRI NP-4498, Vol. 3, Electric Power Research Insti-tute (1986).

[12] S. BIAN, "Application of reactivity weighting to rodejection accident analysis in a pressurized water reac-tor," Nucl. Technol., 41_, (1978), 401 .

[13] K. O. OTT, Introductory Nuclear Reactor Dynam-ics, American Nuclear Society, LaGrange Park, Illinois,USA (1985).

[14] S. LANGENBUCH, "QUABOX/CUBBOX-HYCA:Ein dreidimensionales Kernmodel mit parallelenKuehlkanaelen fuer Leichtwasserreakioren," GRS-A-926, Gesellschaft fuer Reaktorsicherheit, Garching,Germany, January 1984.

[15] L. J. SIEFKEN, C M . ALLISON, M.P. BOHN, S.O.PECK, "FRAP-T6: A computer code for the transientanalysis of oxide fuel rods," NUREG/CR-2148, EGG-2104, May 1981.

[16] J. M. PARATTE, K. FOSKOLOS, P. GRIMM, C.MAEDER, "Das PSI Codesystem ELCOS zur sta-tionarcn Berechnung von Leichtwasserreaktoren", Pro-ceedings of Jahrestagung Kerntechnik, TravemUnde,1988, 59.

[17] P. GRIMM and J. M. PARATTE, "Validation of theEIR LWR calculation methods for criticality assessmentof storage pools," EIR Berichl Nr. 603, 1986.

[18] P. GRIMM and J. M.PARATTE, "PWR core followcalculations using the ELCOS code system," to be pre-sented at the Int. Topical Meeting on the Physics ofReactor Operation, Design and Computation, April 23-27, 1990, Marseille, France.

[19] S. LANGENBUCH, W. MAURER, W. WERNER,"Coarse mesh flux-expansion method for the analysisof space-time effects in large LWR cores," Nucl. Set.Eng., 63 , (1977), 437.

[20] S. LANGENBUCH, "Ein neues Grobgitterverfahrenzur LOsung der orts- und zeitabhgngigen Neutronendif-fusionsgleichungcn," Dissertation, TU Munchen, Ger-many (1976).

[21] S. LANGENBUCH, "Das dreidimensionale Kem-modell QUABOX-HYCA mil parallelen Kiihlkanalen,"Reaktortagung Hannover, 15-18 April 1978.

[22] D. SAPHIER, E. KNOGLINGER, M. A. ZIMMER-MANN, "Improving the RETRAN-02 core modelingfor the simulation of the rod ejection accident," 6thInt. RETRAN Meeting, Washington, D.C., September1989.

[23] E. KNOGLINGER, L.A. BELBLIDIA, G. SKOFF,J.M. KALLFELZ, "Spatial effects of the rod-ejectionaccident in a PWR," to be presented at the AmericanNuclear Society Annual Meeting, Nashville, Tennessee,June 1990.

26

Thermomechanical Behaviour of a Subassembly Hexcan under totalInstantaneous Blockage Conditions

R. Atlinger-, W. Heer», P. Wydler», D. Desprez', J. Louvet', A. Zucchini*

* LWR-Safety Programt Laboratory for Reactor Physics and Systems Technologyt Commissariat à l'Energie Atomique, CEN Cadarache, SLPaul-lcz-Durance Cedex - France§ ENEA, CRI E. Clementel, Viale G. Bercolani 8, Bologna - Italy

Abstract

Since 1987, PSI (EIR) collaborates with CEA and ENEAin the ihcrmomechanical analysis of the hexagonal wrapper(hexcan) of a sodium-cooled subassembly which is sub-jected to a total instantaneous blockage of the coolant at theinlet. The work plays an important role in the framework ofthe development and validation of a French accident codefor simulating such a (hypothetical) blockage accident in asodium-cooled fast reactor. In a first phase of the work,the relative performance of different heat transfer and stressanalysis codes for this type of melting problem has beenassessed. It was shown that the behaviour of the wrappercan be modelled up to failure, the time of failure, how-ever, being rather sensitive to the numerical formulation.The following account of the study has been published as aconference paper [1]. In Switzerland, the work was fundedjointly by PSI and the Federal Office of Energy.

1 Introduction

The total instantaneous blockage (TIB) of a subassembly atnominal power plays an important role in the safety philoso-phy of future fast reactors. In the thermal melt-propagationscenario of the TIB, a central question concerns the modeof failure of the subassembly wrapper. To provide a ba-sis for validating data and methods used in the thermaland mechanical analysis of the behaviour of the wrapper,CEA-DRP at Cadarache is carrying out the COTHAA ex-periments [2], in which circular and hexagonal tubes areheated and pressurized to failure under controlled condi-tions. In this context three calculational benchmarks havebeen formulated and analysed using the heat transfer codesDELFINE (CEA) and ADINAT (PSI) in combination withthe stress analysis codes INCA (CEA), SOLVIA (PSI) andABAQUS (ENEA).

The aim of the first benchmark problem was to validatethe heat transfer codes against a simple problem for which

there exists an analytical solution. In this exercise emphasiswas on the treatment of the latent heat. The second bench-mark problem dealt with the heat conduction in a hexago-nal wrapper (hexcan) subjected to TIB thermal loads. Themechanical response of the wrapper was determined in thethird benchmark problem using a decoupled thermomechan-ical theory. This benchmark exercise allowed to comparethe methods for treating weakening of materials near themelting point and to investigate the influence of geometricnonlincarities.

2 Heat Conduction in a Semi-InfiniteRegion

A one-dimensional heat conduction problem with meltingis calculated: A semi-infinite region has an initial temper-ature which is constant over the whole space, suddenly atemperature larger than the melting temperature is main-tained at the free surface. The resulting temperature tran-sients arc determined using the DELFINE and ADINATcodes with one- and two-dimensional meshes; the latterwas used for verification purposes. Constant but differentmaterial properties were taken for the solid and the liquidphases. The numerical results arc compared to the exactanalytical solution given by Carslaw and Jaeger [3]. Forthe melt front x(t) the analytical expression is

x(t) = 2A JäT^t.

The dimensionless melting constant A and the ther-mal diffusivity of the liquid region a, are 0.301 and 3.319mm2/s, respectively.

Two methods for modelling the melting process areconsidered in the finite element codes. These involve amelting model based on the enthalpy method [4] and anequivalent variation of the heat capacity, the so-called c-variation model. In the ADINAT code, the melting model isrestricted to a lumped approximation for the specific heal

27

capacity and to the Euler backward time integration; themelting model in the DELFINE code and the c-variationmodel require no such limitations.

In general, the numerical and the analytical results agreequite well. Fig. 1 shows the position of the melt front £(()calculated with two-dimensional meshes. Both the melt-ing and the c-variation model give a good representationof the melting phenomenon, provided the spatial mesh andthe time steps are chosen appropriately. In combinationwith a fine spatial mesh the melting model can accomo-datc larger temperature increments per time step or elementsize than the c-variation model. The melting model ofDELFINE appears to model the passage of the melt-frontthrough element boundaries more accurately than the melt-ing model of ADINAT; whereas ADINAT still converges incases where DELFINE fails because of too large tempera-ture increments. With the c-variation model, the time stephas to be sufficiently small to ensure an adequate samplingof the effective specific heat capacity near the melting tem-perature. Some inaccuracies are observed at the beginningof the calculation. They are caused by the infinite temper-ature gradient at the hot boundary; therefore, the calcula-tions were started from the analytical solution at times asindicated in the table of Fig. 1. The infinite temperaturegradient is particular to this benchmark problem.

two-dimensional mesh and material properties as given inFig. 3.

Fig. 4 shows the calculated temperature evolution inthe midst of the face of the hexcan. Up to the onset ofmelting at around 22 s, the variations of the temperaturereflect the variation of the heat flux at the inner surface andthe temperature of the sodium at the outer surface. The on-set of melting has a small effect on the rate of temperatureincrease at the inner surface. At 27 s, the heat exchangeto the surrounding sodium is interrupted, simulating evap-oration of the sodium. In response, the temperature at theouter surface rapidly approaches the melting temperature,the ADINAT code predicting a nearly 1 s long plateau be-fore complete melting is achieved. Upon melt-through, thedisappearence of the mitigating latent heat effect furtherenhances the rate of temperature increase.

On the whole, the DELFINE and ADrNAT predictionsof the hexcan temperature evolution are in close agreement.Numerical instabilities at the onset of melting were encoun-tered due to too large changes of the material properties pertime step; they can be overcome by reducing the time steps.It could be shown that the c-variation model for the latentheat is applicable. Furthermore, it was confirmed that, forboth codes, the range of convergence extends to beyond thepoint of complete melting of the hexcan.

3 Temperature Evolution in a Hexcan 4 Structural Response of the Hexcan

The temperature evolution is determined in a subassemblywrapper of the form of a Superphenix hexcan. The heat fluxai the inner surface of the hexcan, the heat exchange to thesurrounding sodium and the temperature of the sodium aregiven (Fig. 2). The resulting temperature transients arecalculated using the DELFINE and ADINAT codes with a

The aim of the third benchmark was to determine the struc-tural response of a Supcrphenix hexcan under thermal loadtaken from the second benchmark and a superimposed pres-sure load. A further aim was to see how the numericalproblems caused by loss of mechanical strength near themelting temperature are solved.

ANALYTICAL

MELTINGMELTINGC-VAR.

DELFINE

0.5 s *2 s •2 s •

ADINAT0.01 s A2 s a2 s o

0 2 4 6 8

TIME [s ]

Fig. 1: Propagation of the melt front in a semi-infinite region.

10

28

A ihcrmoelaslic-plasiic bilinear maierial law and ma-lerial properties according lo Fig. 3 were used in the fi-nite clement analyses. A plane strain slate, the von Misesyield condition and isotropic hardening were assumed. Theloss or mechanical strength near the melting icmpcralurecaused no numerical problems in the ABAQUS code. Theconvergence problems are overcome in the INCA code bysetting temperatures above a certain cut-off temperature tothis temperature. Inactive elements are used in the SOLVIAcode Tor the simulation of the molten material and the radialreturn method was implemented in a user supplied mate-rial model Tor assuring convergence at high temperatures.The lime steps in the INCA calculation were chosen to beconstant, those in the SOLVIA calculation were chosen toproduce temperature increments of about 50 "C, whereasthe ABAQUS code adjusted the time steps automatically inthe final part of the calculation. Basically, a small strainand small displacement formulation was used, but geomet-ric nonlincaritics were also considered in a complementarycalculation using the large displacement formulation in theABAQUS code.

The calculated midfacc and corner displacements arcshown in Fig. 5. The midface displacement clearly reflectsthe effect of the internal pressure ramp as given in Fig.2. The influence of the modelling of the molten materialin the codes is small, the INCA and SOLVIA codes pre-dicting a slightly stiffer behaviour of the structure than theABAQUS code with the corresponding formulation. Withthe large displacement formulation, the ABAQUS code al-lows to model the temporary inward motion of the cornerwhen - under the internal pressure - the hexcan tends toapproach the more resislant geometry of a cylinder (Fig.7).

A useful quantity for predicting collapse of the hexcanis the ratio of the effective strain to the temperature depen-dent ultimate strain. The evolution of this ratio is shownin Fig. 6 (it has lo be taken into account that the resultsfrom the INCA code refer directly to the nodes while theresults from the other two codes refer to the integrationpoints). An interesting feature of the thermomcchanicalhexcan problem is that the strain ratio at the inner surfaceof the hexcan reaches a maximum after 12 s to 15 s and thenbecomes rather small when, in response to the temperatureincrease, the ultimate strain starts to increase more rapidlythan the effective strain. At the outer surface of the hexcan,the large displacement formulation predicts a similar strainratio behaviour with the difference that due lo the sloweroutside temperature rise the maximum is reached only about2 s before melt-through. The largest sirain ratios arc ob-served in the midst of the outer surface (point A in Fig.6). For the small displacement case, the analyses indicatecollapse at around 25 s due to rapidly increasing sirains.The large displacement calculation predicts the hexcan tocollapse at 27.5 s through a uniform swelling. This timeof collapse has been confirmed by an analytical analysis inone-dimensional ring geometry.

Up to nearly 24 s, corresponding to about 2 % strainand 3 mm to 4 mm maximum hexcan displacement, theINCA, SOLVIA and ABAQUS predictions or the imporlantparameters arc found lo be in good agreement. In particular,

there is good agreement for the maximum of the strain-ratio at the inner surface. A large displacement formulationseems to be necessary for modelling the phenomena in thelast phase just before melt-through.

5 Conclusions

• The predictions of the temperature evolution and themechanical response agree well.

• In the temperature analyses the effect of the latentheat may be simulated by a variation of the specificheat capacity. Numerical problems can be circum-vented by an appropriate choice of the mesh andsmall time steps.

• In (he mechanical analyses the method of modellingthe molten material has a small influence on the re-sults. The small displacement calculations agree withthe large displacement calculation provided the strainsdo not exceed about 2 %.

• Geometric non-linearities have lo be considered forreliable solutions up to collapse.

References[1] R.O. ATTINGER, W. HEER, P.WYDLER, D. DE-

SPREZ, J. LOUVET, A. ZUCCHINI, "Thermal andmechanical behaviour of a subasscmbly hexcan undertotal instantaneous blockage conditions; a code com-parison", Trans. 10th Int. Conf. on Structural Mechan-ics in Reactor Technology, Anaheim (Ca, USA), 1989,Division E, 311-316.

[2] D. DESPREZ, J. LOUVET, P. HAMONET, P.ANZIOU, "COTHAA programme: An Experimentalapproach to ihermomechanical aspects of ihe TIB of asubasscmbly", Trans. 10th Int. Conf. on Structural Me-chanics in Reactor Technology, Anaheim (Ca, USA),1989, Division E, 317-322.

[3] H.S. CARSLAW, J.C. JAEGER, "Conduction of healin solids", Clarendon Press, Oxford, 2nd Ed. (1959),282-296.

|4] W.D. Ill ROLPH, K.J. BATHE, "An efficient algo-rilhm for analysis of nonlinear heat transfer with phasechanges". Int. J. Num. Mclh. Engng., 18 (1982), 119-134.

29

TN o = 950°Ca

• S - 12 W/mmz

V

1.3 MPa

10 20

TIME [s]

Fig. 2: Loading conditions.

30 40

HEAT FLUX 0INITIAL TEMPERATURE To « 430 °CENVIRONMENTAL TEMPERATURE TNo

COEFFICIENT OF HEAT EXCHANGE hBEFORE 27 s 0.096 W/Kmm 2

AFTER 27 s 0.000 W/Kmm 2

PRESSURE p

1000

TEMPERATURE °C

Fig. 3: Material properties.

2000

HEAT CONDUCTIVITY kSPECIFIC HEAT CAPACITY

eLATENT HEAT Lr 1.96

THERMAL EXPANSION a 17.85*10'YOUNGS MODULUS E 161 GPaPOISSON RATIO v 0.3YIELD LIMIT ery 4 8 4 MPaHARDENING MODULUS H 7 GPaULTIMATE STRAIN e , 1B.BO»1O

MELTING TEMPERATURET, 1385 °C

VALUE AT 400 °C

20.05»10"3 W/Kmm

4.30«10~3 J /Kmm 3

J / m m 3

'/K

, - 3

3000

I 2000 •

<

a:2 1000

OUTER

INNER

Tm e , t = 1385 °C

INNER

OUTER

10 20TIME [ s ]

Fig. 4: Temperature evolution in the hexcan.

DELFINE

ADINAT

30 40

30

NCASOLVIAABAQUSABAQUSLARGE

TEi—

LJLdO<_la.a

15.0 •

12.5 •

10.0 •

7.5 •

5.0 •

2.5 •

o.o17.5 20.0 22.5 25.0 27.5 30.0

TIME [s]Fig. 5: Displacement in the hexcan.

0.017.5 20.0 22.5 25.0 27.5 30.0

TIME [s]

0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 22.5 25.0 27.5

TIME [ s ]Fig. 6: Strain raiio to determine collapse at midface.

ABAQUS LARGE27.5 s

MOLTEN MATERIAL

SOLVIA25.0 s

INITIAL MESH :

Fig. 7: Deformed hexcan at collapse.

31

Boil-off Experiments with the PSI-NEPTUN Facility: Analysis and CodeAssessment Overview Report

S.N. Aksan, F. Stierli, G.Th. Analytis

Laboratory for Thermal-hydraulics

Abstract

Postulated small break loss of coolant accidents (SBLOCA)in PWRs may involve partial core uncovery; during such atransient, it is important to be able to predict the historiesof the dry-out point and the rod surface temperature. Inthis work, we shall report on a serious of boil-off experi-ments performed in the healer rod bundle NEPTUN, com-parison of the experimental results with predictions fromthermal-hydraulic codes and, based on this, specific modelimprovements in these codes.

1 Introduction

After the TM1-2 accident, in LWR safety analysis, increasedattention has been paid to low flow and intermediate or lowpressure transients which may sequentially lead to uncov-cry, over-healing and damage of the core. Consequently,there is an obvious need for assessing thermal-hydraulictransient analysis codes as far as their ability to correctlypredict a number of phenomena (e.g. level swell, rod sur-face temperatures in the uncovered region etc.) is con-cerned. Hence, considerable effort is being spent not onlyin improving the available codes, but also in performingcarefully controlled experiments in test facilities; the re-sults of these separate effect tests can be directly utilizedfor assessing these codes or even for developing new, morephysically sound models.

At PSI, the NEPTUN healer rod bundle was originallydesigned for low pressure (< 5 Bar) reflood experiments[1]. Additionally, a number of core-uncovery (boil-off) ex-periments have been performed to investigate the mixturelevel decrease and resulting fuel rod heat-up above the levelthat may occur in a PWR during a SBLOCA.

In this work, we shall report on a number of boil-offexperiments performed in NEPTUN under different initialand boundary conditions, the cooling influence of the exter-nal thermocouples and the subsequent utilization of theseexperiments for assessing and improving thermal-hydraulictransient analysis codes.

2 Description of NEPTUN Facility andTest Procedure

The NEPTUN heater rod bundle was originally built to sim-ulate a PWR core. It contains 33 electrically heated rodsand four guide tubes. Each heater rod has an axial heightof 1.68 m and radial dimensions similar to a PWR nuclearfuel rod. There arc five fuel assembly spacer grids, axiallylocated at equal distances and a continuously variable axialpower profile can be achieved. The instrumentation allowsthe measurement of cladding (at 8 cqui-distant axial lev-els), housing thermal insulation and coolant temperatures,absolute and differential pressures, flow-rates, carry-overrates and heating power. For a complete description of ihefacility, the reader is referred to [1] 12].

The NEPTUN boil-off tests were initiated by turning onthe full defined power and terminating the coolant flow tothe bundle at a given pressure. Subsequently, the coolantswells in the boiling ienght of the test-section and a certainamount of water is expelled out, depending on the powersupplied to the heater rods. As the upper parts of the test-section are uncovered, heat-up of the rods is initiated. Therod surface temperature data shows no multi-dimensionaleffect. Finally, we should point out that each experimentwas performed more than once and the data were found tobe consistent and rcpcatablc.

3 Experimental Results and SystemResponse

Ten boil-off experiments were performed, and for more de-tails, see [2]. Three parameters were varied - rod power,system pressure and initial coolant subcooling. Rod powerlevels were chosen to represent the nuclear decay power,and the system pressure was varied over 1 - 5 bar.

Experiment Nr.50O7 (Table 1) was chosen as the "basecase" because it was conducted at higher pressure and in-termediate power.

32

Table 1: Boil-off experiments analyzedby TRAC.BD1/MOD1

Pressure, bar

Subcooling, K

bundle power, kW

Experiment Number

5002

1

0

24.6

5006

5

12

42.1

5007

5

12

24.6

5008

5

12

10.5

5011

5

39

75.1

3.1 System Response - Base Case, Experiment5007

Fig.l shows an overlay of the core power history, core fluidlevel as measured by the core total A/> measurement, andtypical response of the heater rod thermocouples for thebase case. Notice that the power is increased at about SOs, and that a rapid drop in the core fluid level occurs atabout 100 s. This delay lime between initial power andthe rapid initial level decrease is a result of heating thesubcoolcd water to the saturation temperature. Shortly afterthis (betwen 100 - 110 s), large voids arc formed due tovapour generation, expelling some of the liquid from thecore. After this initial liquid swell, the liquid continuesto slowly boil-off, as shown by the decreasing core A;> inFig.l. The cladding thermocouples dry out and heat upwith different heal-up rates for each axial elevation. At ~800 s the power is shutl-off and the rods are allowed tocool down before the system is refloodcd.

TOO £

• 100 Z00 300 400 500 600 700 BOO 900 1000

Time. S

Fig. 1: NEPTUN system response-base case (lest 5007-high system pressure, intermediate rod power).

3.2 General Remarks on System Response [2]

(a) As expected, increased core power resulted in morerapid boil-off and claddig heal-up, while decreasedpower had ihe opposite effect.

(b) Decreasing the sytcm pressure from 5 lo 2 bar re-sulted in more rapidly decreasing liquid levels androd dry-ouls ranging from 100 - 200 s earlier thanfor the base case.

(c) Many of the experiments were repeated under ihesame conditions; the experimental reapeatabilily wasvery good for all pressures.

4 Evaluation of the External SurfaceThermocouple Response

One of the objectives of the boil-off experiments in theNEPTUN facility was to obtain experimental data for as-sessing any perturbing effects of the external surface ther-mocouples used in LOFT, during simulated small-breakcore uncovery conditions. Prior to the NEPTUN Experi-ments, it was hypothesized that the external surface ther-mocouples might cause additional selective cooling of therods, which would result in delayed dryoul for a slow coreuncovery experiment in LOFT. Also, the added increasein surface area for heat transfer (fin effect) might result inadditional atypicalitics.

The first series of experiments performed in NEPTUNconsisted of eleven tests, ten boil-off and one adiabatic heat-up test. In these tests, three parameters were varied: rodpower, system pressure and initial coolant subcooling asalready described in chapter 3. The effects of the externalsurface thermocouples were determined by comparing thecladding temperatures (as measured by the external LOFT-lype surface thermocouples) to cladding temperatures fromthermocouples within the cladding of the NEPTUN heaterrods [1]. There is only one active external surface ther-mocouple on each of the five LOFT-lype rods, the otherexternal surface thermocouples arc replaced by dummy el-ements.

Overlay plots of the thermocouple responses for manydifferent thermocouples at each axial elevation are con-tained in ihe appendices of |3] for each experiment. Theseplots indicate that the readings of the external surface ther-mocouple is well within the response spread of the internalthermocouples. In this report, as bounding cases, experi-ments 5007 (base case) and 5011 corresponding lo smalland intermediate break LOCA decay heal levels, respec-tively, arc discussed.

Overlay plols of Ihc thermocouple responses at level 4corresponding lo the maximum linear heal generation posi-tion on the heaters are shown in Figs. 2 and 3 for exper-iment numbers 5007 and 5011 respectively. Notice that asystematic lower temperature is measured by ihe externalsurface thermocouples. This temperature difference can betaken as an estimate of the cooling effect of the externalsurface thermocouples and is less than 20 K. Notice alsoin Figs. 2 and 3 thai Iherc is less than 5-10 s difference inthe initial dry-oul times for all level 4 thermocouples, bothembedded and external.

5 Code Assessment

The NEPTUN boil-off experimental data were used forassessing the thermal-hydraulic transient computer codes

33

ouu

600

400

200

n

Center bare rods

: /

m*rf • • • • • • •

1 1 1 l i t

0 200 400 600 800 1000

Time, sFig. 2: Comparison of center rod internal and LOFT ther-mocouples (test 5007, axial elevation, 946 mm - level 4).

800

— Center bare rods

LOFT thermocouple

200 800 1000400 600

Time, s

Fig. 3: Comparison of center rod internal and LOFT ther-mocouples (test 5011, axial elevation, 946 mm - level 4).

TRAC-BD1/MOD1 and RELAP5/MOD2. The assessmentcalculations performed with TRAC-BD1/MOD1 will be gi-ven in more detail in the next subsections. During the earlyphase of the assessment of the RELAP5/MOD2 code, somesimulation and calculational difficulties were encounteredfor boil-off cases e.g. very large discrepancy in calculat-ing the amount of expelled water out of the lest section asshown in Fig. 4 [4]. Further calculations were not per-formed with RELAP5/MOD2, until the reasons for suchdiscrepancies were identified. Two other attempts usingRELAP4/MOD6 and RELAP5/MOD1 were also not verysuccesful [5]. The RELAP4/MOD6 calculations could beperformed until the on-set of nucleate boiling and as soonas vapor was produced, very large pressure spikes wereobserved, resulting in lime consuming and costly calcula-tions. RELAP5/MOD1 which employes five-equation hy-brid model calculated about 100 s earlier critical heat fluxoccurance with respect to RELAP5/MOD2 and the amountof expelled water out of test section was even more over-predicted.

E

(U

• D

j Li

qu

s(0a

Col

la

a.o

1.6

1.2

0.8

0.4

-

-

-

-

; — Calculated

:v>.v Experiment

' \

\ . '

-— • " '

A

- i r~0

1200 r

1000

800

600

200 400 600 800

400

• TRAC-BD1 component2 ROD GROUP 1 NODE 7

• Experimental roddata TRS-946

200 400 600 800

Time, s

Fig. 4: Comparison of measured and calculated collapsedliquid level and peak axial power level rod surface tem-perature histories, using frozen version of TRAC-BD1 forNEPTUN boil-off experiment 5007.

5.1 Assessment of the Frozen Version of TRAC-BD1/MOD1 and Poblem Areas

A number of NEPTUN boil-off experiments have been uti-lized for assessing the predicting capabilities of the thermal-hydraulics transient analysis code TRAC-BD1. Originally,the experiments were analyzed with version 12 of TRAC-BD1 [6] and the problem areas were identified [7,81. Sub-sequently, five of these experiments were re-analyzed byusing a frozen version of the code (version 22) commonlyknown a.. MODI [8],

Sine . the models related to the dominant physical phenom-ena in these core uncovcry experiments arc the same in bothversions of (he code and the problem areas identified withMODI were almost the same with the ones of version 12,we shall concentrate on the results obtained by employingTRAC-BD1/MOD1. For more details, the interested readeris referred to a series of reports (7,8); here we shall out-line our findings related to the problem areas as well asthe model improvements already implemented in MODI.One of these improvements is already included in the BF1version of the code recently released. Four boil-off exper-iments at 5 bar and one at 1 bar were analyzed using thefrozen version of TRAC-BD1/MOD1 and experiments areseparatly summarized in Table 1. A number of numericalproblems have been revealed in the course of the analysis of

34

these experiments with TRAC and have been extensivelyanalysed and reported elsewhere [7,8]; here, we shall re-strict our attention on the problem areas of the code relatedto the actual physical modeling of the phenomena takingplace.

Comparison of measured and calculated collapsed liq-uid level (CLL) and peak axial power level rod surfacetemperature histories for exp.5007 are shown in Fig.4. Onecan readily draw the following conclusions regarding thepredicting capabilites of the code:

(a) TRAC-BDl/MODl underpredicts the CLL historiesand hence, predicts an earlier CHF than the mea-surements show. Clearly, the code overpredicls theamount of water expelled from the test section. Thesedifferences are more pronounced for the 1 bar exper-iment.

(b) Generally, TRAC-BDl/MODl predicts an earlier CHFthan the measurements show; hence, sudden expul-sion of water from the test section is predicted tooccur earlier [7,8]. Also, the predicted rod surfacetemperatures during nucleate boiling are 8-15 K be-low the measured ones.

(c) It was noticed that after the rod power was turned off,the slopes of the predicted and measured rod surfacetemperatures were different, indicating that the heattransfer coefficient in this region was overpredicted.This was changed as we shall discuss in due course.

5.2 Model Improvements

As Fig.4 shows the main problem of TRAC-BD1 is thatit overprcdicts the amount of water expelled in the boil-off tests. The origin of this was traced back to the ratherhigh intcrfacial shear calculated by the interfacial frictioncorrelation used for bubbly/slug flow. This correlation al-though appropriate for tubes, has recently been shown notto be suitable for rod bundles [10]. The interfacial shearforce per unit volume / ; in TRAC-BD1 for the bubbly/slugflow regime is based on the following vapour drift velocitycorrelation

vd = pf(1)

Starting from this expression, it can readily be shown [9,11]that

- C0V,) (2)

where a is the surface tension, g the gravity constant,

C, =1 - O C Q

1 -a (3)

and Co ^ 1.3. Based on the work of Bestion [10], we im-plemented in TRAC-BD1 a new bubbly/slug / , correlationsuitable for rod bundles; it is based on the following vapordrift velocity correlation

Pi(4)

where DH is the channel hydraulic diameter; this results inthe following expression for /,-:

/ • = \ClV, - ~ CoVl) (5)

Fig. S shows the comparison of the measured and calculatedCLL's (with the new TRAC-BDl/MODl version) and peakaxial power level rod surface temperature histories for ex-periment 5007. There is now excellent agreement betweenmeasurements and predictions. Here, we should note thatthe same excellent agreement between measurements andpredictions was obtained when analyzing the other boil-offexperiments in Table 1 with the modified code, <vitti the ex-ception of the 1 bar experiment 5002, for which the amountof water expelled was still over-predicted. The code devel-opers have already implemented this new / , correlation inthe new code version TRAC-BF1.

We could not trace the origin of the earlier transition tonucleate boiling predicted by the code; though, recent work[12] has shown that since the thermocouples are a little be-low the rod surface, an oxide layer having a thickness of 25fan on the surface could in fact increase the thermocouple

Calculated

Experiment

800

a.1000

800

600

400 B

TRAC-BD1 component2 ROD GROUP 1 NODE 7

Experimental roddata TRS-946

B

200 400 600 800

Time, sFig. 5: Comparison of measured and calculated collapsedliquid level and peak axial power level rod surface tempe-rature histories, using modified version of TRAC-BD1 forNEPTUN boil-off experiment 5007.

35

i 4.0

1 3.0

2 2.0

I »U

0-0.1

- — Standard RELAP 5Modified Inlerlaclal frictionIn bubbly and slug flowExperiment

200 400 600Tims, s

800 1000

Fig. 6: Water entrapment in NEPTUN boil-off experiment50O7, calculated by RELAP5/MOD2, with and without newcorrelation for the intcrfacial friction in bubbly and slugflow.

1300

I 900

S 700

S.« 500

I

Standard RELAP 5

Modified Intarfaclai MotionIn bubbly and slug flow ,<

Experiment /"'y'

200 400 600Time, s

800 1000

Fig. 7: Rod cladding temperature at measurement level 4in NEPTUN boil-off experiment S007, calculated by RE-LAP5/MOD2, with and without new correlation for the in-tcrfacial friction in bubbly and slug flow.

readings by as much as 12 K. This is a plausible explana-tion of the differences between measured and predicted rodsurface temperatures during nucleate boiling.

Finally, the problem of overproduction of heat transferafter the power was turned off was traced back to the steamcooling logic of the code and since this problem was notencountered when analyzing the boil-off experiments withversion 12 of the code [7,8], the steam cooling heat transferlogic of version 12 was re-introduced in MODI.

Specifically, for steam cooling, in MODI the followingheat transfer coefficients arc defined [8]

"v.lam —DH

o.8/pn>o.aa>,

(6)

fc.,.Urt = 0.023(fie,)os(Pr)O33/.- !,/£»H (7)

where all the symbols have, their usual meaning and

Cr - (9)

In MODI, the following selection logic exists for hv if a >0.999:

&„ = hv,iUT (10)

/ / /(„ < hv,ncJl,, =/»„,„„

This was modified as follows (for a >0.999):

(11)

(12)

(13)

The effect of this change can be seen by comparing theslopes of the rod surface temperature histories after thepower is cut-off in Figs. 4 and 5.

6 Conclusions

The NEPTUN experiments have provided thermal-hydraulicdata simulating nuclear reactor core boil-off conditions atlow pressure (1 - 5 bar). The data obtained from these testsproved to be useful in assessing the modeling capability ofavailable computer codes.Analysis of the experimental boil-off data indicate that:

• Increasing core power resulted in more rapid boil-off and cladding heat-up, while decreasing power re-sulted in the opposite trends, as expected.

• The lower system pressures resulted in more rapid de-crease of liquid levels and faster rod dry-outs relativeto the base case.

• Dry-out times of the internal and external surfacethermocouples were within 10 s of each other atany axial elevation for all rods in the bundle. Thecladding external surface thermocouples measure thecladding temperatures that would have been mea-sured in their absence within 0 to -20 K.

Analysis of a number of NEPTUN boil-off experiments andcomparisons with TRAC-BDI/MODl predictions showedthat:

• The collapsed liquid level history is underpredictedand consequently, CHF occurs earlier than in the ex-periments. Clearly, TRAC overpredicts the amountof water expelled from the test section.

• Generally, earlier incipience of nucleate boiling ispredicted; recent investigations indicate the differ-ences between measured and predicted rod surfacetemperatures during nucleate boiling can be due tothe formation of an oxide layer around the electricalheater rods.

• After the rod power was turned off, the slopes ofthe predicted and measured rod surface temperatureswere different, indicating that the calculated steamcooling heat transfer coefficient was overpredicted.

36

To improve the predicting capability of TRAC-BD1/MODIthe following main modifications were introduced:

• An alternative bubbly/slug intcrfacial shear correla-tion, more appropriate for bundles and used in theCATHARE code, is implemented in the code. As aresult of this change, the collapsed liquid level histo-ries are correctly predicted by decreasing the intcrfa-cial friction in this flow regime.

• The steam cooling heat transfer logic, used in version12 is re-introduced in MODI, specifically to eliminatethe differences during the steam cooling phase afterthe power was turned of.

References

[1] H. GRUTTER, F. STIERLI, S.N. AKSAN, G.VARADI, "NEPTUN bundle reflooding experiments:Test facility description", EIR-Report No.386, 1980

[2] S.N. AKSAN, F. SIERL1, G.Th. ANALYTIS, "Boil-off experiments with the EIR-NEPTUN facility: Anal-ysis and code assessment overview report", EIR-Bcricht Nr. 629, 1987

[3] EX. TOLMAN, S.N. AKSAN, "Summary results ofthe NEPTUN boil-off experiments to investigate theaccuracy and cooling influence of LOFT cladding sur-face thermocouples", EGG-LOFT-5554, EG&G Inter-national Report, 1981

[4] G.Th. ANALYTIS, M. RICHNER, M. ANDREANI,S.N. AKSAN, "Assessment and uncertainty identifi-cation for RELAP5/MOD2 and TRAC-BDl/MODIcodes under core uncovery and reflooding condi-tions", 14"1 Water Reactor Safety Information Meet-ing, Gaithersburg, Maryland, USA, 1986

[5] M. ANDREANI, "Brief notes on the simulation ofthe NEPTUN Boil-off experiment 5007 by means ofthe RELAP4/MOD6 and RELAP5/MOD1 (cycle 25)Computer Codes", Letter to S.N. Aksan, 24.11.1986

[6] J.W. SPORE et al., "An advanced best estimate com-puter code program for boiling water reactor loss-of-coolant accident analysis". Vol. 1,2,3. NUREG/CR-2178, 1981

[7] G.Th. ANALYTIS, S.N. AKSAN, "TRAC-BD1 as-sessment under severe accident boil-off conditions".Fifth International Meeting on Thermal Nuclear Reac-tor Safety, Karlsruhe, 09. - 13.09.1984, ProceedingsVol. 3, 1821

[8] G.Th. ANALYTIS, S. N. AKSAN, "Trans. Amer.Nucl. Soc.", 47, 1984, 493

[9] D.D. TAYLOR et al., "TRAC-BD1/MOD1: An ad-vanced best estimate computer program for boilingwater reactor transient analysis". Vol. 1,2 NUREG/CR-3633, 1984

[101 D. BESTION, "Inicrfacial friction determination forthe ID-6 equations two-fluid model used in theCATHARE code", presented at the European Two-Phase Flow Group Meeting, Southampton!, England,1985

[11] G.Th. ANALYTIS, 'Trans. Amer. Nucl. Soc.", 52,1986, 481

[12] G.Th. ANALYTIS, S.N. AKSAN, F. STIERLI, G.YADIGAROGLU, "Dynamics of core voiding duringboil-off experiments", 4"1 Miami International Sym-posium on Multi-Phase Transport and Paniculate Phe-nomena, Miami Beach, Florida, USA, December 1986

37

Effects of Vapour/Aerosol and Pool Formation on Rupture of Vessels ContainingSuperheated Liquid

j . Schmidli, S. Banerjee', G. Yadigaroglu'

Laboratory for Thermal-hydraulics

* Dcpt. of Chemical and Nuclear Engineering, University of California, Santa Barbarat Institute of Energy Technology, Swiss Federal Institute of Technology, Zurich

Abstract

Experiments have been conducted involving Frcon 114 (boil-ing point 3.5 °C at 1 bar), contained in spherical glass con-tainers that were suddenly ruptured. The vapour/dropletcloud formed and droplet sizes have been measured withvideos and high speed movies. Pool formation has beenmeasured by collecting the liquid on a cold tray. The mainparameters varied were the liquid superheat, the height ofrelease and the percentage of container filling. The super-heated liquid shattered into droplets at a vapour volumefraction of « 50%. The velocity and size of the largestdroplets had Weber numbers » 20-30. The droplets veloci-ties were, however, an order of magnitude smaller than pre-dicted by isentropic releases of liquefied gases. The dropletvelocities were overpredicted by isentropic expansion frombubbly to slug flow. The ovcrprediction was caused by en-tropy production, which has been associated with bubblenucleation and growth. The liquid pool appeared to formdue to drop impingement on the ground. It contained onlya fraction of the liquid that would have been deposited,based on adiabatic (or isentropic) flash fractions. A modelfor vapour cloud formation, based on the evaporation of thedrops as they travel outwards appears consistent with theseobservations, and with those made using a smoke curtain.

1 Introduction

Reactive and toxic substances arc often processed and storedunder conditions that would give rise to rapid vaporiza-tion when deprcssurized to atmospheric conditions. Ex-amples of such materials are propane, butane, propylcnc,chlorine, ammonia and hydrogen cyanide. To assess theconsequences following accidental releases of this type ofliquefied gas, it is necessary to model spreading phenomenawhich can be roughly divided into the following (overlap-ping) stages:

• Incrtial expansion (driven by internal energy)

• Gravitational slumping (if the vapour/mist is heavierthan the surrounding air)

• Turbulent dispersion due to the wind and related ef-fects

The purpose of this paper is to focus on the first stagefor sudden releases, to elucidate the source term for cal-culation. Previous work [1] [2] suggested that the liquidfragments into drops driven outwards in a roughly hemi-spherical cloud by the vapour arising from flashing. Bothsets of experiments indicated that even though significantportions of the release were expected to remain as liquidsafter initial flashing, very little liquid was observed on theground in the form of a pool. It was concluded that theliquid was entrained in the vapour and evaporated, as airwas mixed into the cloud.

The first set of experiments were made with propane[1] in cylindrical tanks with dished ends. The tanks haddiameters of 40 mm and 60 mm with a length to diameterratio of 35:1. Many important results arose from theseexperiments.

First, a model, qualitatively verified by experiment, wasmade for the bursting of a tank filled with superheated liq-uid that indicated rapid propagation of cracks, so that thetanks burst open in a short time compared with the timeconstants involved in flashing, i.e. for bubble nucleationand growth. This indicated that catastrophic failures due tocracking were possible, and that the release geometry couldnot be treated as turbulent jets. The releases were roughlyhemispherical (because of interference from the ground).

Second, the initial spread of the release was modelledby a diffusion equation with turbulent diffusion coefficientsbased on a mixing-length type formula. The length scalewas taken from photographs of the eddy sizes for the large-scale motion at the edge of the vapour cloud, and the ve-locity scale was based on the speed with which the frontmoved. The quantitative verification of this model involvedignition experiments that determined the uppper and lowerignition limits of the cloud as a function of time. The test

38 Paper published in J. Loss Prev. Ind., 3 ( 1 9 9 0 ) , 1 0 4 - 1 1 1

results appeared to agree well with the predicitons based onan eddy diffustity that was essentailly constant. The con-centration profiles arising from this model are Gaussian andthe cloud remains between the upper and lower flammabil-ity limits for a relatively short time(0.2 s for the 40 mmtank and 0.3 s for the 60 mm tank), so direct concentrationmeasurements are difficult.

Third, HESS et al. [1] semi-analytically integrated thecloud mass between the upper and lower flammability limitslo establish a formula for the amount of energy held inthe expanding cloud. This paper formed the basis for allsubsequent models thai habe been proposed for expandingpuffs and liquefied gases, e.g. the TNO model [3], theWorld Bank model [4] and MAURER's model [5].

The paper by MAURER et al. [2] was a continuationof previous work [1]. The substance released was propy-lene, in tanks ranging from 40 mm diameter (0.226 x 10~3

m3) to 750 mm diameter (1 m3). The larger releases indi-cated the need to modify the HESS model in some respects.The basic change was to treat the expanding cloud as twozones. The inner zone was considered to be well-mixed, i.e.of uniform concentration, whereas the outer zone was gov-erned by diffusional processes, so the concentration profilewas described by a Gaussian distribution. In approximateterms, about half the release was in the inner zone, and halfwas in the peripheral zone. Furthermore, the eddy diffusioncoefficient in the centre zone had to be given a time depen-dence (but not a spatial dependence) to fit the results. Theeffect of lank size (and hence size of the release) was deter-mined empirically, using the same technique as previouslyfor upper and lower flammability limits.

The paper was also noteworthy for measuring flamespeeds and peak overpressures. Using the KUHL et al.[5] method for calculating pressure waves generated bysteady flames, MAURER et al. calculated flame speedsfrom the pressure data and compared these with direct mea-surements, which were in reasonable agreement. (A factorin the KUHL et al. [S] calculation method had to be modi-fied.) The flame speeds ranged from about 20 ms" 1 for thesmallest tanks lo about 50 ms"1 for the largest, and wereexplained [2] as being effected by the turbulent diffusivityto the square root (i.e. the DAMKOHLER [6] model). Animportant qualitative observation [2] was that the flamesappeared lo be affected by "high intensity small-scale tur-bulence within the large eddies". They implicitly proposedthe following:

• The cloud is turbulent during its expansion due toconversion of a fraction of its internal energy at thestart to kinectic energy in velocity fluctuations.

• Due lo this mixing process, the period when substan-tial portions of the cloud are within the flammabilitylimits occurs when it ist still intensely turbulent.

• The high flame speeds and peak overpressures ob-served when such clouds are ignited are due to theturbulence generated in this expansion process.

This mechanism is important for flammable clouds ig-

niled in the early stages of their development. Thereforeflame propagation experiments in premixed, still clouds canbe misleading since the mechanism of turbulence genera-tion due to expansion is missing. Only after the turbulencegenerated by the initial expansion and slumping becomesdissipated, do other (in some cases weaker) mechanismscome into play. For example VAN DEN BERG et al.[7] suggested that obstacles (or ouVr turbulence generat-ing mechanisms) are necessary lo explain the high flamespeeds inferred from the overpressures that appeared to bereached in large vapour cloud deflagrations. Previous [2]experiments suggest that such obstacles are not necessaryfor, although they might contribute lo the high flame speedsin many of the vapour cloud deflagrations. A specific ex-ample is the deflagration where 5500 kg of propylenc wasreleased and on subsequent ignition, killed people [8], Thisis close to the range of the experiments of MAURER el al.[2] that generated up lo 70 mbar overpressure.

While the works of HESS el al. [ 11 and MAURER et al.[2] arc key lo our present understanding of vapour clouds,they leave unexplained many important aspects. For exam-ple, all (he liquid appeared to be caught in the cloud andvaporized. However, at lower flash fractions (and perhapsfor liquids with different physical properties such as highersurface tension), some portions of the liquid may be ex-pected to fall to the ground and form a pool that evaporatesmuch more slowly. To understand this, we need lo observedrop sizes and velocities (and trajectories) as functions ofrelease size, flash fraction and certain key physical proper-ties. Direct observations of the pool formed, and its tem-perature and geometry are also necessary. This division ofthe mass of a release between the pool and vapour/aerosolhas a first-order effect on consequences, whether due tothe flammability or the loxicily of the material. (This isbecause the pool will evaporate more slowly than the sus-pended drops of the material, and will thus participate lessin the hazards that arise - except for X>1 fire-\

Another issue of importance that remains unresolved re-lates to whether the two-zone model applies to the vapour-mist fraction of the release for materials other than propy-lenc, and if it does, what (if any) are the effects of physico-properties and release conditions on eddy difusivity. Notethat we have no real understanding of the turbulence gen-erating mechanisms, or those for enirainmcnt of air inlo thecloud. Therefore, generalization to a variety of fluids anddifferent intial conditions is not possible. A related issuehas to do with the condensation of water vapour (enteringthe cloud with the air), which not only forms a persistentmist that increases cloud density but also affects its temper-ature. (In the initial stages, condensation warms the cloudsomewhat, but later on warming of the cloud is impededdue to the need to vaporize the water droplets). It is known[8] that releases of important materials such as ammoniaand indeed of propylcnc [2] are accompanied by mist for-mation due ID water vapour condensation. CHAN ct al.[9] explained some of the key discrepancies between ex-periments and computer modelling for ammonia releases asarising from water vapour condensation in the cloud.

Turning now to the mechanisms for entrainmcnl and tur-

39

bulence generation in the expanding cloud, it is clear thatthe mean motion is radial, and there are no simple sheargenerating mechanisms except on the ground. It is there-fore hard to explain the turbulence generation mechanisms.Clearly, the expanding vapour front will experience inter-facial instabilities due to unequal viscosities and densitiescompared with air - however, on preliminary appraisal theywould appear to be rather too weak to explain the intenseturbulence previously observed [2].

A clue may be contained in the recent work of COOKEand KHANDADIA [10] and for this reason, the main resultsare briefly renewed. They exploded cylindrical containerscontaining water, using axial charges. The droplets formedwere observed to form a "shell", and had sizes and veloci-ties that could be divided roughly into two regions. In thefirst, the droplets accelerated due to the pressure inside theshell driving them outwards. In the second, the pressurewithin the shell was probably lower than that outside (al-though no measurements were available) and the dropletsstarted to decelerate. In one sense, the droplet shell couldbe thought of as a porous piston with some gas leaking out-wards, while the internal pressure was high, and some airleaking inwards, while the inward pressure was lower thanatmospheric. The shear could then be generated by flowpast the droplets, and could explain the intense turbulenceseen earlier [2].

COOKE and KHANOADIA's [10] experiments also in-dicated that droplet sizes could be correlated in terms of anEotvos number and a Reynolds number, i.e. the accelera-tion was important as well as the relative velocity betweenthe droplets and the surrounding fluid. They presented anumber of photographs showing the breakup mechanismfor the droplet that showed they were "punched out" intoglove shapes and then fragmented - a mechanism observedfor bubbles that is well correlated by EoivOs and Reynoldsnumbers (see CLIFT et al. [11]).

While COOK'S and KHANDADIA's experiments werethe first to offer some insight into such mechanisms, theywere made in a rather non-representative way as far asreleases of liquefied gas are concerned, i.e. the expandingcloud was initiated by explosive charges rather than bubblenuclean'on and growth in a superheated liquid. A recentset of experiments involving superheated liquid have beenreported by BETTIS et al. [12] using Freon 11. Theyused a spherical container divided into two hemispheresthat could be rapidly pulled apart. They measured a numberof parameters including the amount of liquid that formeda pool and the amount in the vapour/aerosol cloud. Aninteresting result was a dependence on the amount of fillingof the vessel. This can be explained by the fact that thevapour starts to expand rapidly on depressurization, whereasthe liquid has to flash, i.e. bubbles must nucleate and grow.Therefore, the expansion time scales of the two fluids aredifferent.

With this very brief review, which is by no means ex-haustive, we have tried to indicate the main areas of un-certainty. To summarize, it is desirable to know how muchliquid forms a pool and how much is in the aerosol on rapiddepressurization of liquefied gases. The way in which the

size of the release and the properties of the fluid affectthese phenomena is important Second, the mechanismsinvolved arc important, so that reliable models may by de-veloped based on them - to handle predictions for a varietyof scales and fluids. Hence, droplet sizes, velocities and tra-jectories are important in addition to the size, temperatureand extent of the liquid pool (if any), and the temperatureand motion of the cloud front The behaviour of atmo-spheric water vapour within the cloud is also of importanceand should be clarified.

2 Experimental

Based on these considerations, an experimental programmehas been initiated in our laboratories, supported by theSwiss National Fund. The initial experiments to be doneare on a small-scale involving several fluids. The data willserve as a base for model development, with the modelbeing spot-checked against some large-scale experiments.

Complete sets of small-scale experiments will be madewith butane. This will be followed by a smaller set of exper-iments with propane (or propylene) and ammonia. Finally,spot-checks will be made on a large-scale with propane (orpropylene). Table 1 summarizes the main parameters forthe small-scale experiments.

As indicated in Table 1,24 experiments will be made forFroon 114 and butane, with repetitions as required. Basedon these experiments, a smaller test matrix will be under-taken for propane (or propylene) and ammonia. The large-scale experiments with propane (or porpylene) will be onabout the same scale as the largest experiments of MAU-RER el al. [2], i.e. 1 m3. However, approval for theseexperiments is still to be obtained. The relevant physicalproperties for the lest fluids are listed in Table 2, indicatingthat they cover a wide range.

T a b l e 1: Small-scale lest matrix

Fluid Sia

Rashfracl. Vessel(nom.) Height filling Meas.

Treon U 40.25 ground Full Cloud rad.

Butane SO ml level Cloud temp.Pool size

Propane 0.15 Pool temp,(or propy- 500 ml 5 dia. 50% Drop, sizelene) Drop, velocityAmmonia11 0.08 Water vapour

condenseda Partial test matrix only

2.1 Experimental Procedure

The procedure successfully adopted to obtain small-scalereleases is by shattering a glass flask, containing the fluid,on impact. The impact was controlled, so that there was noingress of the "hammer" into the contained liquid, i.e. ihehammer stopped shortly after impact. Photographs showthat symmetrical releases may be obtained by this procedure

40

Table 2: Thermophysical propenies of test fluids

SubstanceFreon 114ButanePropaneAmmonia

Boiling pointat 1.013 bar

°C3.75-0.7-42.1-33.5

SaLpressureat25°C

bar2.12.69.89.9

Liquidspecific

heatkJ kg ~ ' ° C - i

0.984°2.12*2.20'4.74°

Latent heatat 1.013 bar

kJ kg"'1363864261369

Rash fractionfor 25°C

0.150.140.350.24

liquid thermalcond. at 20°CW m -> K " '

0.0660.0910.0780.477

Surface tensionat 25°CN m - '

0.0100.0120.0070.036

» At 20 » C6 At 50 ° C

for elevated containers, and roughly hemispherical cloudsfor ground level containers.

An alternative being developed for all liquid contai-ners is to fracture them using hydraulic pressure, i.e. bypressurizing through a small liquid line connected to thecontainer until the rupture occurs. Methods of scoring glassand plastic containers to ensure symmetrical releases arebeing studied. This technique cannot be used for partiallyfilled containers.

2.2 Measurement Techniques

The techniques developed are primarily based on photogra-phy, except for temperature, pool mass and drop size distri-bution. High speed movies (using a Hycam II) and videosare taken of the release evolution. The larger drops can bedistinguished clearly and drop sizes and velocities are ob-tained. The smaller drops cannot be seen clearly enough toallow drop size determinations, although their velocities areestimated. To determine drop size distributions, a techniquebased on laser holograms is being developed.

The movies also show the extent of the pool formed, itsshape and its regression characteristics. However, the massof liquid deposited in various locations needs to be deter-mined using trays placed on weighing scales. The trays arecooled to temperatures well below the minima observed inthe pools to slow down evaporation and allow good mea-surement of deposition. The most difficult measurementsrelate to the position of the vapour cloud front, as distinctfrom the position of droplets. The droplets appear to flyahead of the expanding vapour and their trajectories can beseen in the videos. To see the vapour front that follows re-quires use of a smoke curtain. By embedding the container(or one side of it) at the edge of the rising smoke curtain, itis possible to follow the progress of the vapour front quiteclearly. The technique works because the curtain is quitethin and clearly distinguisable before it is reached by thevapour. It is pushed aside by the vapour and is partiallymixed into the cloud. However, mixing within the cloud isgood, so that the smoke cannot be distinguished once it iswithin. This allows a sharp image of the front to be pho-tographed using transmitted light. For low boiling pointfluids such as ammonia and propylcnc, it should also bepossible to distinguish the region in which the air mixes,simply by the water mist formed (if the air is sufficientlyhumid). This technique has been used [2] for propylene

release. However, for releases of higher boiling fluids, e.g.the model fluid Freon 114, condensation of atmosphericwater vapour is insufficient to demarcate the front, and thesmoke curtain technique is necessary. As many liquefiedgases of interest have boiling points similar to Freon 114and flash fractions similar to atmospheric conditions, e.g.chlorine, butane etc., it is necessary to explore their rangeof properties as well. The temperature within the pool andthe cloud are also of interest in the development of mod-els. They may be measured directly using good responsethermocouples.

3 Results

This paper reports on experiments completed with one ofthe fluids in Table 1, i.e. Freon 114. The purpose ofthese experiments was to identify the phenomena of inter-est. From these experiments arose the lest matrix shown inTable 1. Many of the experiments required for Freon 114have been completed, as well as some additional experi-ments that were needed to cover the range of parametersthat should be investigated. For example, before choosing50 ml and 500 ml as container sizes for the matrix, experi-ments were made to check the strength of the container sizeeffect using 50 ml and 100 ml containers. As the effect wasrather weak, it was decided to compare results for sizes thatwere different by an order of magnitude. Similarly, exper-iments were done with containers that were full and 75%full. Again the effect was weak and it was decided that100% and 50% filling would be adequate.

3.1 Experimental SetupThe experimental data were obtained using the apparatusshown schematically in Fig. 1. The figure also shows thecold tray with a typical pool sketched in. The smoke cur-tain is to the right of the glass flask and is formed by apipe, through which smoke is injected (the lower horizon-tal pipe), and a pipe through which the smoke is suckedout (the lop horizontal pipe). Note that the top pipe im-peded heavy gases (or mist) from slumping into the curtainfrom the top. However, slumping of heavy gases (or mist)under the action of gravity could be observed for the re-gion above the glass flask and to its left This is becausewhen the whole cloud (roughly hemispherical) started to

41

SO or 100 ml flask

Hammer

Thermocouple

Fig. 1: Schematic of apparatus

Table 3: Parameters varied in experiments

Flasksizeml

50

100

Heightof

release

Ground

5dia.elev.

Fillingof

flask

full

0.5

Super-heat°C

21.5

14.5

6.5

Calcul.equilibr.quality

0.15

0.10

0.046

Calcul.equilibr.void frac.

0.972

0.956

0.904

collapse under gravity, the region above the smoke curtainwas prevented by the top pipe from falling into the curtain.Therefore, the front seen was due to slumping of roughlyhalf the hemisphere, i.e. the portion to the left of the cur-tain.

3.2 Qualitative ResultsThe main parameters varied in the experiments are shownin Table 3. The table indicates that 24 runs were made vary-ing the flask size, release height, flask filling and superheat.Each of the first 3 parameters were varied over 2 values,whereas the last was varied over 3 values. Correspondingto the superheat, we also present the equilibrium quality andvoid fraction (vapour volume fraction) for isentropic expan-sion of the superheat liquid down to atmospheric pressureand its boiling temperature. The calculation assumes thatno air is mixed into the cloud during expansion.

The equilibrium quality (isentropic expansion) is

v = "fl~°r~ (')and the corresponding void fraction a is calculated from

where: s is entropy, J kg "' K ~ l; v is specific volume,m3 kg ~l;x 's quality; a is void fraction; f is liquid; g isvapour; fg is difference: vapour to liquid; 1 is initial state;and 2 is final state (at atmospheric pressure).

Note that in Table 1, even though the qualitaties liein the range 0.046 to 0.15, the vapour volume fraction isbetween 0.904 and 0.972. Thus the volume of the cloudwould be about 10 to 40 times larger than the initial vol-ume for the liquid given isentropic expansion with no airmixing down to atmospheric pressure. This is a pureley hy-pothetical number, since the liquid fragments into dropleiswhich are thrown outward to much greater distance thanare implied by the above volume expansion numbers.

The main qualitative results of the experiments were

• On rupture of the glass flask, the liquid fragmentedinto droplets on relatively low amounts of expansion,e.g. to about twice its original volume.

• The drops were thrown outwards with velocities thatwere not a function of drop size, i.e. the initial ve-locities of all the drops were quite similar in our ex-periments.

• The vapour front during this initial expansion (to thepoint of fragmentation) moved negligibly into thesmoke curtain. No displacement was seen for thecurtain.

• The drops moved through the curtain but did not dis-place it, i.e. there may be dilution due to evaporationof the drops, but the drops did not move the curtainin bulk.

• The curtain starts to move after the drops have passedthrough it Clearly this is due to slumping of thecloud formed by droplet evaporation along their tra-jectories. Also, a contributing factor may be evapo-ration from the pool (formed by droplet impingementon the ground).

• The pool appeared to form extremely quickly aftercontainer rupture, due to impingement of the frag-mented liquid with the ground. For elevated releases,the pool had an annular shape because the dropletsimpinging at the centre were of lower concentration(due to geometrical considerations for an expandingspherical shell).

3.3 Quantitative Results

The first measurements relate to droplet sizes and velocities.Only the larger droplets could be seen from the high speedvideos, so size distribution could not be obtained. Figure2 shows the Weber number ( u ^ p / o - ) as a function ofsuperheat and the other experimental parameters for groundlevel releases. It is clear that most of the data are withinthe range 15 to 30, which is in excellent agreement withprevious studies on fragmentation of superheated liquids[12]. The large droplet sizes were in the range 2.8 to 3.5mm and the velocities were 2.8-3.6 m s"1.

42

25

20

15

10

: 0; x

: °r a

L

: i i

50 ml .50 ml .100 m100 m

i 1 i

filled1/2 filled. filled. 1/2 filled

0*

1 . 1 i 1

A

aA

Observeddiametersvelocities

i 1 i 1 i 1 f 1

X

D

8droplet2.8-3.5mm2.3-3.6m/si 1 i 1 i 1

10 12 111 16 18 20 22 24

Superheat (°C)

Fig. 2: Weber number versus superheat and other experi-mental parameters

0 Saturation temperatureat ambient pressure 998 mbar

Fig. 4: Typical pool temperatures as a function of time:50 ml, 100 % full, ground level, d = 12.5 cm, run V 37.11

12 = -

o o oE E o oo o o oin in - f

c! o.e2

u- 0 . 4

0 .2

0

II Pool

Vapour

6.5 14.5 21.5

Superheat (°C)

Fig. 3: Pool and aerosol mass

Figure 3 indicates the vapour to be expected if the re-lease flashed down to equilibrium conditions and, by dif-ference, the amount of liquid thai was entrained in thevapour stream (the aerosol fraction). In reality, the pro-cess did not go in these two sequential steps, i.e. flashingto equilibrium followed by evaporation of the suspendeddroplets. The mass fragmented into droplets with relativelysmall amounts of vapour being formed before fragmenta-tion. This was followed by evaporation of the droplets asthey were thrown outwards. Some of the liquid impingedon the ground forming a pool, but the rest evaporated fromthe droplets before they hit the ground. Thus the divisioninto vapour and aerosol fractions in the figure is artifical,and is only shown to indicate the amount of liquid abovethe equilibrium flash fraction that entered the vapour stream.As expected, the pool mass decreased with superheat andcontainer fill fraction. There was no reliable trend, how-ever, of the fractional pool mass with container size (at leaston the scales studied).

Fig. 4 shows the pool temperature for a typical run. It isevident that the temperature dropped well below the boilingpoint of 3.5°C, and indicates a high degree of evaporationinto the air, rather than boiling. The minimum temperaturecoincided with complete pool vaporization, the warmingtrend afterwards relating to the substrate temperature.

Qualitative trends are the same for all experiments withthe minimum being shifted in time depending on the poolmass to surface area ratio. The quantitative results are notvery important, since the vaporization rate depends on iher-mophysical properties of the substrate, its temperature, theair temperature and flow rate, insulation. However, in thiscase the substrate was a thin cardboard sheet with low heatcapacity and thermal conductivity. Therefore, the pool re-gression was primarily due to evaporation into the air. Theresults indicate the importance of evaporation as a mecha-nism for vapour generation. In this discussion, "evapora-tion" is used in the sense of mass transfer to the unsaturatedair above the pool in contrast to "boiling", which wouldbe caused by heat transfer to the pool from the substrateetc., i.e. one process is diffusionally controlled, the otheris heat conduction/convection controlled. In reality, bothmass and heat transfer play roles in governing vaporizationof the pool.

It was mentioned previously that the smoke curtain didnot move at all in the early stages of the transient. Videosof the phenomena indicated that drops were formed andihrown outward with subsequent pool formation within about0.1 s. Up to this lime, the smoke curtain showed no signof bulk movement in all cases, indicating that no coherentvapour front was advancing to displace the curtain. It isclear from this that negligible amounts of vapour formedbefore the droplets were formed and thrown outwards. Fur-thermore, the effect of the droplets penetrating the smokecurtain could not be seen, probably because the numberdensity of the droplets was low and did not significantlydisturb the curtain.

However, the smoke curtain did start to move well afterthe initial transience was over. The shape of the vapourfront that moved the curtain is shown in Fig. 5. The left-hand figure shows that in some cases the bottom of thecurtain moved more than the top - usually, this happenedat the lower superheats, when much of ihe liquid mass wasin the pools. For the higher superheats, the shape wasmore like that shown in the right hand figure, where themovement was greater towards the top of the smoke curtain.

Some quantitative data on movement of the cloud areshown in Table 4 for elevated releases. These typical results

43

Small superheatlarge fraction in pool

Large superheatsmaller fraction in poolslumping of the vapour cloud

Fig. 5: Sketch of smoke curtain movement

Table 4; Comparison of droplet velocities from models'

Super-hct' C21.514.56.5

World Bank/TOO model

m s " 1

42.726.711.9

Model withflashing

to 30% vapourvolume fraction

m s - 1

8.36.04.4

QualityTNO/WorldBank model

0.1560.10S0.046

Quality30% voidfractionmodel

0.00470.00380.0029

a experimental results arc 2.8 - 3.5 ms~

4 Modelling AspectsThe experiments indicate that for range of boiling pointsand flash fractions similar to Freon 114, i.e. as 3.5°C and0.15 to 0.2, the most important phenomena relate to dropletformation and movement.

The pools appear to be formed by impingement, and thecloud is formed primarily by evaporation of the droplets asthey are thrown outwards. (Clearly for higher boiling lique-fied gases and lower flash fractions, some of the liquid mayfall into the pool without fragmenting into droplets. At theother extreme, at very high superheats and flash fractions,the droplets that impinge on the ground may evaporate sorapidly that no pool is formed). A model for the initialstages of catastrophic rupture of liquefied gases must there-fore concern itself with droplet formation and movement(including evaporation and impingement).

The only published model for velocity of the cloud frontis based on isentropic expansion down to atmospheric con-ditions. This is used in both the TOO model [3] and theWorld Bank model [4]. It is instructive to compare theresults of such a model with experiments. The general ex-pression for an isentropic expansion from initial conditionsdenoted by 1 to some pressure conditions 2 (which maybe atmospheric or greater) leads to a velocity for the frontgiven by

Table 5: Vapour front position from smokecurtain displacement

Superheat=27 °C; elev.;50 ml flask (full)

Superheat=27 °C; elev.;100 ml flask (full)

Time afterrelease

s

Cloudradiusmm

Time afterrelease

s

Cloudradiusmm

1.241.322.042.242.442.64

358365498525541583

0.480.760.840.921.321.54

225307346388527566

indicate that the movement starts well after the dropletshave vanished due to deposition or evaporation, the pool hasbeen formed and pool evaporation has commenced. Fur-thermore, the front velocity equals as 0.2 to 0.3 m s"1,indicating more of a gravity current than the droplet veloc-ities, which are at least an order of magnitude higher.

From this, the tentative conclusion to be drawn is thatfor low superheats and large pools, the vapour front ismainly due to pool evaporation and occurs well after the ini-tial transience (that forms the pool) is over. The front lookslike the diagram on the left-hand side of Fig. 5. For highersuperheats (Table 5), the vapour front looks more like thatshown on the right-hand side of Fig. 5, and appears tobe due to slumping of a cloud formed by the evaporatingdroplets.

(3)This assumes that the cloud is essentially homogeneous

and is at equilibrium until condition 2 is reached. The TNOand World Bank models assume that condition 2 coincideswith atmospheric pressure, i.e. the cloud expands isentropi-cally, homogeneously and with equilibrium between vapourand liquid phases until atmospheric pressure is reached.

An alternative is to assume that isentropic expansion(homogeneous equilibrium) prevails until the liquid Massfragments, when it undergoes a transition from the bub-ble growth regime. At this point, the mass breaks up intochunks of liquid that fragment rapidly into finer droplets asthey move into the relatively still air. In this case, the ini-tial fragmentation would occur when the cloud has between30% and 50% volume fraction of vapour.

In Table 3, we calculated the velocities that would beobtained if the cloud expanded down to atmospheric pres-sure (the TNO/World Bank model) and down to a pressuresuch that the vapour volume fraction was 30%. Note thatin both cases the experimental droplet velocities (as 3.2 ms"1) are significantly overestimated. The TNO/World Bankmodel overestimates the velocities by an order of magni-tude, whereas the model based on flashing to 30% voidfraction overestimates the velocities by a factor of about2. This is probably because the velocity is lowered due toconduction/convection processes controlling bubble growth.A slightly more sophisticated model would probably giveacceptably close velocities for the droplets, but is not pre-sented here, although the main features are clear.

44

The second step in the model is then to obtain the dropsizes based on a Weber number criterion or equivalent. Forthe range of properties in our experiments, the usual Webernumber criterion of w 20 for the largest drops would ap-pear to be adequate (see Fig. 2). It is still necessary to findthe drop size distribution, but previous experiment indi-cate that a log-normal distribution often correlates with theresults. In any case, using the Weber number criterion andan appropriate distribution, the droplet population shortlyafter fragmentation can be characterized.

The last step following the initial evolution of Ihe re-lease is to follow representative groups of droplets as theytravel outwards (under the influence of their initial velocity,drag and gravity) and evaporate. Some impinge or depositon the ground forming a pool, whereas the rest form acloud. The initial state of the cloud (before slumping) canthen be calculated by keeping account of the amount of theliquid evaporated at various locations and its temperature.

Such calculations arc simplest when the amount of va-porization relative to the surrounding air is small, i.e. ac-count docs not have to be taken of air/vapour movementinduced by droplet motion and evaporation when relativelylarge amounts of evaporation occur rapidly. Momentumsource terms must be added to the air/vapour phase and itsmotion solved itcrativcly at every time step that the dropletsare moved.

Such a model is qualitatively different from previousmodels, but docs put the phenomena on a physical basis.Furthermore, it clears up one of the outstanding issues re-lated to vapour-cloud formation from liquefied gas releases,i.e. that turbulence mechanisms can account for the verylarge dilutions observed within a few ms of such releases.Note that a radially moving front (if it is homogeneous)docs not have strong shear producing mechanisms, such asthose that arc present in boundarry layers and in the shearlayers in jets and wakes. Therefore the main turbulence pro-ducing instabilities at the interface must be relatively weak- perhaps of Raylcigh-Taylor type. The observed rapid di-lution of releases is therefore hard to explain if a homo-geneous droplet/vapour front moves outwards into still air.On the other hand, if the droplets are thrown through thesurrounding air, evaporating as (hey go, then it is clear thatthe extent of the cloud depends on how far the dropletscan go gefore evaporating completely. The cloud is there-fore formed in a different way from the TNO/World Bankmodel, where an eddy diffusivity for mixing into the ex-panding cloud is used.

5 Conclusions

The experiments reported here indicate that catastrophic re-leases of liquefied gases lead to clouds and pools that arcprimarily a result of droplet behaviour. Droplets appear toform before a significant mass of liquid is flashed, i.e. whenthe volume fraction of vapour is about 50%. The problemof determining the cloud and pool characteristics immedi-ately after the release is then a matter of following dropletcalculations. To this end, it is important to know droplet

size distributions and velocities shortly after fragmentation.Measurement indicate that the size of the largest dropletsare adequately predicted by a Weber number criterion of as20. The droplet velocities are, however, an order of magni-tude smaller than those predicted by previous models. If itis assumed that flashing and isentropic expansion occoursto 30-50% void fraction before fragmentation, after whichthe velocities of the droplets are better predicted, thoughthey are still higher than those observed. A somewhatmore sophisticated model based on conduction/convectioncontrolled bubble growth until the flow regime changes toconlinous vapour is probably needed to adequately predictdroplet velocities. It is also necessary to check the conclu-sion of other investigators that the droplet size distributionclose to the region of formation is indeed log-normal.

The cloud and pool parameters can then be calculated.The main aspect of the experiments discussed here is thatthey clarify the main mechanisms in cloud formation fol-lowing catastrophic releases, and indicate that the very large,rapid dilution observed is probably due to droplet motionand associated evaporation through the surrounding air.

Extensive work is still needed with fluids of differentthermophysical properties, to determine whether differentphenomena are seen from those reported here. In particu-lar, it is expected that water vapour condensation will be-come significant in very cold clouds formed by liquefiedgases with low boiling points (e.g. ammonia). Further-more, for these systems, droplet evaporation may be veryrapid, so droplet calculations must take into account move-ment of the surreounding air/vapour mixture induced bydroplet evaporation and motion.

References[1] K. HESS, W. HOFFMANN, A. STOCKL, Int. Loss

Prevention Symp., The Hague, Netherlands 1974,227-234

[2] B. MAURER et al., Int. Loss Prevention Symposium,Heidelberg, FRG, 1977

[3] U. PSCHOOR, in "Evaporation" TNO Yellow Book,Netherlands, 1978

[4] World Bank, in Manual of Industrial Hazard Tech-niques, 1985

|5] A.L. KUHL cl al., 14th Symp. on Combustion, 1972

[6] G. DAMKOHLER, Einfluss dcr Turbulenz auf dieFlammcngcschwindigkcit in Gasgcmischcn, Elcklro-chemic 46, 1940

[7] A.C. VAN DEN BERG, CJ.M. VAN WINGERDEN,J.P. ZEEUWEN, and HJ. PASMAN, Int. Conf. onVapor Cloud Modelling, (Ed. J. Woodwared), AIChE,New York, USA, 1987

45

[8] J.A. DAVENPORT, in Hazards and Protection of Pies- [11] R. CLIFT, J.R. GRACE, MM. WEBER, in Bubblessure Storage and Transport of LPG, Industrial Risk Drops and Panicles, Academic Press, UK, 1978Insurers Report, 1987

[12] RJ. BETTIS, P.F. NOLAN, K. MOODIE. in Two-[9] S.T. CHAN, H.C. RODEAN, D.N. BLEWITT, Int. Phase Flashing Releases Following Rapid Depressur-

Conf. on Vapor Cloud Modelling, (Ed. J. Woodward) ization Due to Vessel Failure, IChemE symp. SeriesAIChE, New York, USA 1987,116 No. 102, 1987, 247

[10] N.E. COOKE, P.S. KHANDADIA, Int. Conf. on Va- [13] R.C. ANDERSON, C.A. ERDMANN, A.B.por Cloud Modelling, (Ed. J Woodward). AIChE, New REYNOLDS, Nuc. Sci. Eng. 1984, 88, 495York, USA, 1987

46

High Temperature Materials

G. Ullrich, K. Krompholz

Laboratory for Materials Technology and Nuclear Processes

Abstract

The problems with the applicability of nickel-base alloysand molybdenum-base-alloys at elevated temperatures aresummarized. It is pointed out that the main aspect fordesign is the time to rupture strength. The usual designcriterion is given and from that it is clarified which inves-tigations are necessary to deliver the suitable data.

Many of nickel-base alloys show a tendency to em-brittlement in the intermediate temperature regime. Thenickel-base alloy Inconel-617 is compared with the molyb-denum-base alloy TZM with respect to its time to rupturestrength. It is shown that TZM is superior to Inconel-617in helium at 1123 K but the problem is the evaporation ofmolybdenum oxide in oxidizing atmospheres. The stabil-ity of different layers in helium with a very low oxidizingpotential is discussed.

1 Introduction

Since 1973, PSI has investigated materials within the frameof the project "High Temperature Gas Cooled Reactor"

(HTGR), the application of which up to 1223 K (950 °C) istaken into consideration for special components as turbine,IHX (intermediate heat exchanger), hot gas ducts etc. [1].High temperature materials are face centered cubic nickel-or iron-base alloys in contrast to ferritic steels with bodycentered cubic structure. The main representatives of thisgroup of materials are listed in Table 1 under the usual tradename.

An example is the body centered cubic molybdenum-base alloy with 0,5 % titanium and 0,08 % zirconium whichis dealt with in a separate chapter.

2 Mechanical Properties

The materials presented are commercial alloys, the mecha-nical short term properties of which are nearly equivalentin a broad temperature region. From that point of view thecentral problem of a selection under the aspect of the longterm behaviour has to be considered: The main criterionfor the material selection is the creep behaviour, especiallythe one percent strain limit over a time of at least 30'000hours. As shown in Fig. 1, the extrapolated stress versus

TABLE INominal Compositions of Candidate Alloys [2]

MaterialX10NiCrAlTi3220,1.4876,Incoloy-800 H

Ni-Cr-22Fe-18Mo2.4603,Hastelloy-X

NiCrl5Mol4Hastelloy-S

NiCr25MolO,Nimonic-86

Ni-Cr-22Co-12Mo,2.4663,Inconel-617

TZM

C

0.08

0.10

0.02

0.05

0.07

0.012

Fe

Balance

18.0

-

-

-

-

Ni

32.5

Balance

Balance

Balance

Balance

-

Cr

21

22.0

15.5

25.0

22.0

-

Co

-

1.5

-

-

12.5

-

Mo

-

9.0

14.5

10.0

9.0

Balance

W

-

6.0

-

-

-

-

Al

0.4

-

0.2

-

1.0

-

Ti

0.4

-

-

-

-

0.5

Other

-

-

-

0.05 Cs

-

0.08 Zr

47

101

3 »•4) ,

i I i rStress rupture strength

lnconel-617

Hastelloy-X

Nimonic-86

Hastelloy-S

Incoloy-aOOH

10' 10* 105

Time, h

TIIC.NI. M,C.M,,C,.M,,C

Fig. 1: Stress rupture strength of several high temperaturematerials at 950 °C, extrapolated to 100*000 hours.

1200

1100

0 1000

<B 900

1 800

® 700

I 600r-

500400

I T,IC.N|.M,C.M,,C,.7'NI,AI TIIC.NI * M,,C.• V-NI^I

*• ^TIIC.NH M,C • M,£~ "~ ',' • i'-NI,*! • LamPlaar

TIIC.NI »M,C "* — ~ ^ . _

1 10 100 1000 1OOOO 100000

Aging Time, h

Fig. 2: Time-temperature-transition diagram of lnconel-617

time to rupture curve shows, that the long term behaviourcan only be decided after comparing the creep resistance ofdifferent materials [2]. The long term behaviour is espe-cially dependent on carbide precipitations M2:iC6 and MCas well as on the formation of intermctallic compounds as7' (Ni3AlTi) (Fig. 2), the formation of which is strongly

As received

•Aged 8000 h at 850°C

1 1 1 1 1 1 1 1 1 I 1 11 2 3 4 5 6 7 8 9 10 11 12

Load point deflection VLL, mm

Fig. 3: Long term cmbrittlement (lO'OOO h, 850 °C) ofNimonic-86, measured on pre-fatigued compact tensionspecimen at room temperature.

time- and temperature dependent [3], The importance ofthe phases upon the long term mechanical properties is dif-ficult to interprcte, not to mention a theoretical prediction.Desirable in all cases arc precipitations within the grains.It is well known today that for Incoloy-800 7'-phase preci-pitations in the temperature regime 550 °C< T / 650 °C<positively effects the material. But nevertheless, this is thetemperature region where ihc strongest material cmbriltle-ment is found. Materials with higher creep strength likelnconel-617 exhibit this region of cmbrilllcmenl at about1023K (750 °C) in only few hours.

Avoiding this for each specific alloy of this type char-acteristic temperature field of embrilllemcnt leads to longterm embrittlement (thermal ageing) if higher temperaturesare considered (Fig. 3).

3 Design Criteria

The above considerations mean that a compromise must befound with respect to selecting material due to its creepproperties and the tendency to embrittlemem. Normally fora judgement of the design the basis is the design value [4]

where <TB,T >S the measured or extrapolated long termstrength at the time 7-. If the calculated value is lower thanthe Si value, the design is allowed, if not, the design mustbe modified. Each additional stress like cold deformationafter solution annealing or fatigue which changes the ki-netics of the precipitation, influences the S( -value. Thequestion, whether the reached time or the strain reached isdominant for the design should be answered in the sensethat the inelastic strain should be limited in most cases. Thisis possible if the notch stress time-to-rupture embrittlement

48

10"1 10° 101 1O2 103 104 105

Time to rupture, hFig. 4: Comparison of TZM with Inconel-617: Rupturestrength versus time to rupture at 850 °C. For a design lifetime of lOO'OOO h, TZM, even in the rccrystallizcd state, issuperior to Inconcl-617.

is also considered or if it is possible to clarify the durationof the remaining life time (strain reserve) at different limes.

The last point is very important with respect to an ac-cident in an HTGR-plant (leak-beforc-fracture criterion).Under the aspect of a preliminarily creep damage the lcak-before-fracture criterion is very often not applicable in thehigh temperature regime.

The problem of the long term embriulement due to thecomplex structural changes which lead to enormously highinvestigation costs gave rise to cooperation between PSIand the industry. During this cooperation an investigationof the molybdenum and the molybdenum-base alloy wasperformed in parallel to investigations on the nickel-basealloys [5],[6]. The molybdenum-base alloy can be preparedby melting and by powder metallurgical techniques. Thealloy TZM with traces of litanum and zirconium is distinctlysuperior to the nickel-base-alloy Inconel-617 (Fig. 4).

The low deformability due to the carbon content is detri-mental. The variants with low carbon content exhibit acreep resistance which is equivalent to the resistance of therccrysiallizcd TZM (Fig. 4). From that even the base ma-terial is a potential high temperature material. Detrimentalis the evaporation of molybdenum oxide at temperatureshigher than 923K (650 °C). This is the reason that the ap-plication of this material is restricted to media with a lowoxidation. An artificial layer technology for this materialhas not been investigated.

4 Gas-Metal Interaction

In addition to the structural changes in the bulk of the metal,gas-metal interactions at the surface to a depth of 100 /imcan affect the materials behaviour. This is not of impor-

tance for thick walled components (creep behaviour in airand helium is nearly equivalent). If a thin walled pipe(e.g. heal exchanger lube, about 3 mm) is considered, thezone of reaction has to be taken into account. In a typicalHTGR-helium (Table 2) different reactions must be consid-ered which are presented in Table 3.

Table 2Impurity Levels of HTGR Helium (//bar)

H2

500

H2O

1.5

System

CO

15

CH4

20

pressure: 1.6

N

<

bar

2

5

Table 3Principal Gas-Metal Interface Reactions in H rGR Helium[7].

"Pure" oxidation by H2OH2O + 5 Cr —T i Cr2O3 + H,

Carburizau'on by CHA (summary reaction)CH., + I Cr — . 1 Cr7O3 + 2H2

Decarburizalion by H2O (summary reaction)H2O + M-carbide —• M + CO + H2

Combined oxidalion/carburization by COCO + 3Cr — . 1 Cr,O3 Cr7 C3

In a helium loop (Fig. 5) kinetic and thermodynamic in-vestigations were performed with respect to the differentgas-metal- or gas-graphite interactions.Three different aims were considered:

- For the long term behaviour, the gas composition isto vary in such a way (especially H2O) that neithercarburizalion nor dccarburizalion takes place where aprotective oxide layer is desirai \

- The short term behaviour (for example an accidenuilwater inlcakage) has 10 be controlled with respect tothe graphite corrosion.

- The adhesive strength and stability of suitable sim-plex, duplex, and triplex layers also wilh respect tofriction and wear had to be investigated. Parame-ters arc dynamic (heal exchanger simulator) and staticconditions. These investigations arc performed in co-operation wilh Ihc industry. Different layers weretested and two were found to be excellent wilh respcello fricUon and wear at 1223K (950 °C). The layerswere Cr2C3/NiCr and ZrO2+Y3Oj with NiCrAlY asan intermediate layer [8]. The reproducibility of theproperties of the layers and their long term behaviourat times longer than lO'OOO hours at 1273K (I'OOO°C) in helium under dynamic conditions has to beproven.

49

fcjsction

H p analysis

metafic ceramic

purified gas

putfkatkxi

0-gasciYonistgraphy

2 dillerent gas atmospheres and up to 3+22*4 dilferent temperatures l<1000 °C)can be obtained simultaneously.

Fig. 5: The helium loop permits the investigation of thebehaviour of metallic structural materials in the coolant (im-pure helium) of a high temperature nuclear reactor.a) Diagram of the loopb) Furnaces with flow meters and water measuring cellsc) Furnace with 4 ceramic tubes (99.7 % AljO;,)

5 Conclusions

It is shown lhat at the very moment the design criteria forthe applicability of alloys for a high temperature regimeare the creep data rcsp. time- lo-tupture strength data.The main candidates for this field are iron-nickcl-basc ornickel-base alloys.

The problem of long term cmbritllcmcnt can be treatedby lowering the design limit, at least for the tolerable elon-gation. The combination creep/fatigue is still an unsolvedproblem and part of scientific research.

A very interesting alloy for the high temperature regimeis the molybdenum-base alloy TZM. Being very brittle atroom temperature, it becomes ductile at elevated tempera-tures and it is superior with respect to its creep rate com-pared with commercial nickel-base alloys.

Gas metal interactions are important for thin walledcomponents like heat exchanger tubes. The moisture con-tent of the atmosphere is responsible for carburization ordccarburizalion.

The formation of artificial layers which are stable in theHTGR atmosphere is also important for the molybdenumbase alloy TZM.

In all cases results from long term tests are necessary.

References[II W. JAKOBEIT, J.-P. PFEIFER, G. ULLRICH, "Eval-

uation of high-temperature alloys for helium gas tur-bines", Nucl. Tech. 66 (1984), 195 - 206

[2] F. SCHUBERT, U. BRUCH, R. COOK, H. DIEHL, P.J. ENNIS, W. JAKOBEIT, H. J. PENKALLA, E. TEHEESEN, G. ULLRICH, "Creep rupture behaviour ofcandidate materials for nuclear process heat applica-tions", Nucl. Tech. 66 (1984), 227 - 240

[3] H. KIRCHHOFER, F. SCHUBERT, H.NICKEL, "Precipitation behaviour of Ni-Cr-22Fc-19Mo (Hastelloy X) and Ni-Cr-22Co-12Mo (Inconel617) after isothermal aging", Nucl. Tech. 66 (1984),139 - 148

[4] A. ANGERBAUER, H. J. SEEHAFER, "SeminarFcsligkeitslehrc", INTERATOM, Bergisch-Gladbach,1981

[5] W. JAKOBEIT, W. BULLA, R. ECK, G. ULLRICH,"Mcchanische Eigenschaftcn und Langzcitvcrhaltenvon TZM-Schwcissverbindungen", Metall 41 (1987),476 - 480

[6] G. ULLRICH, K. KROMPHOLZ, "MechanischcEigcnschaftcn von TZM und Rcinmolybdiin unter-schiedlicher Qualitatcn", VDI Berichtc 670,1988,439- 4 4 8

50

[7] K.G.E.BRENNER,L.W.GRAHAM,"Thedevelop- [8] U. FRICKER, G. ULLRICH, H. P. ALDER, "Wearment and application of a unified corrosion model for stresses in ceramic layers at elevated temperatures un-high-temperature gas-cooled reactor systems", Nucl. der helium atmosphere". High Temp. Met. Mat. forTechn. 66 (1984), 404 - 414 Gas-Cooled Reactors Proc. Spec. Meeting, Cracow,

Poland, 20 - 23.06.1988, 196 - 200

51

Experience with the VS-Decontamination Process Developed at PSI

E. Schenker, D. Buckley, H. P. Alder, W. Francioni

Laboratory for Materials Technology and Nuclear Processes

1 Introduction

In the past decade many different decontamination pro-cesses have been proposed and introduced to reduce ra-diation fields in nuclear power plants in order to facilitaterepair and maintenance work. Most of these procedures todecontaminate pressurized water reactors (PWRs) consist ofseveral steps with oxidizing and reducing solutions. Someof these treatments need relatively high temperatures (90to 120 °C) and solutions with rather high concentrations.High temperatures make the handling difficult, and highconcentration of chemicals leads to voluminous secondarywastes.

The new VS-process (Very Soft Process) was devel-oped in order to reduce the treatment temperature and sec-ondary waste volume. The process consists of an oxidizingtreatment by a solution containing permanganic acid andchromic acid, followed by a reducing step with a solutioncontaining oxalic acid, with the possible addition of otherorganic acids, complexing agents, and corrosion inhibitorswhere necessary [1]. The permanganic acid is preparedby treating a solution of a permanganate salt with a strongcation exchanger. The solution is then completed by addingchromium-VI oxide (CrO3). In some cases a solution ofchromium-VI oxide with addition of a permanganic saltmay be suitable. Both solutions may be applied at ambientor elevated temperatures.

This process, first developed to decontaminate parts ofPWRs may also be applied with some modifications to otherreactor systems, i.e. boiling water reactors (BWRs) and toother nuclear installations. In the past year it has not onlybeen successfully applied in the decontamination of somemain coolant pumps of PWRs (Mulhcim-Karlich and Biblisin FRG), but also on the gas cleaning system of a hightemperature gas cooled reactor (THTR 300 in FRG).

2 Decontamination of Main CoolantPumps of PWRs

The first industrial application of the new decontaminationprocess was on the four main coolant pumps (RCPs) inthe Miilheim-Karlich PWR in FR Germany [21,[3], Since

the decontamination treatment was on the critical path, theprocess could not be applied at room temperature. In orderto shorten the needed treatment time, the temperature of theoxidizing step was to be increased.

The equipment used in the decontamination of a PWRpump (Fig. 1) consists of a l.S m 3 preparation tank withheating and stirring facility, where the oxidising and re-ducing solutions are dissolved prior to the treatment of thepump. A second tank with an approximate volume of 1.65m3 is used as process vessel, and is equiped with ultra-sound transducers and a heating facility. The whole systemis plumbed for the transfer of the hot oxidation and reduc-tion solutions, deionised water supply for the rinsing stageand for the transfer of contaminated solutions to the util-ity's waste treatment and disposal facility. The vessel has areinforced flange on the top to support the pump with a ringspray system around the orifice where the pump impellerpenetrates the flange. Before placing the pump unit (up-per pump housing, shaft, impeller) onto the support flange,the dose rates at different points around the impeller wererecorded.

Fig. 1: Mobile equipment used for decontamination of themain coolant pumps from the Miilheim-Karlich and Biblisnuclear power plants.

52

Oxidizing step: To prepare the solution, the chromicacid (CrO.-i) was dissolved in deoinized water and heatedto approximately 80 °C. Subsequently the potassium per-manganate (KMnO.,) was added and dissolved, then thesolution was pumped to the decontamination tank. Duringthe treatment the solution was recirculated with intermittentuse of ultrasonic transducers. During process the tempera-ture was held at approximately 70 "C. After the oxidizingtreatment, the solution was pumped to the utility's liquidwaste storage lank.

Reducing step: For this treatment the solution contain-ing oxalic acid was prepared at room temperature and thentransfcrcd to the decontamination tank. The reducing treat-ment solution was rccirculalcd and intermittently subjectedto ultrasound.

Rinsing: The pump assembly was rinsed by sprayingwith dcioni/.cd water, after draining the reducing solutionto the radioactive waste storage lank.

Process control: During all decontamination steps, sam-ples from the solution were taken at intervals to measurethe radionuclide activity and to analyse the concentrationof decontamination reagents. The nuclidcs were measuredby gamma spcctromclry, the permanganate was determinedby UV/VIS spcclroscopic methods, and the free oxalic acidcontent by measuring p ; / .

The samples may also be analysed for their iron andnickel content, thus making possible an estimate of theoverall corrosion attack.

Control samples: Beside the pump being decontami-nated a cage with non-radioactive and contaminated sam-ples of materials present in a main coolant pump (chromiumsteel DIN G-X 5CrNi 13 4), chromium plated chromiumsteel, stcllitizcd chromium steel), were inserted. The fur-ther investigations of these samples enable to measure thecorrosion attack during the decontamination treatment ofthe different materials and also to control the decontamina-tion on "standard" samples.

3 Results from Main Coolant Pumps

After the decontamination treatment the pump was inspectedvisually and the dose rales measured at the same points asprior to decontamination. Fig. 2 shows the pump after thedecontamination. Before decontamination, all water wet-ted parts of the pump were tarnished with a brown blackcolour. After the treatments, the surfaces were clean andshiny mclallic. The results of the decontamination of thefour main coolant pumps from the Miilhcim-Kelrlich powerplant arc shown in Table 1, details of the activities dis-tribution in the first two complete treatments arc shownin Table 2. The results differ from pump to pump in ac-cordance with different treatment parameters, temperature,ultrasound, treatment time, and concentrations of reagentsall of which have been altered and used in different com-binations.

Fig. 2: The clean pump wheel after the first cycle ofdecontamination treatment

Table 1: Results from RCP decontamination atMiilheim-Karlich

dose ralemSv/h

decont.factor

durationh

RCP 1before decont. 50 -100decont. cycle 1 1.5-10decont. cycle 2 1.5 - 8

8 - 3 3 1010-50 7

RCP 2before decont. 26 - 50decont. cycle 1 0.04 - 0.12 200 - 1100decont. cycle 2 0.03 - 0.11 300 - 1300

RCP 3before decont. 8 -20decont. cycle 1 0.6 - 4.8 4-25

RCP 4before decont.decont. cycle 1

18*2.2'

* measured only on the nut

4 Other Applications

The VS-process was also applied successfully on the gascleaning system of the THTR 300 in Hamm/FR Germany[3]. At this plant decontamination factors as high as 40 to300 were achieved. The process has also been successfullyimplimcnted at both Biblis A and B power plants, FRG.

53

Table 2: Dissolved activities during the decontamination ofthe reactor coolant pump (RCP 1) from MQlheim-KSrlichnuclear power plant

1. Decontamination Cycle:Isotopes Oxydation* Reduction**

Bq % Bq %

Co-58 4.8 E+10 90 1.1 E+10 86Co-60 4.9 E+9 9 1.2 E+9 8Mn-54 3.8 E+ 8 0.7 6.0 E+ 8 5Sb-124 2.9 E+ 8 0.6 3.8 E+ 8 1

2. Decontamination Cycle:Isotopes Oxydation* Reduction**

Bq % Bq %

Co-58 1.2 E+ 9 87 4.9 E+ 8 78Co-60 1.3 E+8 9 7.8 E+7 12Mn-54 6.0 E+7 1 3.7 E+7 5Sb-124 3.8 E+7 3 2.9E+7 5

* Oxydation step, solution of chromic acidand permanganate

** Reduction step, solution of oxalic acid

ReferencesHI E. SCHENKER CH-Patent Nr. 03846/87-4-2.10.1987

[2] W. HEESS, E. SCHENKER, G. PETZOLD, H.D1EWALD, "Very Soft works at room temperature",Nucl.Eng.lnt. 34 (1989) no 420, 35-36

[3] E. SCHENKER, D. BUCKLEY, H.P. ALDER, W.FRANCIONI, W. HEESS, A. CONRATH, "The VS-dcconiaminalion process", (Contribution to:) Int.Conf. Water Chemistry of Nuclear Reactor Systems5, 23. - 27.10.1989, Bournemouth

[4] P. KUBASCHEWSKI, R. STRAEHL, W. SCHULZ,E. SCHENKER, "Dekontamination einer Warme-tauschergruppe der THTR-Casreinigungsablage imeingebauten Zustand", Jahrestagung Kemtechnik '89,09. - 11.05.1989, DUsseldorf, 599-602

54

The Effects of Non-Linear Sorption on Radionuclide Transport

P.A. Smith, A. Jakob

Waste Management ProgramLaboratory for Reactor Physics and Systems Technology

Abstract

As part or the research program for the safety assessment ofnuclear waste repositories, a model has been developed atPSI for radionuclide transport in the geospherc. An impor-tant component of this model is sorption, the mathematicaldescription of which has, until now, been limited to linearisotherms. Here, we discuss the effects of the incorporationof non-linear sorption into the model; non-linear sorptionis considered a step towards greater realism, reducing theneed for conservative assumptions and predicting lower fi-nal concentrations. A simplified version of the model isapplied to a diffusion-sorption laboratory experiment andis shown to reproduce the behaviour of radionuclides withdifferent sorption characteristics.

This report summarises work described in detail in ref-erences [1] and [2].

1 Introduction

Radionuclide transport through the gcosphcre is an essen-tial component in the safety assessment of repositories forhigh-level nuclear waste in deep-lying geological forma-tions. The geological barrier forms the last of a system ofartificial and natural barriers to radionuclide migration andshould result in long transport times due to a high retarda-tion potential.

In crystalline rocks, such as granite or gneiss, radionu-clides are transported by flowing groundwatcr within con-ducting zones: fissures and fractures of essentially planarstructure or lube like veins. These form complex 3D-flowsystems which may interconnect The hydrogcological andgeochemical characterisation is based on inspection of borecores and experiments in drilled wells, which yield the flowsystems and mineral distributions responsible for sorption.Within fluid-conducting zones, radionuclides may be trans-ported much slower than the groundwater. This is of im-portance from the point of view of safety assessment sincemany of the relevant radionuclides decisive in a safely as-sessment have half-lives which are essentially shorter thanthe transport times through the geosphere; they may there-fore decay to negligible concentrations before reaching thebiosphere. The main retardation mechanisms are diffusion

into adjacent zones of stagnant water and sorption onto thesolid phase.

The partitioning of a species between the liquid andsolid phases is usually expressed in terms of a sorptionisotherm, the simplest form of which is the linear isotherm,where the partitioning is independent of concentration. Un-til now, the modelling of radionuclide transport has assumeda linear isotherm, firstly because of lack of sorption dataand secondly because the resulting system of differentialequations is linear and is therefore much easier to solve.The assumption of a linear isotherm may be appropriatewhen, for example, the concentration of a sorbing radionu-clide is well below the natural stable isotope concentration,in which case sorption is determined by isotopic exchange.However, sorption experiments have shown that linear sorp-tion is the exception rather than the rule. For reasons ofconservatism, safety assessments based on linear isothermsare carried out using low retardation factors, correspondinglo high concentrations and, in a safety assessment, this canlead to unrcalislically high total doses to man.

Several other phcnomenological relationships betweenconcentration in the liquid phase and on the solid phasehave been proposed. In this report, we describe calcula-tions in which the effects of introducing one such non-linearrelationship, the Freundlich isotherm, are investigated.

Firstly, we consider a conceptual model for the migra-tion of a radionuclidc pulse through fractured, crystallinerock. The model comprises: a fluid-conducting zone, inwhich transport is by advection and dispersion, rock ma-trix with connected pore spaces, in which transport is bydiffusion normal to the direction of fluid flow, and intactrock, in which transport is neglected. The model was devel-oped by HADERMANN and ROESEL [3] and applied tosmall-scale infiltration experiments by HADERMANN andJAKOB [4], using linear isotherms to describe sorption onthe surfaces of the fluid-conducting zone and on the pore-surfaces within the matrix. Non-linear sorption isothermsare considered lo be a step towards greater realism; their an-ticipated effects are a steeper front to the migrating peak, alonger tailing part, a decreased maximum height and greatertransport times due to the enhanced retardation. For de-caying radionuclides, where transport limes are around orgreater than the nuclide half-life, final concentrations maybe lowered by orders of magnitude.

55

Secondly, we consider a small-scale diffusion experi-ment, carried out by BRADBURY et al. [5], in which aconstant concentration gradient of different radionuclidcs,125/, *&Sr and l37Cs, was maintained across a sampleof rock. The significance of the diffusion experiment isthat it permits both a sorption isotherm and a diffusion co-efficient to be determined from a single, intact sample ofrock. In the original analysis, however, the calculated sorp-Uon was considerably below that anticipated on the basisof batch-sorption tests, with i37Cs exhibiting the greatestdiscrepancies. A further observation was that the diffu-sion coefficient for * sSr was higher by a factor of 4 or5 compared with 1 2 5 / and U7Cs. The analysis assumedthat the sorption isotherm was linear and that the experi-ment had proceeded to a steady state. For1 2 5 / and 8 55r,batch-sorption tests indicated that the former is a valid as-sumption. For i37Cs, however, batch-sorption tests indi-cated non-linear behaviour; the calculated sorption was in-terpreted as a mean value for the concentration range of theexperiment. A non-linear sorption isotherm is, therefore,more realistic representation and is included in the presentcalculations. The model of the experimental system maybe considered a simplification of that described above forfractured crystalline, in which only one-dimensional diffu-sive transport is retained. By including non-linear sorptionand by modelling the transient, as well as the steady statephase of the experiment, the aim is to resolve some of thediscrepancies arising from the earlier analysis.

2 Geosphere Transport

We describe the migration of contaminants in fractured orheterogeneous porous media by taking into account the fol-lowing mechanisms (see also Fig. 1).

1. Advective transport of the radionuclidcs at the aver-age groundwater velocity in the water carrying zones.

lmp»rm»«bl» rock

maxima)panauallon dtplh

Inclurt half-width

Imparmtibl*. rock

Inltt Gullet

Fig. 1: Geometry and relevant mechanisms for radionu-clide transport in a fracture zone.

2. Hydrodynamical dispersion, arising from the super-position of different interconnected flowpaths, char-acterised by different water velocities and migrationlengths and including molecular diffusion within theflowpaths.

3. Matrix-diffusion. Diffusion of the radionuclidcs outof the flowpaths into zones of stagnant water of thealtered rock.

4. Reversible adsorption of the radionuclides on thefracture/vein infill and onto the surfaces of the sur-rounding rock matrix.

5. Radioactive decay and, in the case of the nuclidcchain, ingrowth of the radionuclidcs.

Without retarding processes a contaminant pulse wouldmigrate with the averaged velocity of the groundwater. Hy-drodynamical dispersion affects the concentration distribu-tion of a tracer as it moves along the flowpath and is re-sponsible for its gradual longitudinal (in flow direction) andtransversal spreading. Molecular diffusion takes place in-dependently of any advective motion, but can generally beneglected with respect to dispersion. Matrix-diffusion andsorption can decrease the nuclide release to the biosphere byseveral orders of magnitude provided that the extent of thealtered zone surrounding the water-carrying fracture/vein issufficiently large.

A well know phenomenological relationship betweenconcentration in the liquid phase (C) and on the solid phase(5) for steady-state conditions is the Frcundlich sorptionisotherm:

S= K CN ; N > 0

where A' and N are empirical constants. (For N ~\ andK = Kd this reduces to the linear sorption isotherm.)

Including all the relevant mechanisms mentioned above,and assuming a non-linear (Frcundlich) sorption isotherm,we obtain a conceptual model for the migration of radionu-clides through the gcosphcrc, which is represented math-ematically by a system of coupled, non-linear partial dif-ferential equations. Such a system may be solved by asuitable numerical method (e.g. the "method of lines", thefinite clement method), yielding the space and time depen-dent concentration of radionuclidcs in the gcosphcrc.

As an example, Fig. 2 shows the effect of a Frcundlichsorption isotherm on 135Cs break-through compared withthat from a linear isotherm. Due to stronger retardationin the case of non-linear sorplion, the peak maximum ofthe migrating pulse is significantly reduced in intensity andmore delayed than in the linear sorption case. Moreover, asa consequence of the radioactive decay ( r ^ j = 2.3106 y)the second maximum is absent.

3 The Diffusion-Sorption Experiment

Due to the long transport times (some hundreds of thou-sands of years or greater) over realistic migration distances

56

Effect of linear/non-linear sorption isotherms

Time, years

Fig. 2: 135C'.s flow at a migration distance of 500 m, nor-malised to the water flow through the repository, as functionof time. (Hydrogcological parameters arc taken from [3].)The upper line is calculated with a linear sorption isothermwith a A',( of 3-10"2 n^k-g'1 [6], the lower with a Fre-undlich isotherm with K = 510~- mola3m- '*•;/"'; .V= 0.70 (sorplion on the vein surfaces has been neglected.)Adapted from [1].

and also to the inaccessibility of potential repository ar-eas, we are compelled to test our theoretical concepts on asmaller scale: field experiments on a scale of few metres(instead of hundreds of metres) and laboratory experimentson a scale of centimetres. Here, the results of a diffusion-sorption experiment on sandstone arc modelled by a one-dimensional porous medium approach. From the relevantmechanisms 1 - 5 we consider only molecular diffusionand reversible sorption onto inner surfaces of the sandstone.Sorption is described cither by a linear isotherm, in whichcase the governing equations may be solved analytically,or by a non-linear Frcundlich isotherm, where a numericalsolution technique must be employed.

The diffusion-sorption experiment was conceived as ameans of determining diffusion coefficients and sorplioncharacteristics of different radionuclidcs from a single ex-periment on an inlacl rock sample. It thus has advantagesover batch-sorption experiments. Balch-sorplion experi-ments are relatively simple to perform, but, in order lo com-plete the experiments on a reasonable lime scale, they nor-

mally use crushed material; crushing may seriously modifysorplion behaviour by disrupting the internal pore structureand creating fresh fracture surfaces. Further experimentsarc also required to determine diffusion coefficients.

In the diffusion sorplion experiment, a constant concen-tration gradient of the radionuclides 135/, S5Sr and 137Cswas maintained across the rock sample and the flux of eachradionuclide through the sample was measured. 1257 wasassumed to be non-sorbing, and the results were used toextract the porosity of ihe sample. The transport modelwas fitted lo the experimental results for S5Sr and 137Cs,using the method of least squares, allowing the diffusioncoefficients and the parameters of the sorplion isotherm tovary. The best fits, assuming both a linear and non-linearisotherms, are shown in Fig. 3 for sr 'Sr and in Fig. 4 for

0.3 -

O 0.2 -

0.1 -

-0.0.00 .20 .25.10 .15

Time, yearsFig. 3: Diffusion-sorption curve for S5S>\r is the concentration gradient across the sample andC, = /„' T{t')dt', where T is the nuclide flux into the mea-surement cell. From ref.[2].• « • * experimental data transport model with lin-ear sorption (coincides with ihe curve for non-linear sorp-lion lo within plotting accuracy) asymptote (linear-sorption model)

0.14

.00 .15 .20 .25

Time, years

Fig. 4: Diffusion-sorplion curve for VJ'Cs. From [2].• • • • experimental data transport model wilh linearsorplion transport model with non-linear sorplion.Notice especially the different asymptotic behaviour.

57

! 3 7 Cs. In the case of ssSr, the fits overlie each otherclosely and match the experimental results well. The fittedcurves approach an asymptotic steady state, which is alsoshown in the figure. In the original analysis of these results[S], it was assumed that this steady state had been reachedat the end of the experiment; this analysis therefore under-estimated both diffusion and sorption. The (linear) sorptionparameters extracted from the fitted curves agree with theresults of batch-sorpiion tests, although the high diffusioncoefficient is still unexplained. In the case of 1 3 7 Cs , thenon-linear isotherm gives an improved fit, most noticeablywhere the integrated flux C, begins to increase with timeand I37C'.s breaks through on the low-concentration side ofthe sample. Although the exponent ;V of the Freundlichisotherm extracted from the fitted curve is in reasonableagreement with batch-sorption tests, batch-sorption testsgive a consistently lower value for the constant K.

4 Discussion and ConclusionsIn modelling the migration of a radionuclide pulse throughthe geosphere, it was found that, where the transport timesof a nuclide is around or greater than its half-life, the per-centage surviving decay may be lowered by orders of mag-nitude. The final concentration is, however, highly sensitiveto the parameter values chosen for the Freundlich isotherm.There is considerable uncertainty in these parameters formost radionuclides of interest and, furthermore, the Fre-undlich isotherm is one of several phcnomenological rela-tionships which have been proposed between concentrationin the liquid phase and on the solid phase. Work is thereforecurrently in progress to incorporate general experimentalisotherms into the model.

In the future, the model will be tested using dynamicrock-core infiltration experiments. It is also foreseen to useit in modelling the on-going field migration experiments atNAGRA's Grimsel Test Site. A further extension of theconceptual model is planned, which will deal with piece-wise constant parameters along a migration path. This willallow modelling of transport through layered systems, beit a series of sediments or a single host rock with spatiallyvarying properties.

By applying a simplified form of the model to the diffu-sion-sorption experiment, it was shown that the behaviourof 85SV is adequately represented by a linear isotherm,whereas a non-linear isotherm is required to represent thebehaviour of 1 3 7 C s . This result was anticipated from batch-sorption tests, although l37Cs is sorbed less than predicted.

A probable reason for this disagreement is the disruptionof internal pore structure and the creation of fresh surfaceswhich results from sample preparation in the batch-sorptiontests.

The fitting of time-dependent solutions of the govern-ing equations to the experimental results allows parametersto be extracted from an experiment which has not reachedsteady state. Experiments of relatively short duration cantherefore be used. Furthermore, we have shown that boththe diffusion coefficient and the Freundlich isotherm can beextracted from a single experiment on an intact rock sam-ple. From an experimental point of view, this means thatdiffusion-sorption experiments are an attractive alternativeto batch sorption tests.

References[1] A. JAKOB, J. HADERMANN, F. ROESEL: "Ra-

dionuclide chain transport with matrix diffusion andnon-linear sorption", PSI-Bericht Nr. 54, VilligenPS1, November 1989 and NTB 90-05 , NAGRA,Baden, 1990

[2] P.A. SMITH: "Modelling of a Diffusion-Sorption Ex-periment on Sandstone", PSI-Bericht Nr. 53, VilligenPS1, November 1989 and NTB 90-09 , Baden, 1990

[3] J. HADERMANN, F. ROESEL: "Radionuclide ChainTransport in Inhomogeneous Crystalline Rocks - Lim-ited Matrix Diffusion and Effective Surface Sorption",EIR-Bericht Nr. 551, Wurcnlingen 1985 and NTB85^40, NAGRA, Baden, February 1985

[4] J. HADERMANN, A. JAKOB: "Modelling smallscale infiltration experiments into bore cores ofcrystalline rock and break-through curves", EIR-Bericht Nr. 622, Wiirenlingen 1987 and NTB 87-07,NAGRA, Baden, April 1987

[5] M.H. BRADBURY, A. GREEN, D. LEVER, I.G.STEPHEN: "Diffusion and permeability based sorp-tion measurements in sandstone, anhydrite and uppermagnesian limestone samples", AERE Report 11995,March 1986.

[6] Project Gewahr 1985; "Nuclear waste management inSwitzerland: Feasibility studies and safety analyses",NGB 85-09, NAGRA, Baden, June 1985.

58

Complexation of Cu2+ , Ni2+ and UO2+ by Radiolytic Degradation Products ofBitumen

Z. Kopajtic, L.R. Van Loon

Waste Management ProgramLaboratory for Materials and Nuclear processes

Abstract

The influence of the radiolytic degradation products of bi-tumen on the speciation of radionuclides was studied un-der conditions which reflect those in the near field of acementitious rad waste repository. The potential complex-ation capacity of the degradation products was studied andcomplexation experiments with Cu2+, Ni2+and UO|+wereperformed. In general 1:1 complexes with logK values ofbetween 5.7 and 6.0 for Cu2 + , 4.2 for Ni2+and 6.1 forUO]+were produced at an ionic strength of 0.1 M. Thevalues are in good agreement with those found in the li-terature for oxalate complexes and thus suggest that ox-alate determines the speciation of Cu2+, Ni2+and UO2+inthe bitumen water below pH = 7. However, under the highpH conditions typical of the near field of a cementitiousrepository, competition with OH-ligands will be large andoxalate, therefore, will not play a significant role in thespeciation of radionuclides. The main conclusion of thestudy is that the degradation products of bitumen will haveno influence on radionuclide speciation in a cementitiousnear-field and, as such, should not be considered in theappropriate safety assessment models.

1 Introduction

For more than twenty years bitumen has been used worldwide to solidify intermediate- (ILW) and low-level (LLW)radioactive wastes. Currently, the spent fuel from Swissnuclear power facilities are reprocessed outside Switzer-land and the conditioned wastes will be returned for finaldisposal. A portion of this material (ca. 34001) from LLWand ILW will be bituminous [1]. A small amount (concen-trates) also comes from the nuclear power plant in GOsgen(Switzerland).

A possible disadvantage of the use of bitumen could beits organic composition. It can be degraded by radiolytic,

and microbial processes and may form potential ligands forcomplexation with radionuclides. These ligands may influ-ence the eventual fate of radionuclides, released from thesolidified waste, in the near and the far fields of a repos-itory. Although the gaseous degradation products of bitu-men have been well studied [2] [3] [4] [5] no informationis available on the water soluble degradation products andtheir complexation potentials. The present work describesthe radiolytic degradation of bitumen under near field con-ditions by external irradiation.

2 Materials and Methods2.1 Bitumen Degradation

Two types of bitumen were studied: distilled bitumen Mex-phall 80/100 and blown bitumen Mexphak R 90/30. Analiquot of the bitumen was placed in a stainless steel con-tainer and approximately 1 liter of Millipore-Q water, pHadjusted to 12.5 with 1 M NaOH, was added to it Thecontainer was placed in a Co-60 cell and irradiated untila total absorbed dose of about 5 MGy was reached (doserate: 11 kGy/h). The dose to apply (5.2 MGy) was esti-mated from the activity and composition of a bituminizedwaste product (WA-2) [6] (Fig. 1).

IE+00 IE+061E+02 1E+04Timet yeors

Fig 1: Total absorbed dose for a bituminous waste product(Cogema, Marcoule) with a specific activity of 7.4 GBq/L(/3, i) and 74 MBq/L (a) as a function of time.

59

Non-irradiated blanks were set up simultaneously withthese tests. After irradiation, the water was collected andfiltered through a 0.45 ;im nominal pore size membranefilter and the solutions were stored in glass vials at 4°C.

2.2 Quantitative and Qualitative Analysis ofthe Degradation Products

The total organic carbon (TOC) in the solutions was mea-sured by a combined UV- promoted persulphate oxida-tion method (Dohrmann DC-180). The inorganic carbon(carbonate) was analyzed by acid-base tilralion. The qua-litative analysis of the degradation products in irradiatedand non-irradiated water samples were carried out by GC-MS (EMPA; Diibcndorf). Some of the carboxilic acidshave been determined quantitatively by ionchromatography(Dionex 2010i).

2.3 Complexation of Cu2+by Bitumen Degra-dation Products

For a 1:1 complex, the reaction between a metal M andligand I. can be written as:

M + i ^ ML (1)

The tilralion behaviour can be derived from the stabilityconstant

[ML]K '

and the mass balance

CM =

(2)

(3)

:] + {UL] (4)

•KCL)+K{M]-K[M]

(5)

where:

CM

CLtotal concentration of metal M in solutiontotal concentration of ligand L in solution

25 cm3 of the bitumen water (described in 2.1) wereplaced in a 50 cm3 titration vessel. The pH of the solutionwas adjusted to 3.5 with concentrated HCIO,, after whichthe system was purged with prewettcd nitrogen for 5 min toremove CCK. Afterwards the pH was finally adjusted to thaidesired with 1 M NaOH and the ionic strength was adjustedto 0.1 M with NaClO.,. The solution was titrated with a0.005 M Cu(NO3)2 solution. After eaci, addition of Cu thepH of the system was measured and kept constant automati-cally with 0.01 M NaOH and 0.01 M HC'1O4 al the desiredpH. The uncomplexcd Cu2+in solution was measured bya Cu-ion sensitive electrode (Cristal membrane electrodeMctrohm) in combination with a Ag/AgCI reference elec-trode (Metrohm). All manipulations were performed at 25°C under a nitrogen atmosphere and were controlled by atitroprocessor (Titroprocessor 670, Metrohm).

Additionally, some titrations with pure oxalate solutions(Na-oxalate 5 ppm) were performed, under the same expe-rimental conditions.

2.4 Complexation of Ni2+and UOj+by Bitu-men Degradation Products

No ion-sensitive electrodes are available for Ni-+ andUOij+ and so other methods have to be used for studyingthe complex formation between these metals and organicliganJs. Ion exchange is a method which has been appliedsuccessfully to the study of complex formation in solution.The technique was developed by Schubert et al. [7] |8] |9|.

The working equation of the Schubert method can beexpressed as:

log GH- logI\ML + »tog[L] - log.A (6)

where A'J

l<d

[L]

A

partition coefficient of M in the ab-sence of ligand Lpartition coefficient in the presence ofligand Lstability constant of complex ML,,free ligand concentrationcorrection term when a buffer, B, withcomplexing properties has been used

Measuring the partition coefficient of tracers of metalM in the presence and absence of different concentrationsof the ligand L, and then plotting log I jfi - 1J against log

[L] yields a straight line. The slope n of the line representsthe stoichiomctry of the complex while the intercept withthe y-axis, log KA/£, can be derived after correction forlogA. The method works only when the formed complexis neutral or negatively charged and docs not adsorb on theion exchange resin.

2.4.1 Preparation of the Na-Resin

10 g of a wet Dowex 50 W X-4 cation exchange resin inthe H+-form (= 21 meq) were converted to the Na-formby washing the resin twice with 250 cm3 0.1 M NaOH andequilibrating 3 limes with 0.1 M NaClO^. After equilibra-tion, the resin was air dried and stored in a glass vial.

2.4.2 Complexation Experiments with Ni'-'+, UO;+andOxalate

About 100 mg of the Na-rcsin were transfcred to 50 cm'1

centrifuge tubes. 25 cm3 of a solution containing 0.12M NaClO.(, 0.12 M NaOAc and different oxalate amountswere added. The volume was then adjusted to 30 cm3 with5 cm3 of a 6 x 10"6 M Ni2+solulion, spiked with Ni-63, ora 8.4 x 10~7 M UO.]+(U-233) solution. The final compo-sition of the solutions was: 0.1 M NaClO,,, 0.01 M NaOAcand, in the case of Ni2+, 10~6 M Ni and 0, 2,4, 8,16 ppmOxalalc. In the case of U, the concentration of UOi;+was

60

1.4 x 10~7 M and oxalate varied between 0 and 2 ppm. Thesolutions were buffered at pH = 5 by the NaOAc to preventhydrolysis as a side-reaction. The resin/solution mixtureswere placed in an cnd-ovcr-cnd shaker and equilibrated for20 hours al room temperature. After equilibration, Ni-63and U-233 in solution were measured by liquid scintilla-tion counting (Packard, Tricarb 2250 CA) using Instagel(Packard) as a scintillation cocktail. The partition coeffi-cients of Ni and U (K°d and Kj) were calculated from thedifference in radioactivity in solution before and after equi-libration.

2.4.3 Complexation Experiments with Ni'-'+, UOii+andBitumen Degradation Products

Different amounts of bitumen water (10, 15 or 20 cm3 inthe case of UO;+; 35 or 70 cm3 in the case of Ni-+) wereadjusted to pH = 3 with concentrated HCIO4 to remove anycarbonates. The pH was then adjusted to 5 and a NaOAcbuffer and NaC104 added. The volume was finally made up10 100 cm3. The composition of these solutions was 0.012M NaOAc and 0.12 M NaClO^. The oxalatc concentrationcould be calculated from the dilution. With these mixtures,complexation experiments with Ni3+and UO?,+were per-formed as described in 2.4.2.

3 Results and Discussion

3.1 Composition of the Bitumen Water

Organic matrices arc generally very susceptible to radia-tion influences and radiation effects are strongly dose ratedependent. Short-term irradiations al high dose rates arcoften of limited value for the evaluation of long-term irra-diation effects. In reality, the radiation will mainly comefrom internal sources and the bitumen matrix will be ex-posed to much lower dose rates than in present work. Thus,the interpretation of the results obtained from tests on thesamples irradiated with high external dose rates must betreated with caution.

Most products found in the solution arc fatty acids. Thisis not surprising because fatty acids arc abundant in bitu-men. Their concentration in solution, however, is very low.These fatty acids, in fact, cannot be considered as bitumendegradation products. Beside fatly acids, carbonate andfour low molecular weight compounds arc present: oxalicacid, phthalic acid, acclylacciic acid and the dibutyleslcrof phthalic acid. The dibulylcslcr of phlhalic acid is oftenused as a softening agent in plastic and can be consideredas a contaminant. Oxalate is probably a reaction productof two • COO" radicals formed by decarboxylation of fattyacids.

The concentration of oxalatc in the solution depends onthe absorbed dose. Fig. 2 shows the relationship betweenthe amount of oxalale produced per gram of bitumen, andthe total absorbed dose. The higher the dose, the higher Ihcamount of oxalate produced. The saturation level will bereached at a dose larger than 6.0 MGy. The concentration

0.03

IJ> 0.02

0.01

8 to0 1 2 3 4 5 6Don, MGy

Fig 2: Amount of oxalate produced per gram of bitumenas a function of the absorbed dose (dose rate: 1 lkGy/h).

of oxalatc in the different bitumen waters is given in Table1. Together with the TOC and carbonate concentration. Alarge amount of carbonate is also present in solution, pre-sumably because irradiation of bitumen produces CO?. Theso produced CO2 is scrubbed by the alkaline solutions andreacts to COJ}~. With ionchromatography, two additionallow molecular weight products - acetate and formiale havealso been detected.

Table 1: Concentration of oxalate, TOC and inorganic car-bon in irradiated (B) and non-irradiated (U) bitumen water(n.d. = no data).

Bitumenwater

BlB2B3UlU4B4

Oxalatemg/1

5.43.37.6

<0.1< 0.1

5.2

TOCmg/1

6.04.510.10.91.66.6

CO5- + HCO3meq/1

10.6n.d.20.61.72.014

pH

9.112.510.712.712.711.8

3.2 Complexation of Cu-+by Bitumen Degra-dation Products

In order to provide an insight into the complcxalion be-haviour of the degradation products, tests with Cu-+hasbeen used because the complcxalion capacity can be easilyestimated by titration where the free, uncomplcxed Cu2+ismonitored by a Cu ion-sensitive electrode. Fig. 3 displaysthe titration curve of the irradiated bitumen water (B2) andFig. 4 a pure oxalate solution (5 ppm) with Cu2+ . Thefree Cu'-'+concentration (pCu) is given as a function of thetotal Cu concentration (pCu(0(). A straight line representsthe case where no complcxation occurs.

61

3.5-

4-

! 4.5

I 5

6

6.5

7

+ Eaperimento.1 points

— Calculated by equation 5

5.5 4.5 4- log (Cul,Ql.

3.5

Fig 3: Titration curve of irradiated bitumen water (B2) withO r + (pH = 6; I = 0.1 M; T = 25 °C).

3.5

2I

5.55.5

+ Experimentalpoints

— Colculoted

3.55 4.5 4- log[Cu],ot.

Fig S: Titration of non-irradiated bitumen water (III) withCu 2 + (pH = 5.0; I = 0.1M; T = 25 °C).

+ Experimental points

— Calculated by equation 5

7.54—5.5 4.5 3.5

Fig 4; Titration curve of pure oxalatc (5 ppm) withCu 2 + (pH = 6; I = 0.1 M; T = 25 °C).

The experimental data can be well described by equa-tion (5) with logK = 5.85-6.00 and l o g Q = - 4.45 for theB2 solution and logK = 5.65 and logCL = - 4.10 for the B3solution. This means that only one ligand in the solution isimportant for the complexation of Cu 2 + . The other ligandseither form much weaker complexes with Cu2 +or their con-centration is low and their contribution can be neglected.From the composition of the solution it is clear that onlyoxalatc can contribute to the complexation of C u 2 + . A di-rect comparison of the stability constant with the literaturedata was not justifiable because of the large spread observedin the reported values. Therefore, litrations of pure oxalatcsolution were performed. The stability constant found forthe pure oxalatc solution under the same conditions (logK= 5.82 ± 0.17) is similar to the one observed for the bitu-men water (B2 and B3). This verifies the fact that Cu2 + ismainly complcxcd by oxalatc in the bitumen water.

Fig. 5 displays the litralion of the non-irradiated bitu-men water (Ul) (contact lime: 3 months) with C u - + . Thetilralion yields a straight line meaning that no complcxalionof CuJ+and degradation products can be observed. This isconsistent with the analytical results in Table 1 (no oxalatein the solution).

3.3 Complexation of Ni2+and UO^+by Bitu-men Degradation Products

To verify the Cu-results additional studies with safety re-levant nuclidcs (N i 2 + , UOij+) have been performed.

Fig.6 and 7 show the results of the complcxation study

S'

0.8

0.6

0.4

0.2

I 0

= -0.2

-0.4

-0.6

I + Oxalate (experimental points)

I O bitumen degradation products

Uexpholl 90/30

3.9 3.8 3.74.7 4.6 4.5 4.4 4.3 4.2 4.1- '09 [L]

Fig 6: Complexation of Ni2 +with oxalate and bitumendegradation products at pH = 5.0 and I = 0.11 M (Schubertmethod).

0.8

0.6

0.4

iI o

-0.2

-0.4

-0.6

Mexpholt 80/100

Mexpholt 90/30

+ Oxolote (experimental points)

0 Bitumen degradationproducts

5.8 5.6 4.8 4.65.4 5.2

- log [LJ

Fig 7: Complexation of UOr,+with oxalate and bitumendegradation products at pH = 5.0 and I = 0.11 M (Schubertmethod).

62

of Ni2+and UO?,+with pure oxalate and with bitumen degra-dation products. The regression has been calculated for thepure oxalatc solution. It was assumed a priori, that ox-alalc is the most important ligand present in the bitumenwater. The plot of log (j£ - l\ versus log[X] results ina straight line, the slope of the line equals 1 for both Ni-oxalate and UO2-oxalate indicating that a 1:1 complex hasbeen formed:

= NiOx° (7)

UOn+ + Ox-- ^UO.Ox" (8)

The stability conslants of the complexes above havebeen calculated by applying equation (S) to the individualexperimental data, with n = 1. This procedure gives a moreaccurate value of logK than the extrapolation method be-cause the latter involves extrapolation to L concentrationswhich are 4 to 5 orders of magnitude higher than the Lconcentration range covered by the experimental measure-ments [10]. The values of the stability conslants (logK.)have been summarized in Table 2 together with some datafound in the 'literature.

Table 2: Comparison of experimental stability constants ofcomplexes of oxalate, with Ni=+and UOj"1", at pH = 5 andI = 0.11 M with literature values(* corrected for I = 0.1 M by the Davics equation [14]).

Ni2+ UO;

Oxalate 4.27 ±0.05 6.10 ±0.11Bitumen water 4.38 ± 0.02 6.09 ± 0.11Mexphak 80/100Bitumen water 4.17 ± 0.02 6.13 ± 0.11Mexphalt 90/30Literature 4.32 [11] * 5.96 [11] *

3.83 [12] 6.36 [121 *

It is evident from Fig. 5 and 6 and the rcsulls in Table2 that the experimental data are completely consistent withthat of the pure oxalate system. It can therefore be con-cluded that the a priori assumption was correci: oxalate isthe most important ligand present in the bitumen water. Thenon-irradiated samples, Ul and U4, display some degree ofcomplcxation of Ni2+and UO?,+ . The complexing capacityof this system was equivalent to a solution containing 0.1- 0.3 ppm oxalate.

4 Speciation of Ni and U in CementPore Water (SPP)

In order to get an idea of the influence of bitumen degrada-

tion products on the behaviour of radionuclidcs released tothe geosphere from the near field, a speciation calculationwith the PHREEQE gcochemical code has been performed[13]. A modified thermodynamic database has been used[14]. The data for meial-oxalale interactions were selectedfrom the PSI-MINEQL-database. The concentrations ofNi and U were set arbitrarely at 10~10 M. The oxalatc con-centration was varied between 10~3 M and 10"9 M.

Table 3: Specialion of Ni and U in a cement pore water inequilibrium with calcite at three different levels of oxalatcconcentration (Ni: 10-10M; U: lO"1" M).

Oxalatc

Species

Ni(OH)JNi(OH)2"UO2(OH)SOx2"CaOx

10"9 M

«

72.027.910070.729.3

10-'3 M

%

72.127.910070.729.3

10-:l M

%

72.127.910073.226.8

A Swedish standard Portland cement pore water (SPP)was chosen as reference. Such a water would be repre-sentative of that in the near field of a low and intermediatelevel waste repository for a period of hundreds of years afterrepository closure [14]. The results summarized in Table3 shows that the presence of COrJ" and oxalatc in solu-tion has no influence on the speciation of Ni and U. Thespeciation of both elements in this alkaline environment iscompletely determined by the OH~ anions.

5 Conclusions

The biluminization of low- and intermediate-level radioac-tive wastes operates satisfactorily world-wide. However,bituminization is still debatable because of the organic na-ture of the material, which could supply potential complex-ing agents to the near-field. The main objective of lhcpresent work was to study the radiolytic degradation of bi-tumen under near field conditions and to get some availableinformations about the water soluble degradation productsand their complcxalion potentials.

Oxalate and COj" are the main radiolylic degrada-tion products of bitumen and arc the only compounds thatform relatively stable complexes with heavy mcuils and ra-dionuclides below pH = 7. However, calculations with thePHREEQE-code show that under the high pH conditionsin a cement pore water (representing near field conditionsfor a LLW/1LW repository) the degradation products do notinfluence the spccialion of Ni and U, so lhal they can beneglected for the safety assessment studies.

63

Acknowledgements

The authors would like to thank G. Heyer and H.P. Un-der for their assistance with the experimental work andDr. Mutter (EMPA Dubendorf) for the GC-MS analysis.We acknowledge very much the discussions with Prof. R.Grauer, Dr. B. Baeyens and Dr. U. Berner. We also liketo thank Dr. R. Alexander for his critical view on the ma-nuscript. Partial financial support by NAGRA is gratefullyacknowledged.

References

[1] "Nuklcare Entsorgung Schweiz; Konzept und Ubcr-sicht Uber das Projekt Gewlihr", NAGRA NGB 85-01(1985)

[2] H. ESCHRICH, "Properties and long term behaviourof bitumen and radioactive waste-bitumen mixtures",S: KBS-TR-80-14, 1980

[3] Z. KOPAJTIC, D. LASKE, H.P. LrNDER, M. MO-HOS, M. NELLEN, H.U. ZWICKY, "Characterisa-tion of bituminous, intermediate-level waste prod-ucts", Mat. Res. Soc. Symp. Proc. 127, 1988, 527-534

[4] S. KOWA, N. KERNER, D. HENTSCHEL, W.KLUGER, " Untersuchungen zur Alpharadiolysc vonLAW/MAW Bitumenprodukten aus der Wiederaufar-beitung", KfK-3241, 1983

[5] S.G. BURNAY , "Comparative evaluaUon of a and7 radiation effects in a bitumenisate", Nuclear andChemical Waste Management 7 (1987) 107 - 127

[6] A. SAAS, "Synthese des rdsultats de caractcn-sation obtenus par l'EIR de Wurenlingcn et dcCEN/Cadarache sur les futs reels de bitume 80/100 dcla Cogema/Marcoule", (1982). SCECA 88/01, CEA,Cadarache, 1988

[7] J.SCHUBERT, :on exchange studies of complex ionsas a function of temperature, ionic strength and pres-ence of formaldehyde", J. Phys. Chem. 56 (1952), 113- K8

[8] J. SCHUBERT, A. LINDENBAUM, "Stability of al-kaline earth-organic acid complexes measured by ion-exchange", J. Am. Chem. Soc. 74 (1952), 3529-3532

[9] J. SCHUBERT, E.R. RUSSELL, L.S. MYERS "Dis-sociation constants of radium organic acid complexesmeasured by ion exchange", J. Biol. Chem. 185(1950), 387 - 397

[10] J. Dc BRABANDERE, "Humic acid complexalion ofeuropium in Boom Clay", Disscrtationcs de Agricul-tura nr. 177, Agricultural Department, Catholic Uni-versity of Louvain, Belgium, 1989

[11] S. KOTRBY, L. SUCHA, "Handbook of ChemicalEquilibria in Analytical Chemistry", Ellis HorwoodLimited, 1985

[12] IUPAC, "Chemical Data Series 22, Stability constantsof metal-ion complexes; part B-Organic ligands",Compiled by Douglas D. Perrin, Pergamon Press,1979

[13] D.L. PARKHURST, D.C. THORSTENSON, L.N.PLUMMER, "PHREEQE - A computer program forgcochcmical calculations", U.S. Geological Water Re-sources Division, 80 - 96, NT1S Techn. Rep. PB 81-167801, 1980, rev. 1985.

[14] B. BAEYENS, I.G. McKINLEY, "A PhrceqeDatabase forPd, Ni and Se", PSI-BcrichtNr. 34,1988

64

Local Monitoring Technique in Acoustic Emission at PSI

R. Attingei", B. Tirbonod', L. Hanacek", E. Schneider*, K. Wulle*, M. Emmenegger'

• LWR-Safcty Programt Department of Electro Engineering

Abstract

In reactor technology, acoustic emission is conventionallyused Tor the global surveillance of reactor components. Re-gions of activity can be detected but the precise position ofthe sources and their identification remain uncertain due tousually too large distances between flaw and transducers.We have developed in the last years the local monitoringtechnique of a known crack [1] to circumvent these short-comings, i.e. to locate the position of the sources three-dimensionally und lo determine a few of their characteris-tics. The full potential of the local monitoring techniquewas realized by extending the frequency spectrum of themeasuring system from the conventional upper limit of 1MHz up to 5 MHz.

1 Introduction

The PSI measurement system was designed to measureacoustic emission signals at the in-service temperature ofa nuclear reactor pressure vessel (310 °C) in a broad fre-quency band. Transducers from Interatom, FRG, using apiezoelcment of LiNbO3 fulfil these requirements: theywork at temperatures up to 400 °C, are broadband with vari-ations of amplitude better than ± 10 dB between 500 kHzand 5 MHz and arc sensitive to displacements of 0.01pm; unfortunately, their damping seems to be insufficient.The 4-channcl measurement system includes a low-noisecharge amplifier, a high-pass filter (100 kHz), a main am-plifier and an antialiasing low-pass filter; the gain afterconnection to the transducer is fixed lo 40 dB. The triggerunit which has lo sample the events out of the continuouslymeasured signal works on the basis of a floating thresholdlevel. The signals of the events arc digitized and stored viaa transient recorder (storage length of 2048 words per chan-nel, resolution of 10 bits, sampling frequency of 20 MHz,15 cvents/s) and a computer on a magnetic tape (1 event/s)for off-line processing.

In the following, fundamentals of the local monitoringtechnique are summarized: the three-dimensional sourcelocation and the form of the signals to define recurrent sig-

nals as well as a model based on a microcrack growth todetermine some source parameters. Further, the local mon-itoring technique is applied to a fatigue test on a pressurevessel.

2 Three Dimensional Source Location

The measurement array includes 4 transducers, the first onein the epicentre of the known crack, the other ones at thevertices of an equilateral triangle around the first transduceron a radius of 80 mm (Fig. 1), This setup allows for athree-dimensional location of the position of the sources,for an improved information on the sources mainly fromthe response of the first transducer and for shielding themeasurement array from events or noise from outside thesurveyed area. The position of the sources is determinedby the differences in the arrival times of the signals. Thestart of the signal is assumed to be formed by longitudinalwaves coming directly from the source.

Accurate source location requires a precise definition ofthe differences in time. The differences in time have tobe defined in the local monitoring technique by the arrivaltimes; location by the times at peak amplitude, as it is usualwith the conventional global surveillance systems, leads to

Fig. 1: Standard setup for the local monitoring of a knowncrack with the 4 transducers T,.

65

o) Z [mm]

-V - 4 0 - 3 0

8<p

- 2 0

1

-« j

<P oo |P «

10 2 0 JO 40 J50

w

0

X [mm]

NOMINALDEPTH

b)

^P -*o - 3 0

11 'A-'.1..,

"* ' • - - -

-so -10

• • ^ ^

Z [mm]

110

. . • A

10

A

2 0 30

•—•a-

*" *"

40 60X [mm]

DEPTH

Fig. 2: Determined depth of pencil fracture experiments on a 124 mm thick plate with 6 mm cladding.a) as measured. o Measurement.b) averaged. • Average.

A Corrected average (undiffracted ray at the foot of the transducer).

unrealistic results. The initial edge of the signal has to beidentified by visual inspection in order to surely distinguishevents from noise, to eliminate errors in the automatic de-termination of the arrival times and to take into accountnumerical inaccuracies due to the digitizing of the signal.Pencil fracture tests on a 124 mm thick plate with a 6 mmthick cladding were performed in the non-noisy environ-ment of the laboratory. An accuracy in the arrival limes of± 0.15 /is, which is mainly dependent on the precision oftheir definition at large incidence angles of the wave fieldto the transducers, was achieved. This leads to a scatter ofIhc positions of the sources, which is shown for the depthin Fig. 2. An accuracy belter than ± 2 mm in ihe planarcoordinates and of ± 10 mm in depth were found in theexperiments, the theoretically determined values arc ± 3mm and ± 8 mm (Table 1). The systematic deviation fromthe nominal depth can be traced back to the incidence angle[2], Other factors which influence the accuracy of sourcelocation systematically are described in the following:

• The position of the transducers is defined in generalquite well, but it is very important to take the curva-ture of the pressure vessel into consideration.

• The phase velocity is assumed to be constant; this isvalid since the sources are localized in the near field.The phase velocity is assumed to have a constantvalue of 6 mm//js for the longitudinal wave. How-ever, it is known that the phase velocity is reducedby about 5 % with a temperature increase from room

temperature to 310 °C, the in-service temperature ofnuclear reactor pressure vessels.

• The assumption of a direct wave between source andtransducers holds for a homogeneous material. Thepresence of an open crack would diffract the wavesand increase the paths between source and transduc-ers if the direct paths transverse the crack. However,this situation cannot be simulated in the cvalulionsince the actual size of the crack can hardly be de-termined by a 4-channel measurement system. Ne-glecting diffracted waves may cause the deviation ofthe source locations apart from the crack (e.g. in Fig.6).

The accuracy is shown in Table 1 for a theoretical ex-ample. A 100 mm long and 25 mm deep crack in a 135 mmthick component is investigated; this situation is similar toihc one described below for the pressure vessel ZB2. Thestandard setup was chosen (Fig. 1) together with sourcesalong the positive J1 axis and point source as well as pointtransducers. The differences A given in Table 1 representthe maximal errors, Ac is positive for an increase in depth.

3 Form of the Signal

The form of the signal is very important for the interpre-tation of acoustic emission sources since it contains all the

66

Table 1: Accuracy of three-dimensional location of acous-tic emission sources during local monitoring of a crack (the-oretical example).

Case

Statistical error:- Arrival time ± 0.15 (is

Systematical errors:- Position of transducer 1

X\ + 1.0 mmyi + 1.0 mmzi + 1.0 mm

- Curvature with radius1000 mm

- Phase velocity reduced

by 5 %- Presence of a crack

Source at y = +0.1 mmSource at y = - 0.1 mm

Axmm

±2.8

+1.2-0 .1+2.7

+3.1

+0.1

-2 .2- 3 . 8

A.Vmm

±1.4

-0.1-0 .1-0.1

-2 .5

-0 .2

+1.5

-5.1

Armm

±7.8

+3.4-0 .1-7 .3

-9 .6

+7.0

-5 .6- 10.0

obtainable information on the source mechanism. The formof the signal is the result of successive convolutions be-tween the lime law of the source, the impulse response ofthe component with transducers and of the electrical re-sponse of the piezoelement. The signal will also be alteredduring the wave propagation due to the presence of flawsin the component.

The similarity of the forms of signals is based in ourstudies on a visual inspection; mathematical procedures likecross correlation would be more justified. The descriptionis focused on the following features: the relation betweenthe amplitudes of signal and noise at the start and at theend of the signal within the observed time window, thepresence of different frequency components, the time ofinitiation and the duration of high frequency componentsand the overlapping of different signals.

The influence of the incidence angle of the wave fieldto the transducers on the form of the signal was studied byhigh frequency content signals generated by laser pulses.The laser pulses were applied perpendicularly to the curvedsurface of one half of a cylinder with radius 100 rrm atdifferent angles to the transducer which is fixed in the centreon the plane surface. Fig. 3 shows the resulting signals forsome incidence angles. The main findings are as follows:

• The response of the transducer to a laser pulse isgiven for an incidence angle of 0 ° as a steep frontof the signal with rise time of about 95 ns. Therefore,frequencies up to 10 MHz could be expected.

• One recognizes well the arrival of the transversalwave about 14 (is after the arrival of the longitudinalwave. Also fast and lazy oscillations are observed,which at the start of the signal are mainly attributedto an insufficient damping of the transducer used andlater on to surface waves and reflected waves.

30 -

0

-30

30

0

-30

30•i

\t

-30

30

0

-30

30

0

-30

15°

30°

45°

60°

TIME

Fig. 3: Form of the signal generated by laser pulse for dif-ferent incidence angles of the wave field to the transducer.L Arrival of the longitudinal wave.T Arrival of the transversal wave.

• The form of the signals is quite similar up to an inci-dence angle of 30 °. The amplitude of the signal frontdecreases with increasing incidence angle and disap-pears for aii incidence angle of 45 °. The rise timeof the wave front increases with increasing incidenceangle.

The analysis of the form of the signal has shown that sim-ilar forms of the signals exist in the fatigue test mentionedbelow. Similar forms of the signals together with the sameposition of the source have enabled us to determine recur-rent signals.

4 Determination of Source Parame-ters

The source parameters (expansion time, expansion rate anddiameter of the source) are estimated by assuming that themechanisms can be modelled by a microcrack growth [1].This model of microcrack growth can also be applied to thecrack growth with appropriate corrections [3]. The micro-crack (isolated or situated at the tip of a crack) is described

67

Fig. 4: Mean expansion rate am of the microcrack anddiameter D of the source for measured amplitude of thewave front umaI and expansion time r as determined inthe fatigue cycles at the pressure vessel ZB2. The grid wasdetermined for sources at a depth of 100 mm.

a) During loading.• Recurrent sources.* Non recurrent sources.

b) During unloading.o Recurrent sources.+ Non recurrent sources.

Deteciability limit (after [3]).

Detectability limit by the model of microcrackgrowth.

by a continuous distribution of edge dislocation loops. Thisdescription provides for the microcrack the same stress fieldas the one calculated by linear elastic fracture mechanicsfor fracture mode I. The wave generated by the expansionof the microcrack is calculated by means of the dynamicclastic Green's function and by assuming an unboundedmedium; ihe distance between microcrack and transduceris assumed to be large.

The model of microcrack growth predicts for the lon-gitudinal wave an impulse of a duration equal to the ex-pansion lime r, its displacement u(() is proportional to ihecube of the mean expansion rate a,,, and to the square ofthe expansion time r. These results are qualitatively and

quantitatively in good agreement with experimental resultsof other authors.

The model of microcrack growth allows one to deter-mine the mean expansion rate n,,, of the microcrack formeasured amplitude of the wave front »,,,ar and expansiontime r. It was shown [1] that the expansion time r of themicrocrack for any incidence angle of the wave field to thetransducers can be estimated by the frequency content ofthe signal. The diameter D of the source is determined byD = 2 • (i,,, • r .

The results for the fatigue test mentioned below arcshown in Fig. 4. The amplitude of the wave front »,,,ax islow during loading; it does not exceed 1 pm and lies oflcnnear to the detection threshold of about 0.01 pm. The val-ues of the mean expansion rate («„, about 80 m/s) and ofthe diameter of the source ( D about 60 /im) are generallysmall as compared to the ones reported in the literature.The diameters of the recurrent sources were similar to thepropagation rate of the crack, which leaves open for thesesources a primary or a secondary mechanism, i.e. mecha-nisms related or unrelated to crack growth. At unloadingthe amplitude of the wave front umw always lays near tothe detection threshold; it appears that the values of themean expansion rate (a,,, about 50 m/s) and of the diame-ter of the source (D about 40 /<m) arc slightly lower thanat loading and generally small as compared to the ones re-ported in the literature.

Another feature can also be shown by the model of mi-crocrack growth. Il is well known that acoustic emissionsources may be detected if their expansion rate dm and di-ameter D are large enough. The detectability limit reportedin [3] is given also in Fig. 4 for an assumed penny shapedcrack; it matches to the detection threshold determined bythe model of microcrack growth and a sensitivity to dis-placements of umax = 0.01 pm.

5 Fatique Test on the Pressure VesselZB2

To demonstrate and to extend the performance of acousticemission testing as a method of detecting and classifyingflaws, six institutes performed acoustic emission measure-ments in the course of various loading tests in 1986 on themedium-sized, thickwalled pressure vessel 'ZB2' contain-ing natural flaws. Within ihe framework of this programhosted by the German 'Bundesministerium fur Forschungund Tcchnologie' the Swiss contribution [1] consisted inacoustic emission measurements by local monitoring of crack8, a crack of length 100 mm and depth 24 mm starting fromthe base material and penetrating through the cladding tothe inner surface.

The pressure vessel ZB2 had a diameter of 2000 mmand a wall thickness of 134 mm. The base material of thesegment with the flaws was made out of a fcrritic reactorsteel (A508B) and the cladding out of an austcnitic steel.Different hydrotcsts and fatigue tests at low or high tem-perature as well as tests at constant loading and high tem-

66

CLASS

1-4L5

L 7

L8

L10L11L12L«L-14E

U1U 2

u4

U7U 8

UoE1

PRESSURE AT DETECTION

HH

—e—i• • • • •*•]

hen

-e 1

NUMBER

1064

120364

18410B

32203

6637150

197

91388380

161176

3011240

2

0 5 10 IS 20 25

PRESSURE AT DETECTION [MPo]

Fig. 5: Mean pressure at detection (o), range of pressure atdetection and number of delected signals during the fatiguetest at the pressure vessel ZB2.

/., : Recurrent sources active during loading.E : Non recurrent sources active during loading.I'i : Recurrent sources active during unloading.E' : Non recurrent sources active during unloading.

pcraturc were performed. Out of these tests, Ihe fatigue testat low temperature will be described in the following. Thefatigue test included 14170 cycles at 50 °C in the pressurerange 2.5 MPa to 21 MPa with a loading rale of 14 MPa/minand an unloading rate of 70 MPa/min. The crack grew by2.6 //m per cycle (according 10 unpublished fraclographystudies performed by Siaatlichc Matcrialpriifungs- anstaltMPA, Stuttgart).

Acoustic emission measurements were performed overa total number of 4500 cycles, 4200 signals were recorded,4040 of them could be utilized for further investigations.Most of the signals were recurrent. These recurrent sig-nals arc grouped in classes, each characterised by the sameposition of their sources inside the location error and a sim-ilar form of the signal. 14 recurrent sources were found atloading and 8 recurrent sources at unloading. The classesof signals arc numbered according to their first detection;the class Un contains signals from the same position withhardly definable forms of the signal. The signals as func-tion of loading phase and pressure at detection lime is given

in Fig. 5. The signals at loading were detected (in gen-eral) at intermediate to high pressure, those at unloading atintermediate to low pressure. Signals are seldom detectedat the maximal pressure pmax or at the minimal pressurePmin of the fatigue cycle. One may conclude from Fig. 5that acoustic emission is generated disconlinuously over thefatigue cycle and in different pressure ranges according tothe loading phase. It follows further from the frequency ofdetection that this happens only once per fatigue cycle.

The sources are differently distributed along the artifi-cial region of the crack according to the loading phase (Fig.6a). The sources at loading were shifted from the centre ofIhe artificial crack towards its right end as the experimentprogressed; those at unloading were situated at the left ofthe artificial crack. The depth of the sources, except one,docs not exceed the initial depth of the crack inside thelocation error (Fig. 6b).

The signals arc described by measurement parameters(loading phase, cycle and pressure of detection; in addition,for recurrent signals, the continuous or discontinuous char-acter of the emission), signal parameters (position of thesource, frequency content) and source parameters (expan-sion time, expansion rale and diameter of the source). Thecharacterisation of the source mechanism [1] was made bycomparing these values with those from the literature. Therecurrent sources active at unloading can be related to afriction mechanism, i.e. to a contact between the faces ofcrack 8 or small cracks around it. No proposal can be madefor the recurrent sources active at loading although certainfeatures suggest a secondary mechanism, i.e. decohesionbetween asperities or between corrosion products. The di-ameter D of the recurrent sources determined by the modelof microcrack growth arc similar to the propagation rate ofthe crack.

Only one non recurrent source is located in depth out-side the artificial crack (by 19 mm at cycle 9422, Fig. 6b).Due to physical reasons and data presented in the literature,this source was attributed to a fracture of an inclusion. Itsposition agrees well with the calculated crack front and fur-thermore on the basis of linear clastic fracture mechanics,it could give evidence that the crack has grown.

6 Summary

• The local monitoring technique allows for an accu-rate three-dimensional location of acoustic emissionsources at the expense of surveilling a known crackand a limited region. An accuracy of belter than ±2 mm in the planar coordinates and of ± 10 mm indepth is typical for our local monitoring system anda 135 mm thick plaie. Factors which increase theprecision of defining the arrival limes would reducethis statistical error.

• The form or the signal undergoes no draslical dis-torsion for incidence angles of the wave field to thetransducers of up 10 30 °. The amplitude of the signalfront decreases and the rise time of the signal front

69

o) xy-plone:

ARTIFICIAL CRACK, + • ° + o

b) KZ—plone:

v •

IS

. (i I

(FERRITIC)

I CRACK AFTER 14170 CYCLES I ,f

7 - - •CALCULATED CRACK FRONT

_ — AT CYCLE 9*22

8• o

° +» • ARTIFICIAL CRACKCLADDING

(AUSTENITIC)Fig. 6: Three-dimension]!) location or acoustic emission sources during the fatigue lest on the pressure vesselZB2.

• Recurrent sources active during loading.o Recurrent sources active during unloading.* Non recurrent sources active during loading.+ Non recurrent sources active during unloading.

a) in the plunor coordinates r, y.b) in depth ;.

increases with increasing incidence angle.

• A model of microcrack growth was used to deter-mine the mean expansion rate of the microcrack andits diameter from the estimated frequency content ofthe signal and the measured amplitude of the wavefront The determined source parameters differ lilUcbetween the loading and the unloading phase.

• The frequency of detection of the signals was about1 signal per fatigue cycle at the pressure vessel ZB2.22 classes of recurrent signals were detected. Thoseat unloading are attributed to a friction mechanism;certain features suggest a secondary mechanism alsofor those at loading. Only one source was located inthe base material outside the artificial crack, this nonrecurrent source could give evidence of crack growth.

Acknowledgments

We wish to express our gratitude to Dr. R. Rosel (PLS/PS1) for his valuables suggestions and to C. Liichingcr and

Dr. F. Kottmann (LKE/PSI) for their support in performingthe experiments with laser pulses. We gratefully mentionthe financial assistance from the Swiss Nuclear Safety In-spectorate and from the Swiss Federal Office of Energy.

References

[1] B. TIRBONOD, L. HAN ACER, "Schallemission-smessungen am Druckbchaltcr ZB2 - Lokale 0 -berwachung des Fchlers 8", PSI-Bcricht Nr. 57, Vil-ligen 1990.

[2] R. O. ATTINGER, B. TIRBONOD, L. HANACEK,"Contribution to the three-dimensional location ofacoustic emission sources during local monitoring",Trans, of the 10th Int. Conf, on Structural Mechanicsin Reactor Technology, Anaheim (Ca, USA), 1989,Division G, 263-268.

[3] C. B. SCRUBY, "Quantitative acoustic emission tech-nique", UKAEA Harwell, July 1984, AERE-R 11262.

70

Residual Stresses Due to Multipass Welding in Radioactive Waste Overpacks

R. Attinger

LWR-Safcty Program

Abstract

The procedure to calculate the residual stresses due to mul-tipass welding as presented in this study consists of a suc-cession of steps, each concerning only one physical aspect.In the weld heat source model, the localized intense heatof the welds is the source of the timc-dependent heat fiuxwhich drives the temperature evolution. In this calculation,latent heal effects during phase transformation have to betaken into account. The resulting temperature transientsconstitute the loading in the mechanical pan of the prob-lem. There, a material model is proposed to overcome thenumerical problems connected with the high temperature.

1 Introduction

One of the concepts for final disposal of high-level radioac-tive waste in Switzerland consists of a mined repositoryapproximately 1200 m deep in the crystalline bedrock ofnorthern Switzerland. In order to delay the return of theradionuclides to the biosphere, and to reduce their concen-tration there to acceptable levels, this concept relies on theprinciple of multiple safety barriers. Further to the naturalbarriers (host rock and overlying sediments) the followingengineered barriers are envisaged: the waste form itself (vit-rified high-level waste), an overpack, the purpose of whichis to ensure isolation of the radionuclides from groundwa-ter for a period of at least 1000 years, and a compactedbentonite backfill within which the cylindrical overpack isplaced horizontally in the axis of the repository gallery;this backfill is aimed at reducing the transport of water anddissolved species from the waste (if the overpack shouldfail) to the host rock. Since bentonite undergoes swelling,the backfill exerts an isostatic pressure of approximately30 MPa on the overpack [1].

The present paper concentrates on the modelling of theovcrpack. It consists of a hollow cylinder with integratedhemispherical bottom and pre-assemblcd additional shield-ing; a hemispherical lid, also with prc-asscmbled additionalshielding and with a thread for a gripping device, is pressedonto the body and held in place by means of a conical threadand subsequently welded to it [2], Through the weldingprocess residual stresses arc induced in the overpack. Theyare difficult to relieve by the usual methods, due to possiblerecristallisation phenomena in the glass inside. If stress cor-

rosion cracking cannot be ruled out, it is important to knowaccurately the residual stresses. A low strength cast steelof type GS-40 was chosen as material for the ovcrpack, thereason of this choice being that the possible problems as-sociated with stress corrosion cracking arc thus minimized.

The lid of the overpack is shown in Fig. la. The 13 mmwide and 100 mm deep groove will be filled by submergedarc welding. The energy input rate through welding is 10fcW. The welding speed of 4.7 mm/s results in deposinga layer of 4 mm thickness in each pass; a pass typicallytakes 600 s. The calculations are performed on a simplifiedmodel of the overpack (Fig. lb): the lid of the overpackis represented by a cubical block with a 13 mm wide and40 mm deep groove. The two-dimensional clement meshused in the finite element calculations is shown in Fig. lc.

The procedure to calculate the residual stresses due tomultipass welding can be divided into 3 steps: weld heatsource, temperature calculation and mechanical calculation.Since the heal input into the welding is much larger than theheat produced i.e. by plastic structure changes, the lattermay be neglected. For cast steel, the austenite-martensitetransformation with big pseudo-plastic effects is also absentand therefore, a decoupling of the mechanical process fromthe thermal one is possible.

470 b) 150 c)

330 13I 1 H

13H

- i n

460

Fig. la: Lid of a radioactive waste overpack.b: Model used for the calculations.c: Two-dimensional finite element mesh.

71

2 Weld Heat Source

The non axialsymmetric three-dimensional heat source modelproposed by Goldak [3] is used in this investigation. Goldak'smodel is the most realistic one to be found in the literature;in it the heat input into the welding is effectively distributedthroughout a volume due to arc 'digging' and stirring of theweld pool. The interaction of a heat source with a weld poolis a complex physical phenomenon that at present cannotbe modelled closely yet. Therefore, direct modelling of thefluid flow phenomena is not attempted.

Goldak [3] has derived the theoretical formulation ofthe model for weld heat sources based on a Gaussian dis-tribution of power density in space (Fig. 2). The 'doubleellipsoidal heat source' model nay be adopted to submergedarc welding as well as to electro-beam welding. Calcula-tion experience with it has shown that the steep temperaturegradient in from of the heat source and the gentle one at therear can only be achieved with non axialsymmetric models.The best correspondence between measured and calculatedsize of the fusion zone is obtained when the ellipsoid size isequal to that of the weld pool [3]. The nondimensional sys-tem suggested by Christensen [4] can be used to estimatethe ellipse parameters.

The heat source is advanced with constant welding ve-locity in the ^-direction (Fig. 2). The time-dependentpower density during the crossing of the reference plane lobe investigated is taken as heat input for the temperaturecalculation.

REFERENCE PLANE

6.5 mm

Fig. 2: Double ellipsoid heat source configuration togetherwith the power density distribution functions.

3 Temperature Calculation

The method of calculating the temperature field was vali-dated in a previous study [S] for a single pass weld. It wasshown that the strategy in finite element calculation has noinfluence on the temperature field; only the way the healflow is introduced into the finite element code and the formof the specific heat capacity matrix are relevant. It couldfurther be concluded that three-dimensional calculations arenot necessary for the high welding speed considered.

HBO °C750 °C500 °C250 °C200 °C150 °C

Fig. 3: Peak temperature with extension of the fusion zone(1480 °C) and the heat affected zone (750 °C) in a 10 passwelding.

The temperature field for multipass welding is calcu-lated with the finite element code SOLVIAT [6] on a sim-plified two-dimensional model of the overpack (Fig. lc).The initial temperature of the workpiece is 20 °C. Thetemperature-dependent thermal conductivity and volumet-ric specific heat capacity of low-carbon structural steel(0.23 % C) are taken from [3]. In the liquid range above1480 °C a thermal conductivity of 120 Wm-'K" 1 is as-sumed, in order to simulate to a first approximation theheat transfer by convcciive stirring in the molten pool. Aheat of fusion of 2100 MJrrr3 and a heat of transformationof 55 MJm~3 are associated with the melting (fusion, 1480°C) and transformation (750 l1C) temperatures, respectively.The workpiecc is insulated at the bottom and just under thearc during the time the arc is playing upon the surface; acombined heat transfer coefficient 17] is assumed for iheconvective and radiative heal losses on the remaining sur-faces.

The peak temperatures calculated pass-by-pass arc givenin Fig. 3. For the first 8 passes the same si/.e of the fusionand heat affected zones is obtained. As the heal sourcereaches the surface, the power density increases for a con-slant energy input rate Therefore, greater fusion and heataffected zones result at the top of the workpicce than below.

In the first pass only the top 250 mm of the workpicccshow a temperature change due to welding. The lowerregions arc continuously healed in later passes through heatconduction. At the end of the welding after 6000 s thetemperature is increased at the top by 87 °C and at thebottom by 31 °C. At the end of each pass, the temperaturefield is nearly constant over the widih of the workpicce.The difference in temperature between end and initial slateof one pass is nearly constant for all passes.

The pass-by-pass calculation is quite time consuming.

72

The cost can be substantially reduced by the following ap-proach. For the first pass, the end state is calculated in thenormal way. Starting from this end state, the cooling curveis determined up to the next pass to be calculated. The endstale after the nth pass is evaluated by a linear combina-tion of the temperature increase in the first pass and of thecooling curve for different cooling times after each pass. Inthe calculation of the cooling curve large time incrementscan be taken, since temperature gradients are small andthermal properties arc nearly constant. Compared to thepass-by-pass method, the superposition principle appliedto 10 passes leads to errors of up to 10 % in the calculatedtemperature field, these errors being built up continuouslyin each pass. An improved result with an error of 6.7 %is obtained when the last pass is again calculated in thenormal way.

The temperature transients serve as loading in the me-chanical calculation where the time steps arc chosen to re-sult in temperature increments of 50 °C.

4 Mechanical Calculation

The finite clement code SOLVIA [6] is used to determinethe residual stress field due to the temperature transientsproduced by welding. However, the severe loading con-dition of rapid heating and cooling in combination witha sleep drop in yield strength at high temperatures resultin large gradients which pose numerical problems. Theimplementation of a user supplied material model reducesthe numerical instabilities by controlling stress and strainovcrshoLs. In this material model one evaluates the stress-strain relation for isotropic, thcrmoclaslic-plaslic materialbehaviour at the integration points, considering a von Miscsyield criterion and isotropic hardening. The material modelincorporates especially the commonly used so-called 'ra-dial return' method: the non-admissible stresses outsidethe yield surface obtained in the iteration procedure arcforced back to the yield surface radially in the deviatoricstress plane. It was shown [8] thai the implementation ofthe radial return method has small influence on the evolu-tion of the stresses and that it is important to correct theclastic strains and the plastic strains according to the stressdrop effectuated in the radial return procedure.

Different methods are known to simulate molten mate-rial by a solid, which is the standard way to implement thisbehaviour in a running code without modifications:

• Finite elements which have temperatures above themelting temperature are set inactive at liquefactionand arc reset active at solidification. Since a wholeclement is set inactive/active because of a temperatureat a single point, the transition temperature is notsharply defined. The liquefaction gives no problems;the solidification results in an abrupt and sometimeslarge jump of the displacements.

• Temperatures above a certain cut-off temperature arcreset to this temperature, i.e. the material is kept solidat a temperature just below the melting temperature

and the final drop of the yield limit may be avoided inthe heat-up phase, Ihus preventing possible numericalproblems.

• Thermal expansion above the melting temperature isprevented by setting the coefficient of thermal expan-sion to zero. This method is in principle the same asthe reset of the too high temperatures but it provokeslarge thermal strains at liquefaction. To avoid themno thermal expansion is assumed for the molten ma-terial. This is obtained if the coefficient of thermalexpansion is chosen so that its product with temper-ature is constant, i.e. equal to the product at meltingtemperature.

• 64-bit computers can handle large flexibility and lowstrength without problems. This facility was not atour disposal since we had an APOLLO workstation.

Two further assumptions were implemented:

• The hydrostatic stress is set to zero for the moltenmaterial since welding produces a pool which is opento the surface.

• The solidifying molten material is reset in a virginstate, i.e. the reference temperature at the stress freestate is assumed to be the melting temperature andthe past hardening is reset to zero.

The user supplied material model was verified [8] on thebase of two annuli which are welded together by means ofan electro-beam. The assumptions used in the mechanicalcalculations turn out to be important for the resulting shapeof the workpiece; however, they are insignificant to theresidual stress field since this is defined by the yield limitwith time-independent material behaviour.

In the literature, mechanical material properties for caststeel GS-40 at elevated temperatures and for the weld ma-terial, which is assumed to behave like the parent material,are very scarce. Therefore, uniaxial tension tests were per-formed at temperatures up to 900 °C [9] and at two differentstrain rates to evaluate the influence of cooling rate; typicalstrengths of cast steel are given in Table 1. The Poisson'sratio and the coefficient of thermal expansion were takenfrom 110],

A two-dimensional, plane strain calculation was per-formed on the simplified model of the ovcrpack (Fig. lc)with the user supplied material model. Only material non-

Table 1: Strength of cast steel GS-40 [9]

Temperature"C

strain rate: 1}0

400900

strain rate: :j0

400900

Yield limitMPa

lO-'/s25615828

U)-Vs28016232

Ultimate strengthMPa

45535642

47537658

73

Deformations

Filled groove HMagnification: 10

Effective plasticstrains

Equidistance: 2 %

Tensile stresses

^leraita _ ahyd + a«ff

Equidistance: 100 MPa

Poss 1 V -

\ J

f<<<<k.

ELASTIC

1066 MPa

Pass 2

>

•

\ J

<

<R< \

ELASTIC

15 5! 1148 MPa

Pass 5 p JS

2

::<::^--i

ELASTIC

18I78 MPa

Pass 8 pi

Imm

F

<ii

ELASTIC

Fig. 4: Deformations, effective plastic strains and tensile stresses in multipass welding calculated pass-by-pass.

74

linearities and a rate-dependent material law were consid-ered. The groove was simulated by inactive elements; thedeformed shape and not the initial one is used at the ac-tivation of these elements. Thanks to symmetry, only onehalf of the workpicce was modelled and the bottom of theworkpiccc was fixed in the axial direction.

The results at the end of a welding pass in the top-most 72 mm of the workpiecc are shown in Fig. 4 fora calculation with 8 weld passes performed pass-by-pass.The calculation cannot be performed beyond pass 8 dueto a too deformed mesh; remeshing would be necessary toovercome this problem.

The first pass produces large deformations in the bottomof the groove (Fig. 4); further passes increase these defor-mations, their maximum value, however, occurs at aboutthe same place as during pass 3. The extension of plas-tic deformations is indicated by the doited line in Fig. 4.The largest effective plastic strain is produced in the edgeof the groove at pass 1 and shifted to the centre on thesame level of the workpiecc in the later passes, thereby in-creasing its value asymptotically; the final value of 18 % isreached after pass 3. The highly strained zone is also ex-tended; however, it does not any longer reach the groove.The maximal tensile stresses were considered as the mostrelevant stress component concerning stress corrosion. Thezone with tensile stresses matches with the zone of plasticdeformations. The largest tensile stress (about 1000 MPa)occurs always about 8 mm below the groove. It is not sur-prising that stresses of the order of 1000 MPa are found inthe bulk of the material, since the tensile stress :s expressedas the sum of the hydrostatic and the deviatoric stress com-ponent. At the edge of the groove prevail tensile stressesof 500 MPa.

A shortened procedure to the above very time consum-ing calculation of the residual stresses was tested in ananalysis considering only the first, fourth and eigth pass.The results after the eigth pass are shown in Fig. 5. Thetop of the workpiece is more deformed than in the pass-by-pass calculation whereas the displacements in the initial

groove are smaller and unsteady, thus indicating that eachpass acts on his own. The strained zone has about the sameextension but at lower value as the pass-by-pass calcula-tion. The stressed zones agree in their extension and in theirmaximal value as well. At the edge of the groove prevailsmaller tensile stresses of 300 MPa and the still existingstress peak from the first pass extends over a larger domainin the shortened calculation.

5 Conclusions

• The evolution of temperature, deformations and stressesin a multipass welding can be determined in a decou-pled three stage calculation - calculation of the powerdensity by the heal source model, temperature calcu-lation and mechanical calculation.

• As the heat source reaches the surface, greater fusionand heat affected zones result at the top of the work-piece than below for a constant energy input rate.

• In the mechanical calculation, a material model wasproposed which incorporates the so-called 'radial re-turn' method. The molten material is simulated by asolid which docs not undergo thermal expansion andwhich is reset in a virgin state at solidification.

• The assumptions used in the mechanical calculationturn out to be important for the resulting shape ofthe workpiece; however, they are insignificant to theresidual stress field since this is defined by the yieldlimit with lime-independent material behaviour.

• The residual stresses obtained from the calculationsarc high, as would be expected since the materialmodel used was one of elastic-plastic behaviour with-out creep. Future work should consider the effect ofcreep.

Deformations

Filled groove I IMagnification: 10

Effective plasticstrains

Equidistance: 2 55

Tensile stresses

t̂ensile _ ahyd + aeff

Equidistonce: 100 MPa

Pass 8

rr

<i1

—j

ELASTIC

9 %

I00 MPa

Fig. 5: Deformations, effective plastic strains and tensile stresses in multipass welding determined at the end ofpass 8 where only passes 1, 4 and 8 were calculated.

75

• A simplified approach based on the superposition prin-ciple is proposed for the temperature calculation whichleads to an error in the final temperatures of 6.7 %when cost is reduced by a factor of 5. The cost wasreduced in the determination of ihe residual stressesby calculating only every 4th step; the stressed zonesagree with a pass-by-pass calculation in their exten-sion and in their maximum value as well, the tensilestresses at the edge of the groove are reduced to aboutone half, the displacements are different.

Acknowledgements

Thanks are due to Dr. O. Mercier and Dr. R. RBsel (PSI),R. Meyer (SENAP, Widen), Dr. B. Knecht (NAGRA,Baden), Dr. A. Rosselet and Dr. J. Simpson (SULZER,Winterthur) for helpful discussions and comments. Thepresented work was carried out with partial financial sup-port from the Swiss National Cooperative for the Storageof Radioactive Waste (NAGRA), Baden, Switzerland.

References

[1] "Project GewShr 1985: Nuclear Waste Management inSwitzerland: Feasibility Studies and Safely Analyses",NAGRA Report NGB 85-09, Baden (CH), 1985.

[2] STEAG Kemenergie GmbH, Motor-Columbus Inge-nieurunternehmung AG "BehSHer aus Stahlguss fur dieEndlagerung hochradioaktiver Abfalle", NAGRA NTB84-31, Baden, 1984.

[3] J. GOLDAK, A. CHAKRAVARTI, M. BtBBY, "A newfinite element model for welding heat sources", Metal-lurgical Transactions B, 15B (1984) 6, 299-305.

[4] H. CHRISTENSEN, V. DE L. DAV1ES, K. GJER-MUNDSEN, "Distribution of temperatures in arc weld-ing", British Welding Journal, 12 (1965), 54-75.

[5] R.O. ATTINGER, "Temperature field due to multipasswelding of radioactive waste containers". Res Mechan-ica, 29 No. 1, (1990), 21-30.

[6] "SOLVIA-PRE 87 user manual", SOLV1A EngineeringAB, Vasteras (S), Report SE 87-1, 1987.

[7] V. A. VINOKUROV, "Welding stresses and distor-tion", Translated from Russian, The British Library,1977.

[8] R.O. ATTINGER, O. MERCIER, "Residual stressesdue to multipass welding of radioactive waste contain-ers", Trans. 10th Int. Conf. on Structural Mechanics inReactor Technology, Anaheim (Ca, USA), 1989, VolB, 257-262.

[9] A. ROSSELET, "Ergcbnisse der Zugversuche mit GS-40", Sulzer Brothers Ltd, Winterthur (CH), 1988.

[10J F. RICHTER, "Die wictuigstcn Eigenschaften von52 Eisenwerkstoffen", Stahleiscn-Sondcrbcrichte, 8(1973).

76

Teaching Activities and Lectures

University Level Teaching

At the University of Zurich

Prof. Dr. R. Graucr

Die Endlagerung radioaktivcr Abfiille: ChemischeAspekte (SS)

Quantitativ- und insirumenlell-analytischc Molho-den (mit Ubungen) (WS)

At the Technical University of Hannover (FRG)

Prof. Dr. W. Seifritz

- Kerntcchnik III (WS)

- Hochlemperatur-Prozesswarmereakloren (SS)

Teaching Activities at other Schools and Col-leges

H.HEYCKDiplomKurs "Kemteclinik" (5. und 6. Scmeslcr) Abt.Masctrinenbau, HTL Brugg-Windisch

Lectures

S.N. AKSAN"Some model improvement and assessment for RE-LAP5/MOD2 at PSI",4th Int. Code Assessment Meeting, Belhasda, Washing-ton D.S., USA, 18. - 20.10.1989

H.P. ALDER"EinfUhrung zum Fcslkolloquium 25 Jahre Hotlabor amPSI" 17.03.1989

H.P. ALDER"Vorstcllung des Bercichs Nuklcare Encrgic dcs PSI"Informationsveranstaltung fur Joumalistcn am PSI,21.03.1989

HP. ALDER"Nuclear energy research on district heating reactors andadvanced fuels at PSI in Switzerland"Nuclear Engineering Department Texas A+M Univer-sity, 01.12.1989

H.P. ALDER, E. SCHENKER"Ubersicht iiber LWR-Dekontaminationsverfahren"SVA-Vertiefungskurs "Wasserchemie im Kernkrafl-werk"Winterthur, 22. - 24.11.1989

H.P. ALDER, R. W. STRATTON, G. LEDERGERBER,F. BOTTA"Status of advanced LMR fuel development in Switzer-land"Am. Nucl. Soc. Winter Meeting, San Francisco, 26. -30.11.1989

W.R. ALEXANDER"An overview of ihe swiss rad waste programme".Chemistry Division, ORNL, Oak Ridge, USA, 7.12.89

W.R. ALEXANDER"Migration experiment in the Grimsel test iite",SURRC, East Kilbride, GB, 7.4.89

W.R. ALEXANDER"Swiss colloid programme",MIT, Cambridge, USA, 5.12.1989

W.R. ALEXANDER, I.G. McKinley1, A.B.MacKenzie2, R.D. Scott2, J. Meyer'"Attempted verification of matrix diffusion in granite bymeans of natural decay series disequilibria",MRS Conference, Boston, USA, 30.II. 19891 NAGRA, Baden'-' SURRC, Glasgow, GB3 Univcrsitat Bern

G.T. ANLALYTIS"Analysis of NEPTUN-III reflooding experiment No.6035 by RELAP5/Mod2/36.02(5) and by modifiedcode", NEPTUN-Bcnchmark Workshop, PSI Villigcn,05.09.1989

G. BART"Characterization of TMI-2 core material at PSI, statusreport",OECD/NEA, Principal Working Group No. 2: SystemBehaviour Related to the Prevention and Control of Ac-cidents (In-Vessel), Task Group on Three Mile Island 2,Karlsruhe, FRG, 29.-30.5.I989

G. BART, E.T. AERNE, H.A. THOMI, H.U. ZWICKY"Experience in the characterization of radioactive sam-ples by secondary ion mass spectroscopy",2nd Karlsruhe Internal Conference on AnalyticalChemistry in Nuclear Technology, Karlsruhe, FRG, 5.-9.6.1989

K. BEHRINGER"Dclrending of non-stationary noise data by spline tech-niques",21st Informal Meeting on Reactor Noise (1MORN-21),Villigcn PSI, 20 - 22.09.1989

77

R. BROGLI, D. MATHEWS, S. PELLONI"Design and safety aspects of nuclear district heatingreactors",IWGGCR-Meeting, January 1989, Vienna

R. CHAWLA"Uberblick Uber das Versuehsprogramm zur FDWR-Physik", FDWR-Workshop, Karlsruhe, 28.06.1989

R. CHAWLA"The PROTEUS facility and research capabilities - areview".Meeting on Validation of Safely Related Reactor PhysicsCalculation for Low Enriched HTGRs, PSI, 04. - 06.10.1889

R.M. CORNELL"Influence of additives on the crystallization of amor-phous ferric hydroxide",AECL, Whiteshcll, Can. 13.11.89

C. DEGUELDRE"Colloid samples and characterisation".POLOS de Caldas Workshop, Saancn, 19.-20.01.1989

C. DEGUELDRE"Grimsel colloid exercise, results intercomparison",8th CoCo Club meeting, Paris, 13. - 15.02.1989

C. DEGUELDRE ct al."Grimsel colloid exercise".Seminary GAP, Uni Gen&vc, 10.05.1989

C. DEGUELDRE"Grimsel colloid exercise conclusion " and "Humic acidPSI characterization tests",9th CoCo Club meeting, Imcrlaken, 12.-I4.9., 1989

C. DEGUELDRE"Pocos de Caldas colloid: characterization results",Pocos de caldas modelling workshop, Saancn, 21.-22.9.,1989

C. DEGUELDRE"Grimscl Colloid Exercise"13. Sci. Basis Nuclear Waste Manag., Boston, USA,27.11. -02 .12.1989

J. DRE1ER

"NEPTUN-III-Expcrimcntc fUr FDWR-Notkiihlungs-thcrmohydraulik",FDWR-Kolloquium Karlsruhe, 28.06.1989

J. DRE1ER, N. ROUGE"NEPTUN-Ecnchmark-Expcriment"NEPTUN-Bcnchmark-Workshop, PSI Villigcn,05.09.1989

R. GRAUER,"Die Endlagcrung hochradioaktiver Abfailc: ChemisetteAspekte",Naturf. Ges. Zurich, 13.11.1989

J. HADERMANN, C. McCOMBIE1,1.G. McKINLEY1,P. ZUIDEMA1

"Safety Assessment of HLW Disposal in Switzerland:Lessons Learned",Intern. Symp. on the Safely Assessment of RadioactiveWaste Repositories, Paris, 9. - 13.10.19891 NAGRA, Baden

J. HAMMER"Failure data acquisition at the SAPHIR-Reactor",Meeting on Data Acquisition for Research Reactor Prob-abilistic Safety Analysis Studies, IAEA Headquarters,Vienna, 02. -04.10.1989

K. HOFER"RETRAN-03 problems occurred at the INET stabilitytest post-calculations",RETRAN-03 Working Group Meeting, Madrid, 15.-16.06. 1989

J.P. HOSEMANN"Druckenllastung von Sicherhcitstiiillcn zur Vcrringc-rung des Risikos bei Leichtwasser-Reakiorcn?",SGK-Seminar Druckenllastung in Sichcrheiishiillen vonKemkraflwerken, ETH Zurich, 15.06.1989

E. KNOGLINGER, L.A. BELBLIDIA, G. SKOFF, J.M.

KALLFELZ"Three-dimensional space-time dependent analysis of arod ejection accident in a PWR",2nd Technical Committee Meeting on Safety Aspectsof Reactivity Initiated Accidents, Vienna, IAEA, 14. -17.11.1989

Z. KOPAJTIC"PSI Bitumenprogramm 1988-1990",Workshop, Cadarache, SCECA, 20.-22.9.1989

D. LUEBBESMEYER"ISP-25 ACHILLES reflooding experiment blind-calculations using RELAP5/MOD2 codeCSNI-Int. Standard Problem 25 (ISP-25) Workshop,Winfrilh, UK, 27.-28.11.1989

D. MATHEWS"The proposed HTR-PROTEUS program",Meeting on Validation of Safety Related Reactor PhysicsCalculation for Low Enriched HTGRs, PS!, 04. -06.10.1989

78

D. MATHEWS, R. BROGLI, K.H. BUCHER, R.CHAWLA, K. FOSKOLOS, H. LUCHSINGER, G.SARLOS, R. SEILER"LEU HTR critical rxperiment program for the PRO-TEUS facility in Switzerland",IAEA Technical Committee on Gas-Cooled ReactorTechnology, Safety and Siting, Dimilrovgrad, 21. -23.06.1989

L. NECHVATAL"Thermal-hydraulic stability of a Geyser-type Reactor",RETRAN-02 European User Group Meeting, Madrid,Spain, 14.06.1989

J. SCHMIDLI"Experiments on vapour/aerosol and pool informationon rupture of vessels containing superheated liquid"First Int. Conf. on Loss of Containment, London,September 1989

R. SEILER"Experimental techniques at the PROTEUS facility".Meeting on Validation of Safety Related Reactor PhysicsCalculation for Low Enriched HTGRs, PSI, 04. -06.10.1989

R. SEILER, H.D. BERGER, H.G. HAGER"PROTEUS-Experimente zu engen Gitteranordnun-gen: Analyse der neutronenphysikalischcn Messungen",FDWR-Workshop, Karlsruhe, 28.06.1989

R.W. STRATTON, G. LEDERDERBER, H.P. ALDER"Development of advanced mixed oxide fuels inSwitzerland"IAEA Technical Committee Meeting on Recyclingof Plutonium and Uranium in Water Reactor Fuels,Cadarache France, 13. - 16.11.1989

H.-J. ULRICH"PSI colloid laboratory experiment by electroanalyticaltechniques",NAGRA Colloid Workshop, Baden, 19.10., 1989

M. ZELLER"GEYSER: A now, simple heating reactor of high in-herent safety",ASME Winter Annual Meeting 1989, San Francisco,USA, 10. - 15.12.1989

M.A. ZIMMERMANN"Stability analysis of INET loop with 2 healed sectionin operation",RETRAN-02 European User Group Meeting, Madrid,Spain, 14.06.1989

M.A. ZIMMERMANN"RETRAN-O3/RETRAN-O2 comparison for KKM TT",RETRAN-03 Working Group Meeting, Madrid, Spain,15. - 16.06.1989

G. YADIGAROGLU"Modelling of two-phase flows",4th Int. Tropical Meeting on Nucl. Reactor Thermal-Hydraulics (NURECTH-4), 10. -13.10.1989, Karlsruhe

Scientific Publications

PSI-Reports

(PSI-Berichte)

HP. ALDER, D. BUCKLEY, R. GRAUER, K.H.WIEDEMANN"Corrosion products, activity transport and deposition inboiling water recirculation systems",PSI-Bericht Nr. 45, 1989

M. ANDREANI, G. YADIGAROGLU

"Dispersed flow film boiling"(An investigation of the possibility to improve modelsimplement in the NRC computer codes for the refloodingphase of the LOCA),PSI-Bericht Nr. 51, 1989

B. BAEYENS, I.G. McKINLEY1

"A PHREEQE database for Pd, Ni and Se",PSI-Bericht Nr. 34, 1989 (NAGRA NTB 88-28, Baden1989)1 NAGRA, Baden

G. BART (Editor)"Jubilaums-Jahresbericht Hotlabor 25 Jahre",PSI Bericht Nr. 36, 1989

K. BEHRINGER"Detrending of non-stationary noise data by spline tech-niques",PSI-Bericht Nr. 46, 1989

S. BROSI, R. WANNER"Beanspruchung der Wand des HDR-Reaktordruckbe-halters infolge eines Langzeillhcrmoschocks, gercchnetan einem 2D-Schcibcnmodell",PSI-Bcricht Nr. 21, 1989

S. BROSI, R. WANNER, D. UHLMANN1, H. DIEM1

"Nachrechnungen zum RohrversagensversuchRORV(B)",PSI-Bericht Nr. 29, 19891 Staall. Matcrialpriifungsanslalt, UnivcrsitSl Stuttgart

79

M.H. BRADBURY (Ed.), Contributing authors: S. AK-SOYOGLU, W.R. ALEXANDER, B. BAEYENS, C.BAJO, M.H. BRADBURY, E. HOEHN, R. KEIL, M.MANTOVANI, M. MAZUREK1, J. MEYER1

"Laboratory investigations in support of the migrationexperiments at the grimsel test site",PSI-Bericht Nr. 28, 1989 (NAGRA NTB 88-23, Baden1989)1 University of Beme, Switzerland

R. CHAWLA, J. HAMMER, R. SE1LER, H.D.BERGER, R. CHRISTEN, K. DUTHALER, H.G.HAGER"Contribution to research reactor safety, operations andodificalions",PSI-Berichl Nr. 46, 1989

C. DEGUELDRE, G. LONGWORTHl, V. MOULIN-,P. V1LK.S3 et al.

"Grimsel colloid exercise"PS1 Bericht No. 39, 1989 (NAGRA NTB 90-1 , Baden19891 Harwell Laboratory, UK- Commisariat a l'Cnergie atomique, Fontenay, F3 Atomic Energy Canada Ltd., Pinawa, CD

A. FROMENTIN"Particle resuspension from a multi-layeiideposit by tur-bulent flow",PSI-Bericht Nr. 38, 1989

M. FURRER, R.C. CRIPPS, E. FR1CK"Iodine severe accident behaviour Code IMPAIR 2"PSI Bericht Nr. 25, 1989

V. GERWECK"Rewelling phenomena and their relation to intermole-cular forces between a hot wall and ihe fluid"PSI-Bericlit Nr. 42, 1989

R. GRAUER"Zur Koordinationschemie dcr Huminstoffc",PSI-Bericht Nr. 24, 1989 (NAGRA NTB 89-08, Baden1989)

F. HERZOG"Hydrologic modelling of the migration site in the grim-sel rock laboratory - the steady state",PSI-Bericht Nr. 35, 1989 (NAGRA NTB 89-16, Baden1989)

A. JAKOB, J. HADERMANN, F. ROESEL'"Radionuclidc chain transport with matrix diffusion andnon-linear sorption",PSI-Berichl Nr. 54,1989 (NAGRA NTB 90-05, Baden)1 University Computer Center, University of Basel

E. KNOGLINGER, L.A. BELBLIDIA, G. SKOFF, J.M.KALLFELZ"Three-dimensional space-time dependent analysis of arod ejection accident in a PWR",PSI-Bericht Nr. 50, 1989

O. MERCIER"Abschlussberichldes Schweizerischcn HDRprogramms- PHASE II",PSI-Bericht Nr. 30, 1989

S. OLEK"Solution to a fuel and cladding rcwetting model"PSI-Bcrichl Nr. 33, 1989

S. PELLONI, P. GRIMM, D. MATHEWS, J.M.PARATTE"Validation of LWR calculation methods and JEF-1based data libraries by TRX and BAPL critical experi-ments",PSI-Bericht Nr. 32, 1989

M. RICHNER, S.N. AKSAN"Determination of some thermal properties of the NEP-TUN heater rod and experimental test of the inverse heatconduction program INCON"PSI-Bericht Nr. 43, 1989

P.A. SMITH"Modelling of a diffusion-sorption experiment on sand-stone",PSI-Bericht Nr. 53, 1989 (NAGRA NTB 90-09, Baden)

K.H. W1EDEMANN"Korrosionsuniersuchungcn zur Entwicklung vonDckontaminalionslosungen fur Nuklearanlagen"PSI-Bericht Nr. 44, 1989

Common Reports

"Thcrmohydraulics of emergency core cooling in lightwater reactors"Edited by M.J. LEWIS, HSKContributions by G. YADIGAROGLU, S.N. AKSAN,G.T. ANALYTIS, M. ANDREANI, D. LUEBBES-MEYER, S. OLEK, in addition to other contributorsfrom other organizationsOECD/NEA, SNCI-rcport No. 161, October 1989

G. LONGWORTH1, C. ROSS", C. DEGUELDRE,M. IVANOVICH1

"Interlaboratory study of sampling and characterizationtechniques for groundwatcr colloids",Harwell Laboratory report, AERE R 13393, Feb. 1989'Harwell Laboratory, UK- British Geological Survey, Key worth, UK

80

S.M. MODRO', S.N. AKSAN, V.T. BEARTA2, A.B.WAHBA3

"Review of LOFT large break experiment"US. Nucl. Regulatory Commission Publicalion,NUREG/IA-0028, Oct. 19891 Austria Research Center, Scibcrsdorf- Idaho National Engineering Laboratory, USA3 Gesellschaft fUr Rcaktorsichcrheii, Garching, FRG

Publications in Scientific and TechnicalJournals and Conference Reports as well asother Scientific Reports

N. AKSAN"Investigations on rapid cladding-cooling and quenchduring the blowdown phase of a large break loss-of-coolant accident using RELAP5/MOD2"Proc. 4th Int. Topical Meeting on Nucl. ReactorThermal-Hydraulics (NURETH^t), Karlsruhe, 10. -13.10.1989, Vol. 214 - 220

S. AKSOYOGLU"Sorption of U(VI) on granite",J. Radioanal. Nucl. Chem. 134 (1989) Nr. 2

G.T. ANALYTIS"Suppression of 'numerical' liquid carryy-over inthe semi- and nearly-implicit hydrodynamic solutionschemes of RELAP5/MOD2 during reflooding"Proc. 4th Int. Topical Meeting on Nucl. Reactor 'Thermal-Hydraulics (NURETH-4), Karlsruhe. 10. -13.10.1989, Vol 1, 196-205

G.T. ANALYTIS"Implementation of a consistent inverted annular flowmodel in RELAP5/MOD2"Trans. Am. Nucl. Soc., 1989 Winter Meeting, SanFrancisco, TRANA060 1-792, 26. - 30.11.1989, Vol.60, 670 - 671

J.J. ARKUSZEWSKI"MCNP simulation of two GCFR deep penetration andstreaming configurations", 7lh Int. Conf. on Radia-tion Shielding, 12. - 16.09.1988, Bournemouth, Eng-land (1989), 399 - 408

JJ. ARKUSZEWSKI, J.F. JAEGER"Shielding performance of the NET vacuum vessel", 7lhInt. Conf. on Radiation Shielding, 12. - 16.09.1988,Bournemouth, England (1989), 482 - 486

R.O. ATTINGER, W. HF.ER, P. WYDLER, D.DESPREZ1, J. JOUVET1, A. ZUCCHIN1-"Thcrma! and mechanical behaviour of a subassemblyhexcan under total instantaneous conditions a code com-parison" ,

10th Int. Conf. on Structural Mechanics in Reac-tor Technology, Anaheim, California, USA, 14. -18.08.1989 Vol. E, 311 -3161 Commissariat a l'Energic Atomiquc, DRP, CENCadarache- ENEA, CRI E. Clementel, Bologna

R.O. ATTINGER, O. MERCIER"Residual stresses due to multipass welding of radioac-tive waste containers",Trans. 10th Int. Conf. on Structural Mechanics in Re-actor Technology, Anaheim (Ca, USA), 1989, Vol. B,257 - 262

R.O. ATTINGER, B. TIRBONOD, L. HANACEK"Contribution to the three-dimensional location ofacoustic emission sources during local monitoring of acrack",Trans. 10th Int. Conf. on Structural Mechanics in Re-actor Technology, Anaheim (Ca, USA), 1989, Vol. G,263 - 268

A. BILEWICZ, D. BUCKLEY, H.P. ALDER,E. SCHENKER"Studies on high temperature ion exchange for LWR-cleanup"PSI-Nuclear Energy Research Progress Report 1988, 66- 69 (1989)

A. BILEWICZ, D. BUCKLEY, E. SCHENKER.H. P. ALDER"Sorption of Radionuclides on Inorganic Ion ExchangeMaterials from High Temperature Walc-"(Contribution to:) 5th Int. Conf. on Water Chemistry ofNuclear Reactor Systems, Bournemouth/England, 23. -27.10.1989

K. BEHR1NGER, J. PINEYRO"Remarks about the Displaced Spectra Techniques" Ann.Nucl. Energy, 16 (1989), 43 - 45

H. BOCK'.J. HAMMER"Basic and applied research at the 250 kW TRIGA re-actor Vienna",Proc. of the International Symposium on Research Re-actors, Hsinchu, Taiwan, 06. - 12.12.1988; 1989, 35 -361 Atominstitut dcr Osterr. Univ., Wien

H. BOCK'.J. HAMMER"Experience from 26 years of TRIGA reactor operation"Proc. of the International Symposium on Research Re-actors, Hsinchu, Taiwan, 6. - 12.12.1988; 1989, 427 -4391 Atominstitut der Oslcrr. Univ., Wicn

81

H. BOCK1 ,J. HAMMER, K. VARGA2

"Optical inspections and maintenance of research reactortanks and their components", Nuclear Europe (1989), 181 Atominslitul der Osterr. Univ., Wien- Isotope Research Institute of the Hungarian Academyof Sciences, Budapest

H. BOCK1, J. HAMMER, G. ZUGAREK1

"Operation experience with the TRIGA reactor Vienna"10th Europ. TRIGA Users Conf., 14. - 16.09.1988,Vienna; 1 9 8 9 , 1 . 1 1 3 - 1 . 1 1 91 Alominstitut der Osterr. Univ., Wien

R. BROGLI, P. BURGSMULLER1

"Nukleare Warmequelle" SIA Schweizer Ingenieur undArchitekl, J07 (1989), 759 - 764'Gebr. Sulzer AG, Winterthur

R. BROGU, P. WYDLER"Wcitcrc neuartige Kemreakior-Konzepte"SIA Schweizer Ingenicur und Archilekt, 107 (1989), 765-769

S. BROSI, R. WANNER, D. UHLMANN1, H. DIEM1

"Fracture mechnics studies of a cracked pipe bend underin-plane loading",Trans. 10th Int. Conf. on Structural Mechanics inReactor Technology, Anaheim (Ca., USA), 1989, Vol.G, 19 - 241 Staatl. Materialpriifungsanstalt, Univcrsitat Stuttgart

M. BRUNET, E. KNOGLINGER, M.A. ZIMMER-MANN "Time-domain analysis of density-wave oscil-lations in a natural circulation loop at low quality"(Contribution to:) 6th Int. RETRAN-Mceting, Wash-ington, DC, 18. -20.09.1989

R.CHAWLA, J.HAMMER, R.CHR1STEN, K DUTHA-LER"On-line coolant Gamma-spcctroscopy for detectingsmall fuel element defects in the SAPH1R reactor" (Con-tribution to:) Int. Symp. on Research Reactor Safety,Operations and Modifications, Chalk River, Canada, 23.- 27.10.1989

R. CHAWLA, R. SEILER, H.D. BERGER, H.G.HAGER"Operational safety in the PROTEUS zero-power exper-iments for light Pu-containing LWR lattices"(Contribution to:) Int. Symp. on Research Reac-tor Safety, Operations and Modifications, Chalk River,Canada, 23. - 27.10.1989

R.M. CORNELL, R. GIOVANOLI1

"Effect of cobalt on the formation of crystalline ironoxides from ferrihydrite in alkaline media",Clays and Clay Minerals, 37, (1989), 65-701 Univcrsitiit Bern

R.M. CORNELL, R. GIOVANOLI1, W. SCHNEIDER2

"Review of the hydrolysis of iron(III) and the crystal-lization of amorphous iron(IIi) hydroxide hydrate",J. Chem. Tech. Biotechnol., 46, (1989), 115-1341 Universiiat Bern2 ETH Zurich

R.M. CORNELL, R. GIOVANOLI1, W. SCHNEIDER-"The transformation of ferrihydrite into Lepidocrocite",Clay Minerals, 24. (1989), 549-5531 Universitat Bern2 ETH Zurich

R.M. CORNELL, W. SCHNEIDER1

"Formation of Goethite from ferrihydrite at physiologi-cal pH under the influence of cysteine".Polyhedron, 8, (1989), 149-1551 ETH Zurich

R.M. CORNELL, W. SCHNEIDER1, R. GIOVANOLI2

"Phase transformations in the fcrrihydrile/cysteinc sys-tem",Polyhedron, 23., (1989), 2829-28361 ETH ZOrich2 Universiiat Bern

R.M. CORNELL, G.D. SMITH1, G.K. CRANSTOUN1

"Oxide films on iron and iron/copper alloys: An imag-ing atom probe investigation",J. Chem. Technology and Biotechnology, 44, (1989),9-17• University of Oxford, GB

J.W. DAVIDSON1, D.J. DUDZIAK1, C. E. HIGGS,J. STEPANEK"A one— and two-dimensional sensitivity and unvcr-lainty path of the AARE modular code system", 7thInt. Conf. on Radiation Shielding, 12. - 16.09.1988,Bournemouth, England (1989), 744 - 7511 University of California, Los Alamos National Labo-ratory, USA

C. DEGUELDRE, B. BAEYENS, W. GOERLICH, J.RIGA1, J. VERBIST1, P. STADELMANN2

"Colloids in walcr from a subsurface fracture in graniticrock, grimsel test site Switzerland",Gcochimica et Cosmochimica Acta 53 (1989) 6031 University Namur, Belgium2 EPFL, Lausanne, Switzerland

F. VAN DORP1 , H. GROGAN, C. McCOMBIE1

"Disposal of radioactive waste",Radial. Phys. Chem. 34 (1989) 3371 NAGRA, Baden

82

H. FIERZ, B. SIGG"Analysis of unsteady spatial temperature distributionsin SONACO Experiments"Proc. Liquid Metal Boiling Working (Ed.: H.M. Kot-towski) Directorate-General for Science, Research andDevelopment, Joint Research Centre, Ispra Establish-ment, 1989, Vol. 1, 496 - 516

K. FOSKOLOS"Die Schweizerischen Femwarmenetze"Femwarme Inlemalional, 18/2 (1989), 153 - 166

K. FOSKOLOS, R. BROGLI"Prospects of SMSNR development in Switzerland andthe influence of the evolution of district heating net-works"Proc. Post-SMiRT 2nd Seminar on Small and Medium-Sized Reactors, San Diego, Aug. 1989, III.3.1 - 3.8

K. GOTT1, E. SCHENKER"Build-up of radioactive deposits in the replacement re-circulation piping in Muhleberg", (Contribution to:) Int.Conf. Water Chenistry of Nuclear Reactor Systems 5,23. - 27.10., Bournemouth1 Studsvik Energy Tcknik, Studsvik

R. GRAUER"The corrosion behaviour of carbon steel in Portlandcement",NAGRA NTB 88-O2E, Baden, January 1988

H.A. GROGAN (Ed.)BIOMOVS Technical Report 6, "Scenario B2: Irrigationwith Contaminated Groundwaler", National Institute ofRadiation Protection, Stockholm, Sweden (May 1989)

H.G. HAGER, R. SEILER, R. CHAWLA, H.D.BERGER, R. BOHME"Die Realisierung weiterer Untcrsuchungen zur FEWR-Physik in PROTEUS", Jahrestagung Kemtcchnik 1989,DUsseldorf 9.-11.05 1989 Tagungsbcricht, DcutschesAtomforum, 1 5 - 1 8

J. HAMMER, G. ZUGAREK1, H. BOCK1

"Concepts for the renewal of the TRIGA reactor in-strumentation", 10th Europ. TRIGA Users Conf., 14- 16.09.1988, Vienna; 1989, 2.47 - 2.551 Atominstitul der Osterr. Univ., Wien

L. HANACEK"Accuracy of a three-dimensional source location by thelocal monitoring technique",(Contribution to:) 18th Meeting of the EuropeanWorking Group on Acoustic Emissions, Vienna,04./05.10.1989

W. HEESS1, E. SCHENKER, G. PETZOLD2, H. DIE-WALD3

"Very Soft works at room temperature", Nucl. Eng. Int.34 (1989) no 420, 35-361 ABB Mannheim2 RWE Essen3 KKW Mulheim-KSrlich

H. HEYCK"Leistungsregelung und Sicherheil von Kernkraftwerken(Sicherheit und Risiko)", SIA Schweizer Ingenieur undArchitekt, 8 (1989), 211 - 214

J. P, HOSEMANN, G. VARADI"Ablauf schwerer Unfillle; Quelltermstudien und Ex-pcrimentc als Voraussetzung zur Beurteilung vonStorfallmanagement-Massnahmen"Schweizerische Vereinigung fur Atomenergie (SVA),Vertiefungskurs ilber Stdrfallmanagemcnt in Kem-kraftwerken, HTL Brugg-Windisch, 19. - 21.04.1989

J.F. JAEGER, JJ. ARKUSZEWSKI"Shielding performance of the NET vacuum vessel",Proc. 15th Symposium on Fussion Technology, 19. -23.09.1988 Utrecht, Holland (1989), 1130

J.F. JAEGER, V. HERRNBERGER, W. FRANCIONI"Shielding aspects of a small district heating reactor",7th Int. Conf. on Radiation Shielding, 12. - 16.09.1988,Bournemouth, England (1989), 514 - 518

F. KASPAREC1, J. HAMMER"Operating data documentation system for a research re-,-!Ctor"10th Europ. TRIGA Users Conf., 14 - 16.09.1988, Vi-enna; 1989, 2.47 - 2.551 Atominstitut der Osterr. Univ., Wien

E. KNOGLINGER

"Simulalionsmodelle zur Transienten-Analyse der Rcak-toren in der Schweiz", Energie-Forschung 1988, Jahres-berichte der Beauftragten, Sektion EnergicforschungBEW, Bern (1988), 4.1, 1 - 5

Z. KOPAJTIC, D. LASKE, H.P. LINDER, M. MOHOS,M. NEI.LEN, H.U. ZWICKY"Characterization of bituminous, intermediate-levelwaste products".Mat. Res. Soc. Symp. Proc., Vol. 127, p. 527, Pitts-burgh 1989

P. KUBASCHEWSKI1, R. STRAEHL1, W. SCHULZ2,E. SCHENKER"Dckontamination einer Warmetauschergruppe derTHTR-Gasreinigungsanlage im eingebauten Zustand",Jahreslagung Kerntechnik '89, 09. - 11.05.1989, Diis-seldorf, 599-6021 HRB Mannheim2 THTR300Hamm

83

G. LONGWORTH1, C. ROSS2, C. DEGUELDRE, M.IVANOVICH1

"Interiaboratory study of sampling and characterizationtechniques for groundwater colloids",Report AERE 13393, Harwell, February 19891 Harwell Laboratories, Harwell, UK2 British Geological Survey, Keyworth, UK

D. MATHEWS, R. BROGLI, R. CHAWLA, P.STILLER"LEU HTR Experimente for the PROTEUS critical fa-cility" Jahrestagung Kerntechnik 1989, Düsseldorf, 09.- 11.05.1989 Tagungsbericht, Deutsches Atomforum, 7- 10

D.W. MUIR1, J.W. DAVIDSON1, DJ. DUDZIAK1,D.M. DAVIERWALLA, C.E. HIGGS, J. STEPANEK"A benchmark-problem specification and calculation us-ing SENS1BL, a one- and two-dimensional sensitivityand uncertainty analysis code of the AARE system",LA-AU-88-672 (Contribution to:) Int. Symposium onFusion Nucl. Techn., 10. - 19.04.1988, Tokyo/Japan1 University of California, Los Alamos National Labo-ratory, USA

S. OLEK"Rewetting of a solid cylinder with precursory cooling"Applied Scientific Research, 36, (1989), 347 - 364

S. OLEK"Solution to a fuel and cladding rewetting modcll"Int. Communication on Heat and Mass Transfer, J6(1989), 143 - 158

OVE ARUP & PARTNERS' (ed.), S. HENDRY1,F. LAUDE2 , W. WICK3, W. DEBRUYN4 , J.COLLADO5, R. ATTINGER (contributors)"A. directory of computer codes suitable for stress ana-lysis of HLW containers - COMPAS project".Commission of the European Communities, Nuclear Sci-ence and Technology, Bruxelles & Luxembourg, ReportEUR 12052 EN, 19891 OVE ARUP & Partners, London, UK2 Commissariat à l'Energie Atomique, Bagnols-sur-seze, F3 STEAG Kernenergie GMBH, Essen, FRG4 SCK/CEM, Mol, B5 Equipos Nucleares SA, Saniander, E

OVE ARUP & PARTNERS1 (ed.), T. KEER1,D. BROC2, R. SCHAEFER3, H. BECKERS1, J.COLLADO5, R. ATTINGER (contributors)"Stress analysis of HLW containers - preliminary ringlest exercise",

Commission of the European Communities, Nuclear Sci-ence and Technology, Bruxelles & Luxembourg, ReportEUR 12401 EN, 1989

84

1 OVE ARUP & Partners, London, UK2 Commissariat à l'Energie Atomique, Bagnols-sur-seze, F3 STEAG Kernenergie GMBH, Essen, FRG4 SCK/CEM, Mol, B5 Equipos Nucleares SA, Sanlander, E

OVE ARUP & PARTNERS1 (ed.), S. LI1, M. RYAN1,D. BROC2, R. SCHAEFER3, H. BECKERS4, J.COLLADO5, R. ATTINGER (contributors)"Stress analysis of HLW containers - COMPAS project"Commission of the European Communities, Nuclear Sci-ence and Technology, Bruxelles & Luxembourg, ReportEUR 12053 EN, 19891 OVE ARUP & Partners, London, UK2 Commissariat à l'Energie Atomique, Bagnols-sur-seze, F3 STEAG Kernenergie GMBH, Essen, FRG4 SCK/CEM, Mol, B5 Equipos Nucleares SA, Sanlandcr, E

S. PELLONI, E.T. CHENG1, M.J. EMBRECHTS2,"Self-shielding characteristics of aqueous self-cooledblankets for next generation fusion devices", Fus. Tech.,Aug. 1989, 53 - 641 General Atomics, San Diego2 Rensselaer Polytechnic Institute, USA

S. PELLONI, W. SEIFRITZ, J. STEPANEK, P.STILLER, W. GJESSER1, D. LEITHNER1

"Parameter study on water ingress in a HTR",Atomkemenergie, 53/3 (1989), 233 - 2381 HRB, Mannheim

S. PELLONI, J. STEPANEK. P. VONTOBEL"Analysis of PROTEUS phase II experiments performedusing the AARE modular system and JEF-based li-braries",Nucl. Sc. and Eng., 103 (1989) 247 - 253

S. PELLONI, P. VONTOBEL"New JEF/EFF-based MATXS-formatted nuclear data li-braries", Nucl. Sei. and Eng., 101 (1989), 298 - 301

D. SAPHIER, E. KNOGLINGER, M.A. ZIMMER-MANN"Improving the RETRAN-02 core modelling for the sim-ulation of the Rod ejection accident", (Contribution to:)6th Int. RETRAN-Meeting, Washington, DC, 18. -20.09.1989

E. SCHENKERCH-Patent Nr. 03846/87-4-2.10.1987

E. SCHENKER, D. BUCKLEY, H. P. ALDER, W.FRANCIONI, W. HEES1, A. CONRATH2

(Contribution to:) "The VS-Decontamination Process"5th Int. Conf. on Water Chemistry of Nuclear ReactorSystems, Bournemouth/England, 23. - 27.10.19891 ABB Mannheim2 RWE Muhlheim-Kärlich

W. SEIFRITZ"Nukleare Prozessw&rme und C02-Entsorgung"Jahrestagung Kemtechnik, Diisseldorf, 09. -11.05.1989,695 - 698

W. SEIFRITZ"A. new mixed fossil nuclear energy system for theproduction of electricity with zero emission of carbondioxid"Nucl. Techn. (1989), 201 - 206

W. SEIFRITZ"Melhanol as the energy vector of a new climate-neutralenergy system",Int. J. of Hydrogen Energy, J4, (1989), 717 - 720

W. SEIFRITZ"EmsorgungsmOglichkeiten von Kohlendioxid"Endbericht zum Studienschwerpunkt A.5.3 des Studien-schwerpunktprogrammes fiir die Enquente-Kommission"Vorsorge zum Schutz der Erdatmosph&re" desDeutschen Bundestages, Bonn, Oktober 1989

W. SEIFRITZ, M.J. KLAENTSCHI, G. MEGARITIS,K. LIEBER"Nuclear district heating with small heating reactors",Proc. of Alternative Energy Sources, (Ed.: T.N.Veziroglu), Hemisphere Publ. Compl. N.W., 1989, 41- 5 2

W. SEIFRITZ, J. MENNIG, O. KYMALAINNEN"Modelling the acceptance problem of nuclear energyby non-linear dissipative processes", in:Dissipative Strukturen in integrierten Syste-men, Schriflenreihe zur gesellschaftlichen Entwicklung,Hrsg. Ali B. Cambel / Bruno Fritsch / Jiirgen U. Keller,NOMOS-Verlag, Baden-Baden, ISBN 3-7890-1711-6,1989, 287 - 313

B. SIGG"Survey of SONACO 37-pin bundle experiments"(Contribution to:) 4th Int. ATopical Meeting on Nucl.Reactor Thermal-Hydraulics (NURETH-4) Karlsruhe,10. - 13.10.1989

J. STEPANEK, C.E. HIGGS"A general description of AARE: A modular system foradvanced analysis of reactor engineering", (Contributionto:) Int. Reactor Physics Conf. 1988, 18. - 21.09.1988,Jackson Hole, Wyoming

J. STEPANEK, S. PELLONI, J.W. DAVIDSON1,DJ. DUDZIAK1

"Two-dimensional analysis of LBM experiments at LO-TUS", (Contribution to:) 15th Symp. on Fusion Tech-nology, Utrecht, 19. - 23.09.19881 LANL, Los Alamos

A. STONE1, H.-J. ULRICH"Kinetics and reaction sloichiometry in the reductive dis-solution of manganese(IV) dioxide and Co(III) oxide byhydroquinone",J. CoU.Sc, 132 (1989) 509-5221 John Hopkins Institute, Baltimore, USA

R.W. STRATTON, H.P. ALDER, G. LEDERGERBER"AC-3 completion of a multi-year irradiation programmeon an advanced nuclear fuel in the fast flux test facility"PSI-Nuclear Energy Research Progress Report 1988, 53- 65 (1989)

K. THOMSEN1, C. MORANDI2, F. SOREL2, J. HAM-MER"Laser surveillance system - the engineered prototypeLASSY", Proc. of the 11th Symp. on Safeguardsand Nuclear Material Management Lucembourg 30.5 -01.06.1989, 387 - 3921 IAEA, Vienna2 CEE, Jointr. Res. Center, ISPRA-Estb., Italy

B. TIRBONOD, L. HANACEK"A new acoustic emission measurement system and itsapplication to the local monitoring of a known crack ina pressure vessel".World Meeting on Acoustic Emission, Charlotte (NC,USA) 20. - 33. 03.1989, in Journal of Acoustic Emis-sion, 8 (1989), 8 4 - 8 7

R.C. UPSTILL-GODDARD1, W.R. ALEXANDER,H. ELDERFIELD2, W. WHITEFIELD1

"Chemical diagenesis in the Tamar estuary",Contrib. Sediment. 16, (1989), 1-49,1 Plymouth Marine Lab., Plymouth, UK2 Cambridge University, GB

L.R. VAN LOON, G.M. DESMET1, A. CREMERS2

"The uptake of TcOJ by plants: A mathematical de-scription".Health Physics, 57, (1989). 309-3141 CEC, Brussels, B2 Uni Leuven, B

S. VEPREK1, F.-A. SAROTT, S. RAMBERT2,E. TAGLAUER3

"Surface hydrogen content and passivation of silicon de-posited by plasma induced chemical vapor depositionfrom silane and the implications for the reaction mecha-nism",J. Vac. Sci. Technol. A 7 (4), (1989), 2614-26241 Technische University, Munchen3 Universitat Zurich3 Max-Planck-Institut, Munchen

85

R. WANNER, S. BR0S1, R. ROESEL. O MERCIER"Influence of model geometry on stress intensity for acyclically thermally loaded nozzle corner crack",Nuclear Engineering and Design U S (1989) 1 - 6

R. WANNER, M. SCHMID, O. MERCIER, G. E.NEUBRECH1, E. HANSJOSTEN1

"Computation of temperature and stress in the HDRpressure vessel in presence of mixing effects in the coldleg and mantle downcomer",Trans. 10th Int. Conf. on Structural Mechanics in Re-actor Technology, Anaheim (Ca., USA), 1989, Vol. F,19-241 Kernforschungszentrum Karlsruhe, FRG

T. VON WEISSENFLUH, B. SIGG, T.V. DURY, H.R.FIERZ"Survey of SONACO 37-pins bundle experiments",Proc. 4th Topical Meeting on Nucl. Reactor Thermal-Hydraulics (NURETH-4), Karlsruhe, 10. - 13.10.1989,Vol. 2, 1339 - 1345

H.U. WENGER, P. WYDLER"Benchmark-Rechnungen filr einen schnellen Brutreak-tor mil JEF-1- Neutronendaten", Jahrcstagung Kerntech-nik, DUsseldorf, 9. - 11.05.1989, 79 - 82

G. YADIGAROGLU, M. ANDREANI"Two-fluid modelling of thermal-hydraulic phenomenafor best-estimate LWR safety analysis"Proc. 4th Topical Meeting on Nucl. Reactor Thermal-Hydraulics (NURETH^*), Karlsruhe, 10. - 13.10.1989,Vol. 2, 980 - 995

H.U. ZWICKY, E.T. AERNE, G. BART, F. PETRIK,H.A. THOMI"Evaluation of the radial distribution of gadolinium innuclear fuel pins by secondary ion mass spectromeiry(SIMS)",Radiochimica Acta £7, (1989), 9-12

H.U. ZWICKY, B. GRAMBOW1, C.MAGRABI2, E.T.AERNE, R.BRADLEY2, B. BARNES3 , TH. GRABER,M. MOHOS, L.O. WERME1

"Corrosion behaviour of british magnox waste glass inpure water",Mat. Res. Soc. Symp. Proc. Vol. 127, (1989), 129-1361 Hahn-Meitner-Institut, Berlin2 British Nuclear Fuels pk, Sellafield, Seascale, GB3 Swedish Nuclear Fuel and Waste Management Co,Stockholm

86