detection of discontinuities [gmaw]

7/29/2019 Detection of Discontinuities [GMAW]

http://slidepdf.com/reader/full/detection-of-discontinuities-gmaw 1/8

Contro l o f GMAW: Detec t ion o f

Discontinuities in the Weld Pool

Ultrasonic pulses transmitted into the base of the weld pool

detect discontinuities and signal parameter changes

BY N . M . C A R L S O N , J. A . J O H N S O N A N D D . C . K U N E R T H

ABSTRACT. Ultrasonic methods wereused to detect incomplete sidewall fusion,porosi ty and we ld bead geometry in m o l

ten steel pools formed during gas metalarc welding of T- joints and single-bevel V

grooves. A dual-element, piezoelectr ictransducer mounted on the base metaltransmit ted ult rasonic pulses into the region at the base of the we ld poo l . The return ing echoes were show n to be indicat ive of the geometry of the pool and ofthe presence of incomplete fusion andporos ity. Tw o analysis method s, based onray tracing and transducer field calculations, were used to conf i rm the exper i mental results. An expert system was pro posed to analyze the returning echoesand to prov ide informat ion about theweld pool to an intell igent controller thatdetermines when the weld ing process

parameters need to be changed. A non-contact ing ult rasonic system was discussed as a potential practical alternativeto the contact system used in this work.

Ba c k g r o u n d

In a research program at the Idaho Nat ional Engineering L aboratory (INEL), methods of sensing the physical propert ies ofthe weld during gas metal arc welding(GMAW) are being evaluated as part ofthe development of an automatedGMAW system. In this system, propert ies

of the weld w ou ld be cont ro l led us ing thefeedback from the sensors as input to amodel of the welding process that relatesthese propert ies to welding parameters.This paper descr ibes the port ion of thatwork relat ing to the use of ult rasound todetect weld defects as they are beingmade . Ultrasonic techniques are sho wn tobe capable of obtaining information fromthe region in and around the molten we ldpool that relates to the formation ofincomplete sidewall fusion and porosity.

N. M. CARLSON, J. A. JOHNSON and D. C.KUNERTH are with Idaho National EngineeringLaboratory, EG&GIdaho, Inc., IdahoFalls, Idaho

This information can be sent to the c o n

t roller, which then makes decisions abouthow the weld ing process should be a ltered to prevent further discont inuit iesf rom fo rm ing .

A diagram of an automated weld ingcon trol schem e is sho wn in Fig. 1. O n theleft are the various types of sensors capable of detect ing the physical propert ies ofth e w e l d . These sensors may be categor ized by type — infrared, electr ical, opt icaland ultrasonic — or according to the pro pert ies of the welding process they measure, such as pool geometry (opt ical andultrasonic), cooling rate ( infrared), dro pletformation mode (electr ical) , and discont inuities (ultrasonic).

Infrared sensors can be used to detectcooling rate, which is important for con

t rol of f racture toughness while maintaining strength in some steels (Ref. 1). Workers at Auburn University are also using infrared techniques for pool size andtracking measurements (Refs. 2,3).

Electrical signals include varia tion in current and voltage, as well as the standardDC values. Variat ions provide informationon transfer mode of droplets detachingfrom the melting wire (Refs. 4, 5). This isimpo rtant, since for many applicat ions thespray transfer m od e (Ref. 6) is requ ired forqual ity welds and the boundary betw eenthis transfer mode and other, undesirablemodes may shif t due to factors beyond

the control of the system.

K E Y W O R D S

G M A W C o n t r o lDiscont inui ty Detect ionWeld Pool Discont inuityFillet WeldsSingle Bevel VsIncomplete Penetrat ionSidewall Penetrat ionUlt rasonic MethodsWeld Test MethodsNDE Expert Systems

Optical techniques such as the INEL vision system can be used for joint t rack ingand f itup measurements, as well as forobserving the shape and size of the topsurface of the molten pool and the d r o p

let detachment (Ref. 7). Richardson atOh io State University has dev elo ped a vision system that views the pool through aspecia l hol low torch bo dy, prov id ing someof the same information as the INEL visionsystem (Ref. 8). Ultrasonic methods arealso being used to determine the pool geometry at the molten/solid interface. (Ref.9). Ini tial exper iments were per fo rmed byLott (Ref. 10) at INEL and Katz and Hardtat MIT (Ref. 11). Fenn and coworkers atBrunei University claim to have used u ltrasonic measurements to determine thedepth of jo int penet rat ion and to cont ro lthat depth by vary ing the weld ing cur rent

in submerged arc welding (Ref. 12). Unfortunately, insuff icient detail was presented in their papers to co nf irm o r repe attheir results.

The final sensor in Fig. 1, which is thetopic o f this paper, uses ultrasound to de tect discont inuit ies in the we ld po ol du r ingweld ing. The ultrasonic echoes from themolten/sol id in ter face and the moltenpool prov ide informat ion about the quali ty of the fusion zone and the molte n poo l .Unacceptable welding condit ions that canresult in porosity, incomplete sidewall fus ion or undercut are detected.

Other sensors are also being used,wh ich are not l isted on the con trol-schem ediagram. These include spectroscopic andphotometr ic measurements of the arc forpen etrat ion con trol (Ref. 13), real- time radiography for discont inuity detect ion (Ref.14), and measurements of the audiblesound emit ted by the process for droplettransfer mode detect ion (Ref. 5)

The col um n lab eled Analysis in Fig. 1 iswhere the raw sensor data are convertedinto the desired information on the process, such as cooling rate deduced fromthe raw infrared data. This information is

passed on to two intell igent modules,which use the informat ion to determinethe best method of f i l l ing the weld whi le

256-s | JULY 1990



maintaining the correct heat input to the

we ld . These modules conta in economic

constraints as well as physical ones sincethe f i l l s trategy determines the number of

passes required to fill the joint and thus the

product iv i ty o f the process.The final step is a model o f the weld ing

process, which converts the requestedheat and f i ll inputs to the physical we lding

parameters (Ref. 15). Both Hardt (Ref. 16)

and Eagar (Ref. 17) at MIT are also working on aspects o f m odel ing GM AW , in

cluding the control scheme required for anonlinear, cross-coupled system and the

physical aspects of the wire melt ing.

IR —

E l e c t r i c a l - * -

O p t i c a l - * -

Ul t r as on i c - * -

s e n s o r ;

A na l y s i s

C o o l i n g

Rate

M e t a l

Transfer

M o d e

B e a d /

Groove

G e o m e t r y

De f ec t s

B us

I

Heat

Input

S t ra tegy

Fi l l

S t ra tegy

H

G

Wi re

M e l t i ng

Con t ro l

M o d e l

I

-s

-R

-C T

-Eo

-z

Prel iminary Exper iments

An init ia l study was conducted to determine the capabil i ty of longitudinalacoustic waves to detect discontinuitycond i t ions in the mol ten w eld p oo l dur ingGMAW. Refrac ted long i tud ina l waveswere used primari ly because they can

propagate in to the mol ten weld poo l andthus al low inspection of the pool 's interiordur ing the weld ing process.

A single-element, 5-MHz piezoelectrictransducer, operated in the pulse-echomode, was mounted on a Luc i te wedgesuch that 45-deg refracted longitudinalwaves w ere generated in a 25.4-mm (1.0-in.) thick carbon steel sample. The transducer was pos i t ioned 25.4 mm from theto p edge of a 90-de g sidewall prep arationso that the 45-deg longitudinal wave wasd irec ted in to the bot tom corner o f thewe ld preparat ion and was m ove d in a l ignment wi th the e lec trode throughout the

we lding pass — Fig. 2. The other s ide of thepreparat ion was a t a 60-deg bev e l . Thetwo parts were tacked together wi th a6.35-mm (0.25-in.) steel backer bar with a4.5- to 8.0-mm (0.18- to 0.31-in.) roo topen ing a t the bot t om of the prepa rat ion.A-scan signals from two flaw generatingcondit ions (porosity and incomplete side-wall fusion) and fro m a go od we ld (Fig. 3)we re acqu i red .

Incomplete sidewall fusion wasachieved by mistracking the welding gunrelative to the joint an d simultaneously inc reas ing the c ontac t- tu be-to-w orkp iece

distance from 15.9 mm (0.63 in.) to approximately 25 mm (1.0 in.) and decreasing the wire- fee d ra te . To ach ieve porosity, a f i t t ing with a septum was added tothe argon shield gas line that pe rmi tted theinjection of a volume of air into the covergas. The f inal quality of the weld was determined by radiographic techniques forthese preliminary tests.

The first A scan in Fig. 3 shows echosignals from the benign, geometric reflectors in the weld preparation prior towelding. These reflectors include the bottom corne r (Reflection B), the sidewall

preparation (Reflection C), and internalreflections from the Lucite wedge (Reflect ion A).

Fig. 1 - Conceptual diagram of the parts of an automated GMA W system.

The second A scan in Fig. 3 shows thatwh en the root pass penetra t ion is adequate and the molten pool is al igned withthe transducer, no signif icant reflectors

are observed f rom the bot tom corner orthe sidewall preparation. Since incomplete s idewall fusion had occurred on theroot pass (third A scan), the bottom corner signal was still present because thecorner geometry was not s ignif icantly altered by the welding process. Porosity(fourth A scan) in the molten weld poolresulted in a set of dynam ically c hangingreflections that occu rred later in t ime thanthe or ig inal bot tom corner re f lec t ion . Thelast three A scans in Fig. 3 can be discriminated visually, which implies that an expert system could be developed to distinguish these three welding condit ions (Ref.

8, 18).

This in it ia l study demonstrates that ref rac ted long i tud ina l waves can prov idein format ion about two unacceptab leweld ing cond i t ions , poros i ty and incomp le te s idewal l fus ion. However, mul t ip lereflections introduced by the transducerwedge prov ide no in format ion about thewelding process and complicate signal interpre ta t ion . To unders tand the source o fthe received signals, analysis techniquesare need ed to mod el the interaction of thesound waves wi th the mol ten poo l andthe weld preparat ion. A lso, both rad iog

raphy and destructive evaluation are required to determine accurately the qualityof the weld at locations of interest. Thus,a series of addit ional experiments areplanned that wil l address these shortcomings.

Exper imental Procedures

Improvements were made in the u l t ra sonic transducer and in the evaluation ofthe results for these experiments. Themajor diff iculty in the prel iminary experiments was the presence of the echoes

from the internal reflections in the Lucitewed ge. To ove rcom e th is d i ff icu l ty , a du al-element transducer was designed and

purchased. In this transducer, tw o 12.7-

mm (0.5 X 0.5-in.) square piezoelectric el

ements were mo unte d on separate Luc itewedges that were next to each o ther but

electrically and acoustically isolated. Oneelement served as a transmitter and the

other as a receiver. The reflections from

inside the transmitt ing wedge were notobserved in the rece iv ing wedge. The

tw o e lements were ang led s light ly t ow ard

each other so that the centers of thesound beams intersect at a poin t 25.4 m m

(1.0 in.) deep in the carbon steel base

metal. This point is at the bo tto m corner

of the 90 deg weld preparat ion because

the angle of refraction was again chosento be 45 deg — Fig. 4. Both elements were

matched to have a broadband f requency

response centered a t 5 MHz.The transducer was kept in a f ixed po

sit ion and data were acquired duringwelding at 0.32 mm intervals as the elec

t rode moved a long the weld preparat ion

Fig. 2 —Single element 45 deg refracted

longitudinal transducer aligned with electrode on single bevel V-groove weld sample.

WELDING RESEARCH SUPPLEMENT 1257-s



past the transducer. An encoder was

mounted on the s ide-beam weld ing machine to measure the posit ion of the elec

trode. Thus, the posit ion of the electrode

relative to the transducer was known andthe formation and solid if ication of the

w e l d p o o l , as well as the pool itself, couldbe observed. The e lec trode pos i t ion was

recorded, the transducer was pulsed andthe echo data w ere digit ized using a co m

puter-contro l led works ta t ion (Refs . 19,20).

Two d i f fe rent we ld preparat ions were

used. The first, a s ingle-bevel V groove,

was the same as was used in the prel iminary experiments described above. The

secon d was a f i l let weld o f 25.4-mm steelo n a 6.35-mm s tee l bar. However, u l t ra-

sonically the geometries of the two preparations were identical as the transducer is

mo unted on the 90 deg corner por t ion o fbo th samples. In fact, the be vel s ide of the

groo ve spec imens o f ten was not fused tothe backing bar and did not appear in

some of the sample photos o f the destructive evaluation.

G M A W was perform ed us ing a L indes ide-beam we ld ing machine and a Cobra

mat ic wire feeder. The weld ing parameters were: currents f rom 212 to 310 A,

voltages from 22 to 28 V, travel speedsfrom 0.23 to 0.27 m/min (9 .1-10.6 in . /

min), and wire feed rates of 7.9 to 12.4

Reflections from weld preparationI prior to root pass

A= Lucite wedge multiple rellec!iois

B Bouom corner of we'c nrepa'a''onC- Mooe convertec reflection

Good sidewall penetration

-. ihillM f M M > IIIi "IWRI I MI p 'M\

m/min (311 to 488 in./min) using 1.1-mm

(0.045-in.) diameter welding wire. The average deposit ion rate was 1.4 g/s (11 l b /

hr). The spray transfer mode was used

with a contac t- tube-to-workp iece d is tance of 15.9 mm (5/8 in.) The shielding

gas was 98% argon and 2% oxygen.The flaw types generated for these ex

per iments were incomple te s idewal l fu sion and porosity. The incomplete side-

wall fusion was generated as before bymistracking, increasing the contact-tube-

to-workp iece d is tance, and reduc ing thewire feed speed. Porosity was generated

by manually fanning the cover gas. Theporosity obtained with this technique was

very gross compared to that obta ined inthe prel iminary experiments. Undercut on

the sidewall adjacent to the top of the f i l lwas made by mistracking the electrode up

the sidewall.

Destructive evaluations of the welds

were made at 26 transducer locations onweld root passes to determine the quality

o f the we ld for corre la t ion to the acqu iredultrasonic data. The welds were cut per

pendicular to the preparation in a planecontaining the transducer sound center-

l ine. The destructive evaluations showedthe geometry o f the weld poo l and the

presence of both near-and-far-s ide incomplete fusion, acceptable s idewall fusion, and poros i ty .

Ray-Trac ing and Sound F ie ld

Calculat ions

Echoes we re re ceived from the sidewallpreparation prior to the root pass and

from the weld poo l as the e lec trodepassed the transducer. During welding,the echoes f rom the preparat ion may bereduced in amplitude or changed in shapebecause the molten metal penetrates thesidewall and changes the reflectiv ity of thesidewall. To understand the sound pathsof the signals received before and duringweld ing, two methods o f ana lyz ing thedata were used to validate the arrivalt imes and relative amplitudes of receivedsignals and to d eterm ine th e sound p ath inthe experimental data.

In the f irst method, rays were traced

from the po in t where the u l t rasound en-

^ ^ l

13 2 17.6 22.2Time (JJS)

Fig. 3—A scans of weld preparation prior to andduring the root pass of good and flawed welds

acquired with a single element transducer.

Weld pool

Fig. 4 —Schematic of the center rays from the

dual-element transducer.

tered the part, to various reflectors on th epreparat ion and in the weld p o o l , an dthen to where the u l t rasound ex i ted thepart at the receiv ing transducer. A computer program was wr i t ten t o t race theserays, accounting for the refraction of theray as it ente red or ex ited the molte n po oland for the change in the angle of reflect ion upon mode convers ion (change insound mode f rom long i tud ina l to t rans

verse and vice versa). The input to thecomputer program cons is ted o f the number o f rays , in format ion on whether theray is reflected or transmitted when itreaches a bou nda ry, the mo de of the rayon each leg, and the sound speeds of thelongitudinal and transverse wave modesin the tw o m edia (solid steel and m oltensteel). No attempt was made to accountfor the e f fec t o f the temperature grad i ents near the solid/molten interface. Although these grad ients are known toaffect the sound speed and distort thesound field slightly (Ref.21), such calculat ions are complex and are not needed for

identif ication of the wave paths observedin the data. The output of the programconsisted of the locations of each entryand reflection point of the ray, the transitt ime for each leg, and the total transit t imefor the ray.

Two vers ions o f the program wereused. In the f irst, the user entered the ini

t ia l angle of the ray at the entry point andobse rved the path of the ray on a grap hics terminal. In the second version, theprog ram iterated thr oug h initia l angles un til the final leg exited at the location of thereceiving transducer.

In the second analysis method, anothercomputer program (Ref. 22) was used todetermine the approx imate beam spreadof the sound f ield. The program calculatedthe f ie ld due to a transducer co upled tothe solid through the plastic wedge with athin layer of fluid couplant. In this t ech nique, the transducer face was div ided upinto many small areas, each of which wasassumed to be a source of sphericalwaves. A ray was t raced f rom each sourcepo in t , th rough the wedge and in to thesolid to the requested f ie ld point. Thevecto r sum of the f ie lds of each of the rays

was then calculated to f ind the total incident f ie ld at that point f or a particular f re quency.

From the f ie ld of the transducer, therelative amplitudes of the signals weredetermined. Using this code the magnitude of the sound fie ld at locations alongthe bottom surface of the weld sample (xdirection) and along the sidewall preparat ion (z direction) was calculated for thetransducer placed on the steel sample atcoordinate s (x = — 25.4, z = 25.4). Thecalculated longitudinal beam spread wasdef ined to be the area betwe en the p o in tswhere the ampl i tude o f the f ie ld was

down by 20 dB relative to i ts maximumvalue along the beam centerl ine. Note

258-s | JULY 199 0



that the centerl ine for transverse waves

fro m this transducer was at an angle of 27deg in the steel part, while the angle for

longitudinal waves was 45 deg. For the

longitudinal waves these 20-dB points areat (x = —11, z = 0) in the x direction (on

the bot tom of sample , 11 mm f rom the

bottom corner), and at (x = 0, z = 10) inthe z direction (up the sidewall, 10 mm

from the corner). Between these posit ions

a long the bo t to m and up the s idewal l theultrasonic f ie ld was down by less than 20

dB compared to the f ie ld at the centerl ine

of the transducer. For the refracted trans

verse beam, the beam spread along thebo tto m (x d i rec t ion) was 10 to 23 mm (0 .4

to 0 .9 in. ) f ro m the bot tom corner wi th no

significant sound field on the vertical side-wal l . Because of reciprocity, the effect of

the beam spread on the amplitude of a

received signal from a given point wasexpected to be the same as that on a

transmitted signal to the same point.

Exper imenta l Resul ts beforeW e l d i n g

An A scan of the unwelded part isshown in Fig. 5. Three large signals can beseen and their source can be identif iedusing the two analysis techniques described above. The results of ray-traceanalysis of this unwelded preparation areshown in Fig. 6. The sound entrance andex i t po in ts were 25.4 mm from the edgeof the 25.4-mm-th ick carbon s tee l we ldsample. Only the base metal on the 90 degside is shown since the backing bar andthe base metal on the 60 deg side were

not acoustical ly coupled to the f irst piece.In Fig. 6A, a direct longitudinal wa ve at 45deg re f lec t ing f rom the corner is shown.This ray is actually made up of three legs:one f rom the t ransducer to the corner,one across the corner and one returningto the transducer. The second leg is tooshort to be seen in the figure. For simplicity, the ray is called the L-L ray, meaningthat it consists of t w o lon gitud inal legs.The coordinates in the x-z plane of theentry and exit points and the reflectionpoint are shown in the f igure. The raymoves to some extent in the directionperpendicular to the plane of f igure cor

responding to the d is tance between thecenter of the sending and receiv ing elements — Fig. 4. The ray-trace program calculated the total t ime for this ray to be15.6 yus (including the transit times in thetwo wedges o f the dua l-e lement t ransducer). This correlates well with the f irstecho obs erved in the A scan in Fig. 5. TheL-L signal to the corner is obviously alongthe centerl ine of the transducer and shouldbe a large reflection in the acquired Ascan.

Fig. 6B shows a more com pl ica ted pathconsisting of an initial transverse ray at 27deg , which mode converts a t the bot tomsurface to a longitudinal ray. This inter-

14.0 17.2 23.4 26.50.3

Time (fis)

Fig. 5-A scans of weld preparation acquiredwith dual element transducer.

sects the sidewall preparation at 8.4 mmand reflects to the top surface entering theLucite wedge at a f inal angle of 56 deg.This path is called T-L-L (meaning transverse-longitudinal-longitudinal). The reverse of this path is also possible. Thepredicted t ime for this ray is 19.8 us. Thispath correlates well with the second echoin Fig. 5. The T-L-L path sh ould h ave a largeamplitude since both the init ia l and f inal

rays are within the beam spread of thetransducer. The initial transverse leg (or final transverse leg for the reverse path) intersec ts the bot tom at -12.5 mm (—0.5in.) well within the beam spread of thetransducer. The final (or initial) longitudinalleg intersects the sidewall at 8.4 mm (0.33in .) up f rom the bot tom corner and the20-dB point is at 10 mm (0.04 in.) up fro mthe bo t tom co rne r .

Fig. 6C shows another path , L-L-T, generated by an initial longitudinal leg at 34deg that re flec ts o f f the bot to m and mod econverts to a transverse wave at 12.5 m m

up the sidewall. This wave enters the Lucite wedge at an angle of 63 deg. The total time fo r this ray is also 19.8 /xs, the sam eas the T-L-L ray. However, the L-L-T pathis not e xpec ted t o have a large amp litude.The initial longitudinal ray arrives at thebot tom at 8 .4 mm fro m the corner, w i th inthe 20-dB l imit of the beam, but the f inaltransverse ray originates from a posit ionon the sidewall well beyond the transverse wave beam spread of the transducer. A similar analysis shows that thereverse of this ray is also expe cted to besmall.

Ot her rays that are exp ecte d to be smallin amplitude include a T-T ray, w hich issimilar to t he L-L ray sh ow n in Fig. 6A, anda T-L-T ray, which consists of an initialtransverse ray to the bottom surface at14.8 mm that m ode co nverts to long i tud i nal, intersects the weld preparation at14.8 mm, and mode converts back totransverse (T-L-T). Both these rays havetotal travel t imes of 25 (is. The reflectionreceived from the T-T ray is not expectedto be of s ignif icant amplitude since thecorne r at 25.4 mm is outside the beamspread for both the init ia l and f inal transverse legs. Finally, in the case of the T-L-T

combination, the transverse ray to orf rom the s idewal l is beyond the beam

4 5 r

(-25.4, 25.4)

a)

\ L t-(0 , 0)

5 6 °

27°

b)

(-24.5, 24.5J

A

\ L L(0, 8.4)

(-12.5, 0)

(-25.4, 25.4)

(0, 12.5)

(-8.4, 0)

Fig. 6— Traces for rays before we/ding. A—L-Lray path; B— T-L-L ray path; C —L-L-T ray path

spread of the transducer for transversewaves and the signal is not expe cted to be

large. In Fig. 5, this ray is observed, but it

is smaller than th e other tw o discussedabove and is not observed in al l preweld

sidewall prepar ation A scans. The reasons

for this are not c lear and may be relatedto the roughness of the preparation.

Experimental Resul ts dur ing

W e l d i n g

Digit ized data and destructive evaluations were obtained for a total of 26 locations on root weld passes. Photomicrographs and A scans for f ive of these weldlocations are presented in Figs. 7 - 1 1 . Theweld shown in Fig. 7 is a f i l let w e l d . Theothers are single bevel V-groove welds —the bevel s ide did not fuse to the backingbar and came off during destructive evaluation in the examples shown in Figs. 9and 11 . These examples i l lustrate th e signals obtaine d fo r a go od we ld (Fig. 7), unacceptable w elds (Figs. 8-1 0) and on e

ve ry poo r we ld —Fig. 11. These signalswe re ana lyzed to determine i f incom ple te

W EL DI NG RESEARCH SUPPLEMENT | 259-S



Fig. 7 - Polished and etched weld samp le of adequate sidewall penetra

tion and fill height with A scan acquired during the root pass.

14.00 17.10 20.25 23.40 26.50

Time (^s)

Fig. 9 — Polished and etched w eld sample of incomplete penetration with

A scan acquired during the root pass .

1 m m

Fig. 8 — Polished and etched weld sample of adequate sidewall penetra

tion and low fill height with A scan acquired during the root pass.

**~~>j^~^f~^/v » < l * » ^ mMMMMt tmm^fMy

14.00 17.10 20.25 23.40 26.50

Time (/ts)

Fig. 10 — Polished and etched weld sam ple of near and far side incom

plete penetration with A scan acquired during the root pass .

260-slJULY 1990



Velocity

Density

Acoustic

impedance

m /s6.88 X 103

k g / m 3

27.9 X 10 6

N s / m 3

penetra t ion , poros i ty or undercut cond it ions cou ld be detec ted by observ ing thechanges in the signals due to the varioussound paths in the weld sample. The Ascan be low each photomicrograph is o fthe digitized acoustic signal as the moltenweld pool passed the transducer. The results obtained from the ray trace and f ie ldcode analyses were used to understandthe echoes in the digitized signals.

The mol ten po o l a f fec ted the three signals from the weld preparation (Fig. 5) invarious ways. Consider first the L-L signal.When adequate s idewal l penetra t ion occurred, the longitudinal wave refracted atthe molten interface with the majority ofthe sound entering the pool and with l i t t lesound energy re turned to the rece iv ingelement. The amplitude of the signal ref lec ted f rom the mol ten/so l id in ter facecan be calculated from the impedances of

the two media . The impedance in tu rndepends on the sound speed and density

of the m aterial. For solid and molten steel

at the melt ing point the sound speeds,densities and impedances are (Ref. 23):

So l id Mol ten

4.8 X 103 4.05 X 103

m /s

7.8 X 10 3

k g / m 3

37.4 X 106

N s / m 3

The amplitude reflection coeff ic ient, R,can be calculated from the impedances, Z,

by :R = (Z s - Z m ) / ( Z S + Zm) = 0.146

Thus, with adequate penetration, the L-L

echo should be only 14.6% of i ts preweldamplitude in Fig. 5. Higher amplitudes occur only if the penetration is not c om plete .If d iscontinuity condit ions l ike inadequatesidewall fusion were present, the effect ive impedance over the area of theincomplete fusion is zero, result ing in al lsound energy be ing re f lec ted f rom thediscontinuity at an angle favorable for recept ion by the rece iv ing e lement. Actu ally, because of the increase in attenuationof sound wi th temperature , the re f lec tedampl i tude f rom a s idewal l w i th goo d penetration is expected to be significantly lessthan that predicted from the reflection

coeff ic ient.

Now consider the effect on the T-L-Lray. The centerline of this ray intersectsthe s idewal l 8 .4 mm from the bot tomcorne r. Thus, this ray path can be a ffectedby sidewall fusion or undercutt ing higherthan 8 .4 mm.

Complete s idewal l fus ion wi th adequate f i l l is achieved in the w eld sho wn inthe photomicrograph in F ig . 7 . In the Ascan acqu ired f rom the mol ten p o o l , th esignal from the L-L ray to the bottom corner is not observed as the longitudinalwaves enter the pool because adequate

sidewall fusion is achieved and no soundis reflected back to the transducer. Th e f i ll

height of 8.5 mm is sufficient to alter thewaveform of the T-L-L ray at 20 ^s sincethis ray intersects the sidewall at 8.4 mm.

Complete penetra t ion o f the bot tomcorner is shown in Fig. 8, but w ith a l owlevel of only 4 mm. The L-L signal is smallbut present and the waveform is differentthan that observed before weld ing —Fig.5. This is due to the fact that the cornerrefle ctor is still presen t because th e lo w fill

level of this root weld does not affect al lof the rays that contribu te to the L-L soundpa th . Figure 12 shows that at this low filllevel about 60% of the L-L bea m is sending and receiv ing sound energy from thecorner. In addition, the T-L-L ray is not affected by this weld because the fill level iswell below the centerl ine of the returningL w a v e .

Incomple te penetra t ion o f the bot tomcorner is sho wn in tw o exam ples. The f irst(Fig. 9) example is of incomplete fusion ofthe backup bar but ful l fusion of the side-wal l . The L-L signal has a very low amplitude, indicating acceptable s idewall fusion, and the waveform of the T-L-L is notaltered since the fill level is below thecenter l ine o f the beam. The bot tom corner reflection appears to indicate a complete fusion condit ion, which is not thecase in this weld as the backup bar is notpenetrated. Therefore, this technique isnot appropr ia te for determin ing the quali ty of penetration for the backing bar butonly the quality of fusion at the sidewall.

Figure 10 shows a root pass with bothincomplete sidewall fusion and undercut.Incom plete fusion is evident in the A scansince the L-L signal from the corner, al

though decreased in amplitude, is sti l l asignificant reflector. In this case the fill

Fig. 11-Polishedand etched weldsample of porosity,undercut andincompletepenetration with A

scan acquired duringthe root pass.

height of 11 mm (0.43) is sufficient to affect the corner reflector rays that have aray path to 10 mm in the z direction.Therefore , the bot tom corner re f lec t ioncan only be due to the presence of corn ergeometry caused by incomplete sidewallfusion at the root of the w e l d . Th e T-L-Lwav e is a f fec ted in a minor way by the 11mm fi l l height and the undercutt ing at 12m m , indicating that this fill geometry is not

one that alters this path very much.A new signal, between the L-L and

T-L-L, appears in the A scan in Fig. 10. Using the ray tracing program, this is determined to be due to the path sh ow n in Fig.13 . The angle of the molten pool is suchthat the surface of the pool is a specularreflector that returns the ray sho wn in Fig.13 directly back to the transducer at thetime observed in the A scan.

In Fig. 11 , incomple te fus ion, poros i tyand severe undercutt ing caused by an

Fig. 12 —L-L ray paths in the presense of a mol

ten pool showing some rays transmitted intothe pool and others returning to the transducer.

WELDING RESEARCH SUPPLEMENT 1261-s



EMAT receiver

Fig. 13 —Ray path that accounts for the presence of the extra echo observed in the weldshown in Fig. 10 .

unacceptab le weld ing process are show n.The bottom corner reflector (L-L) remainswhen the molten pool is present, indicating the inco mple te sidewa ll fusion, although the f i l l height is adequate. Thismeans the corner reflec tion is due toincomplete fusion and not low fi l l level asin Fig. 8. The T-L-L signal is red uce d in am p l i tude and the wav efo rm a l tered dramatical ly because of severe undercutt ing andgross porosity at the sidewall.

Discussion

Ultrasonic Discontinuity Detection

The analysis of data acquired duringroo t passes fro m 26 stationary transducerlocations shows that certain discontinuitycondit ions can be detec ted using refractedlongitudinal sound waves to interrogatethe mol ten p o o l . Severe undercut above8 mm can be detected by the T-L-L pa th ,because of the distort ing effect the undercut has on th e signal. Furthermo re, thisdistort ion can be observed as a decreasein the amplitude of the signal and a changein the waveform. Incomple te fus ion canbe detected if i t occurs at the sidewall.However, incomple te fus ion o f the backing bar is not detected using the currentexper imenta l se tup.

Expert System Potential

An expert system could be used to determine discontinuity condit ions during

Steel backer bar Weld pool

T i m e (jiS)

Fig. 15 — Noncontacting ultrasonic data showing the effect of low fill level.

Fig. 14 — Schematic of noncontacting sensing offill level of a fillet weld.

GMAW. By f irst recording the init ia l weldpreparation signals using a dual-elementultrasonic transducer and then monitoringthe signals observed during welding forthe presence of an L-L signal with a largeampl i tude and waveform changes in theT-L-L signal, the system could make deci sions about the presence of discontinuitycondit ions and the quality of the weld geometry . The expert sys tem would thenprovide an input s ignal to a c losed-loopprocess control system to minimize theamount o f unacceptab le root we ld f romthe automated welding process. The system could work l ike one already developed for use in ultrasonic evaluation ofsolid if ied, partia l ly completed welds (Ref.24). In that system , a s imple "If, Then , Else"type of logic is used to evaluate the amplitude and shape of ultrasonic signals anddecide if the A scan contains echoes fro mincomplete sidewall fusion, porosity org o o d w e l d . The system is an expertsystem in the sense that the decis ion pr o

cess is based on the same cues in the Ascans that the developers of the systemwere able to use to discriminate amongthe d i f fe rent f laws and good w e l d .

Noncontacting Ultrasonic System

To make the system truly practical, difficulties in using the piezoelectric transducers must be overcome. The majorpro ble m is the maintenance of c ouplingbet we en the transducer and the part. Thisis normally done with a f i lm of gel betwe en the Lucite we dge and the part. Thisgel must be fairly thin and unifo rm w ith no

gaps or large bubbles. In the preliminaryexperim ents, a special f ix ture w as buil t toprov ide a cont inuous f low o f ge l betwe enthe transducer and the part for the short(1-m) distances of travel. Stroud (Ref. 25)uses more complex devices which maintain coupling by pumping gel into a tube,which then takes the place of the plasticw e d g e .

Anothe r approa ch is to use nonc ontacting transdu cers (Ref. 26). In preliminaryexperiments on fillet welds, a laser spotwas focused on the mol ten weld v ia mirrors and lens to generate ultrasound di

rectly on the surface of the weld pool(note the laser energy was much less than

that used in laser beam welding). The ult rasound was t ransmit ted through thepoo l to an electrom agnetic acoustic transducer (EMAT) on the part as shown schematically in Fig. 14. W ith this e xperime ntalsetup, the sound path went th rough themol ten/so l id in ter face about 3 .9 mm(0.15) up the sidewall and was thereforeaffected by low fi l l level. When the f i l letwe ld geometry was acceptab le , soundgenerated by the laser was received bythe EMAT. Figure 15 shows a sequence ofA scans obtained at 8-mm intervals withthis noncontacting system, start ing in anarea of adequate fill level, passing throughan area of low fi l l level, and then back toadequate fill level. The signal received bythe EMAT at 12 ps th rough the sound pathshown in Fig. 14 was initially present, disappea red in the region of lo w fi l l level, andthen re turne d as the accep table f i ll levelwas res tored. Star t ing wi th the fourth Ascan, a severe drop in fill level resulted inthe disappearance of this signal. This occurred because the change in the f i l let

weld geometry alters the sound path tothe EMAT to such an extent that no so undwas received. As an acceptable f i l l levelwas again achieved , the EMAT rece ived asignal from the weld as seen in the lastthree A scans of Fig. 15.

For this noncontacting system to befieldable, several improvements must bemade. The EMAT is much less sensitivethan piezoelectric transducers and thesignals in Fig. 15 show a poorer signal-to-noise ratio than those presented abovefor the piezoelectric transducer (Fig. 5,Figs. 7-11 ). An o rder of magn itude im

pro vem ent in sensitiv ity is require d fo rpractical use. The laser beam must be del ivered to the weld via a f iber-optic cableto reduce personnel hazards in the workarea and to increase the mobil i ty of thesource of excitation on the weld for interrogation of the total weld interface. Usingonly mirrors and lenses, sound could begenerated only on the upper part of thew e l d poo l .

Conclus ions

Discontinuit ies may be detected duringGMAW us ing u l t rason ic methods. The

types of discontinuit ies detected in thisw or k are incomple te sidewall fusion, grossporos i ty and undercut . The in format ionabout the posit ion and types of discontinuit ies being formed could be derivedfrom the ultrasonic signals by an expertsystem and the results sent to the con trol le r o f an automated GMAW system. Thecontrol ler could change the welding parameters to prevent the cont inuat ion o fthe d iscont inu i ty cond i t ion , thus improving the quality of the welding process. Inaddit ion, the condit ions leading to defectsmay also be detected (Ref. 9), and thepotential exists for preventing the discontinuity condit ions altogether. Noncontact-

262-S | JULY 1 9 9 0



i n g s y s t e m s m a y i m p r o v e e n o u g h i n t h e

f u t u r e t o p r o v i d e a v e r y p r a c t i c a l m e a n s

o f i m p l e m e n t i n g t h e t e c h n i q u e o n t h e

s h o p f l o o r .

Ackno wledgmenls

T h i s w o r k w a s s u p p o r t e d b y t h e U . S .

D e p a r t m e n t o f E n e rg y , O f f i c e o f E n e r g y

R e s e a r c h , O f f i c e o f B a s ic E n e r g y S c i e n c e s

u n d e r D O E C o n t r a c t N o . D E - A C 0 7 -

7 6 I D O 1 5 7 0 . A p p r e c i a t i o n is e x p r e s s e d t oU . S . W a l l a c e w h o o v e r c a m e h i s p r i d e t o

m a k e w e l d s w i t h f l a w s d e l i b e r a t e l y a n d t o

C . L. F l e tc h e r w h o p r e p a r e d t h e d e s t r u c

t i v e e v a l u a t i o n s a m p l e s .

References

1. Lukens, W . E., and M orri s, R. A. 1982. In

frared temperature sensing of cool ing rates forarc weld ing contro l . Welding Jour. 61(1 ) :27 -33 .

2. Covardhan, S. M., and Chin, B. A. 1989.Mon i to r i ng GTA we ld pudd le geometry us ing

measured surface temperature gradients. Re

cent trends in weld ing science and technologyT W R ' 8 9 . Proceedings of the Second Interna

tional Conference on Trends in Welding Re

search, Gat l i nbu rg , Tenn . pp . 383 -386 .

3. Nagara jan, S., G ro om , K. N., and Chin, B.

A. 1989. Infrared sensors for seam tracking in

GTAW. Recent Trends in Weld ing Science andTechno logy TWR'89 , Proceedings of the Sec

ond International Conference on Trends in

Welding Research, Gat l i nbu rg , Tenn . , pp . 9 5 1 -

955.

4. Lui, S., and Siew ert, T. A. 19 89. M eta l

transfer in gas meta l arc weld ing: droplet ra te.

Welding Journal 68(2):S2-s to 58-s.

5. Johnson, J. A.,Ca rlson, N. M ., and Sm artt,

H. B. 1989. Dete ct ion of Meta l -Transfer M od ein GMAW, Recent Trends in Weld ing Science

and Techno logy TWR'89 , Proceedings of the

Second International Conference on Trends in

Welding Research, Gat l i nbu rg , Tenn . , pp .377 -381.

6. Lesnewich, A. 1958. Contro l o f mel t ing

rate and me tal transfer in gas shielded metal arc

we ld ing , pa r t II —control of meta l t ransfer .

Welding Journal 37(4):418-s to 425-s.

7. Johnson, | . A., e t a l . 1986. Automated

weld ing process sensing and contro l . Proceed

ings of the Second International Symp osium on

the Nondestructive Characterization of Materi

als, New York: Plenum Publ ish ing Corp., pp.4 0 9 - 4 1 7 .

8. Richardson, R. W ., Gu to w, D. A., Ander

son, R. A., and Farson, D. A. 1984. Coaxial arc

weld pool moni tor ing for process moni tor ingand con t ro l . Welding Journal 63(3 ) :43 -50 .

9. Car lson, N. M., and Johnson, | . A. 1988.Ul trasonic sensing of we ld po ol pe netrat ion.

Welding Journal 67C\1):239-s to 246-s.

10. Lott , L . A. 1984. Ultrasonic detect ion of

mol ten/so l id in ter faces of weld pools. Materi

als Evaluation 42, p. 337.

11 . Hardt, D. E., and Katz, J. M. 1984. Ul trasonic measurement of weld penetrat ion.

Welding Journal 63(9):273-s to 281-s.

12. Stroud, R. R., and Fenn, R. 1985. The

measuremen t and con t ro l o f pene t ra t i on du r

ing submerged arc weld ing. We/ding Journal

6 4 :1 8 - 2 2 .

13. D e a m , R. T . 1989 . W e ldp oo l f requency :

a new w ay to de f i ne a we ld p rocedu re , Recen tTrends in Weld ing Science and TechnologyT W R ' 8 9 , Proceedings of the Second Interna

tional Conference on Trends in Welding Re

search, Gat l i nbu rg , Tenn . , pp . 9 6 7 - 9 7 1 .

14. Rohkl in , S„ Ch o, K., and Gun , A. C. 1989.

Closed-lcop process contro l o f weld penetra

t ion using real - t ime radiography. Materials

Evaluation 47 , pp . 363 -369 .

15. Sm artt, H. B., Einerson, C. J., Watk ins , A .

D., and Morr is , R. A. 1986. Gas meta l arc process sensing and contro l . Advances in We ld ing

Science and Technology TWR'86, Proceedings

of the First International Conference on Trends

in Welding Research, Gatl inburg, Tenn., pp.

4 6 1 - 4 6 5 .

16. Doum anid is, C. C, and Hardt, D. E.1988. A mode l for in-process contr o l o f thermal

propert ies dur ing weld ing. Trans. ASME, J. Dy

namic Systems, Measurem ents and Control.

17. Kim, Y. S., and Eagar, T. W . 1989. Tem

perature d istr ibut ion and energy balance in thee l e c t r o d e d u r i n g GM A W . Recent Trends in

Welding Science and Technology TWR '89,

Proceedings of the Second Internat ional C o n

ference on Trends in Weld ing Research, Gat l in

bu rg , Tenn . , pp . 13 -18 .

18. Johnson, | . A., Carlson, N. M., andSmartt, H. B. 1986. Gas Metal Arc D efect For

mation Mechan isms: Detection and Control,

EGG-SD-7405.

19. Johnson, | , A., et al. 1988. Ultrasonic andvideo computer ized data acquis i t ion for auto

ma ted we ld ing . Proceedings of the 2nd Inter

national Conference on Comp uter Technology

in Welding, The Weld ing Inst i tu te, Cambridge,

England, pp. 18-1 to 18-9.

20. Johnson, J. A., e t a l . 1987. A C AM AC

based u l trasonic data acquis i t ion workstat ion.

Materials Evaluation 45 , pp . 934 -938 .

21 . Johnson, | . A., Carlson, N. M., Lott, L. A.1987. Ultrasonic wave propagat ion in temper

ature gradients. Journal of NDE 6, No. 3, pp.

147-157 .

22. Johnson, | . A., and Tow, D. M. 1986.Numerical calculations of ultrasonic field/crack

interactions. Review of Progress in QuantitativeNDE, 5A, D. O. Thompson and D. E. Chimenti ,eds., New York: Plenum Publ ish ing Corp., pp.

8 3 - 9 1 .

23. Kurz, W „ and Lux, B. 1969. Berg-und

Huttenmannische Monatshefte 114, p. 12 3-130.

24. Johnson , J.A., Carlson, N.M . 1986. W el denergy reduct ion by using concurre nt n onde

struct ive evaluat ion. NDT International 19 ,

p p .1 9 0 - 1 9 6 .

25. Stroud, R. Seam tracking with ultra

sound, p rep r in t , p r i va te commun ica t i on .

26. Johnson, J. A., and Ca rlson, N. M. 1988.

Nonc ontact Ul trasonic Sensing of W eld Poolsfor Automated Weld ing. Thi rd In ternat ional

Symposium on Nondestruct ive Character izat ion of Mater ia ls, Saarbruecken, FRG.

WRC Bulletin 332April 1988

This Bulletin contains two reports that characterize the mechanical properties of two differentstructural shapes of constructional steels used in the pressure vessel industry.

( 1 ) Characteristics of Heavyweight Wide-Flange Structural ShapesBy J. M. Barsom and B. G. Reisdorf

T h i s r e p o r t p r e s e n t s i n f o r m a t i o n c o n c e r n i n g t h e c h e m i c a l , m i c r o s t r u c t u r a l a n d m e c h a n i c a l ( i n c l u d i n g

f r a c t u r e t o u g h n e s s ) p r o p e r t i e s f o r h e a v y w e i g h t w i d e - f l a n g e s t r u c t u r a l s h a p e s of A 3 6 , A 5 7 2 G r a d e 5 0 a n d

A 5 8 8 G r a d e A s t e e l s .

( 2 ) Data Survey on Mechanical Property Characterization of A588 Steel Plates and Weldments

By A. W. Pense

T h i s s u r v e y r e p o r t s u m m a r i z e s , f o r t h e m o s t p a r t , u n p u b l i s h e d d a t a o n t h e s t r e n g t h t o u g h n e s s a n d

w e l d a b i l i t y o f A 5 8 8 G r a d e A a n d G r a d e B s t e e l s as i n f l u e n c e d b y h e a t t r e a t m e n t a n d p r o c e s s i n g .

P u b l i c a t i o n of t h i s B u l l e t in w a s s p o n s o r e d b y t h e S u b c o m m i t t e e o n T h e r m a l a n d M e c h a n i c a l E f f e c t s

o n M a t e r i a l s o f t h e P r e s s u r e V e s s e l R e s e a r c h C o m m i t t e e o f t h e W e l d i n g R e s e a r c h C o u n c i l . T h e p r i c e o f

W R C B u l l e t i n 3 3 2 is $ 2 0 . 0 0 p e r c o p y , p l u s $ 5 . 0 0 f o r p o s t a g e a n d h a n d l i n g . O r d e r s s h o u l d b e s e n t w i t h

p a y m e n t t o t h e W e l d i n g R e s e a r c h C o u n c i l , S u i t e 1 3 0 1 , 3 4 5 E. 4 7 t h S t . , N e w Y o r k , N Y 1 0 0 1 7 .

detection of discontinuities [gmaw]

Documents