characterisation of ice-structure and fluid-structure interactions on
TRANSCRIPT
Characterisation of Ice-Structure
and Fluid-Structure Interactions
on Polar Vessels using Operational
Modal Analysis
Progress Report for COMNAP
Author
Keith Soal
2015
i
TABLE OF CONTENTS
Page
1. Introduction .................................................................................................. 2
2. Objectives ...................................................................................................... 3
3. Motivations .................................................................................................... 4
4. Progress to date ............................................................................................ 5 4.1. Full-scale Measurements .................................................................... 5 4.2. Validation of a novel OMA implementation ...................................... 5
4.3. Characterisation of Ship Dynamic Response using OMA ................. 5
5. Conclusions ................................................................................................... 8 Appendices ............................................................................................................... 9
2
1. INTRODUCTION
Polar research vessels operating in ice and open water are relied upon by
research institutes and their scientists to re-supply bases as well as to serve as
floating laboratories. Increasing interest in the Arctic's Northern Sea Route as
well as in Antarctica have resulted in countries such as Germany, China, the USA,
Australia and Russia investigating options for new polar vessels.
These vessels operate in unique environments, and are exposed to complex
dynamic loading patterns and vibration responses. The resulting ice-structure
and fluid-structure coupled interactions are still not well understood. The
dynamic response of vessels is strongly related to their performance in terms of
ice resistance, power consumption and ice breaking capacity. Dynamic responses
are therefore of extreme importance during the design and optimisation phases
of new vessels. Several studies have been conducted into the local ice impact
forces on ship hulls, as well as the effect of wave induced vibrations on fatigue
damage, however very few studies have collaborated across scientific disciplines
to study coupled phenomena in more detail.
In 2012 STX Finland recognized the sparse high resolution full-scale data
spanning across disciplines, and formed and international consortium of research
institutions and universities. The aim of the consortium is to create a scientific
basis for the design of ice going ships in terms of ship hull, propulsion, power
requirements and comfort for passengers and crew on board. The consortium
members included Aker Arctic, STX Finland, DNV, Rolls-Royce, Wärtsilä, The
Department of Environmental Affairs (South Africa), Smit Vessel Management
Services, Aalto University, the University of Oulu and the University of
Stellenbosch.
3
2. OBJECTIVES
The aim of this project is the characterization of ship dynamic responses due to
complex ice-structure and fluid-structure interactions, using operational modal
analysis (OMA). Investigations will be conducted into different OMA algorithms
to determine which are best suited to accurate modal model estimation while
dealing with challenges such as harmonic contamination and non-stationary
modal behaviour. Due to the complexity of characterising local ice-structure
interactions as a result of non-linear ice mechanical properties, it is proposed to
use OMA to improve the estimation of ice impact forces from inertial
measurement unit data in two ways. Firstly the Eigenvalues and vectors obtained
in OMA can be used to reconstruct the mass (M), damping (C) and stiffness (K)
matrices. This will allow model updating based on the actual loading and
boundary conditions, which is currently the major limitation of the method.
Secondly OMA will provide an estimation of elastic energy dissipation which will
further improve the technique by not limiting it to the rigid body assumption.
The modal parameters will then be used together with other full-scale predictor
variables such as ship performance parameters, wave heights and directions and
ice thickness, concentration and floe size to develop a predictive model for ship
dynamic responses using multivariate statistical techniques. The determination
of ship hull static stiffness from full-scale OMA will also be investigated and
validated on laboratory structures.
4
3. MOTIVATIONS
In order to design optimal and efficient vessels to deal with complex ice-
structure and fluid-structure interactions, ship designers need an array of tools
to predict the effects of variables such as bow geometry, engine power, thrust,
speed etc. on ship dynamic response which is closely related to ship
performance. These tools are developed from measurements and observations
at model-scale and full-scale, from which numerical models can be formulated.
Ice-structure and fluid-structure interaction phenomena are still not understood
to the level of accurate numerical prediction. Characterisation of these
interactions using OMA and multivariate statistical techniques will provide
further insight into the coupled phenomena in order to update and validate
numerical models.
Currently on most modern ice going vessels, the captain and his/her navigating
officers operate the vessel based on the dynamic response feedback which they
receive through the structure in the bridge. Low response feedback could
therefore result in the vessel being operated at speeds and headings which could
cause impact damage or significantly increase fatigue damage due to thick ice
features or large waves. The determination of local and global ship loading
during operation using whole ship motions as well as OMA, would therefore
provide the navigating officers with more information which would allow them
to make better informed decisions regarding both safe and efficient vessel
operation.
5
4. PROGRESS TO DATE
4.1. Full-scale Measurements
Full-scale measurements were conducted on the S.A. Agulhas II during the
2014/2015 Antarctic voyage. The measurement setup was significantly expanded
to include forty two synchronous measurement channels. Measurements were
also conducted during a voyage to Marion Island in the Southern Ocean. This was
purely an open water voyage, however large levels of slamming had been
experienced on previous voyages and provides and interesting dynamic case to
study. Propeller speed run-up and run-down tests were also conducted and will
be analysed to determine whether this is an effective modal extraction
technique. It is essential for the success of this work to develop high resolution
data sets recorded over long durations and multiple seasons with different
operating and environmental conditions. A research proposal for full-scale
measurements on the German research vessel the R/V Polarstern was accepted
and will allow measurements during a voyage to the Arctic. This will add a new
dimension to the existing data set and allow comparisons of Arctic and Antarctic
sea ice as well as comparisons of the two different vessel designs.
Instrumentation of the vessel is currently being planned for October 2015.
4.2. Validation of a novel OMA implementation
Laboratory testing is currently being conducted at the Technische Hochschule
Ingolstadt (THI) in Germany to validate a novel implementation of OMA to
analyse structures. The idea is currently being patented by Prof Joerg Bienert,
and the rights are therefore reserved. Upon completion of the patent the
method will be published with full details.
4.3. Characterisation of Ship Dynamic Response using OMA
Modal analysis is a multi-faceted field involving both rigorous mathematical
proofs and intricate test procedures. In the authors opinion it is equally
important to study the mathematical theories as it is to conduct tests, observe
the physical phenomenon and analyse the data. For this reason the various
dynamic and acoustic tests which have been conducted at the THI, although
often seemingly unrelated to the current project, have provided invaluable
lessons and insight. The following topics are currently under investigation:
6
MEMS sensor testing -low cost MEMS accelerometers have been tested
against high precision PCB accelerometers by running sweep tests on a
laboratory shaker. Results have shown that MEMS sensors are able to
match PCB sensors to within very tight tolerances, and have a low noise
floor. It has however been found that the intended MEMS sensor
housings have a large effect on sensor performance and new housing are
currently being laser printed for further testing.
Acoustic modal analysis - acoustic modal analysis has been conducted in a
glass fish tank using a self-made microphone array of low cost MEMS
microphones. The fish tank is excited by an electromagnetic shaker and
studies are currently being conducted into the coupling between
structural and acoustic modes, as well as the effect of introducing
different damping materials.
Reverberation chamber testing - reverberation time and sound power
tests have been conducted to characterise the THI's new reverberation
chamber, and as a precursor to further acoustic testing.
Sound power measurements in vehicles - sound power has been
measured in four different new three cylinder cars during different
driving conditions. Acoustic testing has also been conducted inside the
helmet of a motorcycle rider using low cost MEMS sensors.
Modal analysis of a dynamic test bench - modal analysis of a test bench
was conducted to provide further insight into the modal properties
obtained during contract work for Airbus.
Analysis of foam boundary conditions - the effects of so called free-free
boundary conditions when using foam support were determined
analytically and compared to measurement results.
Implementing EMA and OMA in Matlab - an EMA solver capable of
implementing the Least Squares Complex Exponential (LSCE) and
Polyreference Time Domain (PTD) methods has been coded in Matlab
using the ABRAVIBE tool box. Implementations of the OMA methods SSI
and FDD are currently being conducted. This has provided insight into the
mathematics behind the respective techniques, from which it is
hypothesised will aid in the identification of potential strengths which can
be further exploited and weaknesses which should be avoided.
7
Conference papers - conference papers have been published and
presented at two international conference, namely International
Operational Modal Analysis Conference (IOMAC) in Spain, and Port and
Ocean Engineering under Arctic Conditions (POAC) in Norway. The
current research has also been presented at the European Modal Analysis
Users Group Conference (EMAUG) in Berlin. These conferences have
been instrumental in the development of new ideas and have provided
great insight and perspective from discussions with world leading
professors and researchers.
Research collaboration - a research exchange to Aalto University in
Finland is planned from 16 August 2015 and will provide a platform for
collaborative investigations into important phenomena.
8
5. CONCLUSION
I am extremely grateful to COMNAP for this fellowship which has enabled me to
pursue my passion for research into structural dynamics on vessels operating
under polar conditions. The funding has enabled me to conduct my studies under
the expert guidance of Prof Joerg Bienert in Germany which has been a privilege.
The funding has also enabled me to travel and present my research at
conferences in Spain, Norway and Berlin, where I have met and been inspired by
many Professor’s and researchers whose research and text books I have been
studying during my Master’s degree in South Africa. Being exposed to this world
class calibre of researches has challenged my thinking and cultivated new ideas
such as the use of OMA to estimate global ice loads from whole ship motion. It
has also resulted in new and exciting collaborative research. The conference
papers from IOMAC and POAC are included in the appendices, with grateful
acknowledgement to COMANP. The last slide from the presentations at IOMAC,
POAC and EMAUG has also been included which also acknowledges the support
which has made this work possible.
9
APPENDICES
OPERATIONAL MODAL ANALYSIS ON THE POLAR SUP-PLY AND RESEARCH VESSEL THE S.A. AGULHAS II
Keith Soal 1, Jorg Bienert 2, Anriette Bekker 3
1 Mr, Stellenbosch University, [email protected] Prof, Technische Hochschule Ingolstadt, [email protected] Dr, Stellenbosch University, [email protected].
ABSTRACTOperational Modal Analysis (OMA) was conducted on the S.A. Agulhas II while moored in Cape Townharbour to investigate the global structural dynamic characteristics of the vessel. Measurements wereconducted while the vessel was exposed to small wave (ripple) and wind excitation at the quaysidewithout harmonic excitation from the main engines or harbour generator. Twenty three vibration mea-surement channels were recorded in total with eighteen sensors on the hull structure and five on the su-perstructure. The sensor locations and directions were selected to investigate the normal and transversebending modes as well as torsional modes of the structure. The LMS Operational PolyMAX frequencydomain and ARTeMIS CCSSI time domain OMA techniques are used to estimate the modal parameters.Both techniques identified three stable modes which include the two-node (first), three-node (second)and four-node (third) normal bending modes. In terms of frequency the Operational PolyMAX and CC-SSI agree to within 1,2 %. However the damping estimates show less agreement especially for mode 3which differs by 59 %. The Modal Assurance Criterion confirms three unique modes, and the complexityplots reveal real valued mode shapes which confirm the proportional damping model approximation. Theresults of the OMA were then compared to those of the Finite Element (FE) model developed by STXFinland. The natural frequencies predicted by the FE model are found to be bigger than those measuredusing OMA. The FE calculations are based on a vessel draught of 7,7 m which is deeper than the 6,8 mdraught during OMA testing. The effect of the vessel draft on the modal parameters will form part offuture research. This work is a precursor to investigations into the effect of various ice and open waterboundary conditions as well as ship loading and operating conditions on the dynamic characteristics ofthe vessel structure.
Keywords: Full Scale Measurements, Operational Modal Analysis, Polar Supply and Research Vessels
1. INTRODUCTION
Vessels operating in Antarctica and the Southern Ocean are exposed extreme and unpredictable condi-tions, and encounter a number of excitation mechanisms. These excitation mechanisms include waves,ice and wind as well as the engines, propellers and machinery on board. This results in a variety of forcesbeing applied to the structure which can cause structural fatigue, excessive vibration and slamming.
Polar Supply and Research Vessels (PSRVs) play a key role in scientific and logistical support in Antarc-tica and the Southern Ocean. The PSRV S.A Agulhas II is the work horse of the South African NationalAntarctic Program (SANAP). The S.A Agulhas II entered service in 2012, and was designed with anoperational lifetime of 30 years. The vessel spends around 8 months a year at sea in some of the harshestoperating conditions on the planet. Investigations into the structural dynamic response of the vessel aretherefore important to assess its dynamic performance, its ability to operate safely for 30 years as wellas to compare the measured dynamic response to the predicted response used during the design phase.
In order to investigate the vessel’s structural dynamic response to these excitation mechanisms, it isimportant to first determine the structural dynamic characteristics of the vessel. Operational ModalAnalysis (OMA) is used to investigate the global structural dynamic characteristics of the ship. The aimof OMA is to obtain the structure’s modal parameters, namely the natural frequencies, damping ratios andunscaled mode shapes from response only measurements. These modal parameters will provide insightinto the operational dynamic response of the structure and can then be used to investigate phenomenasuch as structural fatigue, excessive vibration and slamming.
2. THE S.A. AGULHAS II
Full scale measurements were conducted on-board the PSRV S.A. Agulhas II, see Figure 1, which wasbuilt by STX Finland at the Rauma Shipyard. The S.A. Agulhas II is designed to carry cargo, passengers,bunker oil, helicopter fuel and is also equipped with laboratories, a moon pool and drop keel to conductscientific research in the Southern Ocean. The main specifications of the ship are presented in Table 1.
Figure 1: The S.A. Agulhas II.
Table 1: The main specifications of the S.A. Agulhas II
Length, bpp 121,8 mBeam 21,7 mDraught, design 7,65 mDeadweight at design displacement 5000 tInstalled power 4 × Wartsila 6L32 3000 kWPropulsion Diesel-electric 2 × 4500 kWSpeed, service 14 kn
3. FULL SCALE MEASUREMENTS
3.1. Measurement Equipment
The data acquisition system (DAQ) and measurement equipment used are presented in Table 2. TheLMS SCADAS were configured in a master-slave setup which allowed simultaneous measurements con-trolled from one DAQ. The LMS SCADAS are equipped with a hardware low-pass anti-aliasing filter.Accelerometers were calibrated according to the South African Bureau of Standards (SABS) by the Na-tional Metrology Institute of South Africa (NMISA). Accelerometers were mounted to rigid structuralmembers in order to measure the global ship vibration response.
Table 2: Measurement equipment
Equipment
1 x 16 channel, LMS SCADAS1 x 12 channel, LMS SCADAS1 x 8 channel, LMS SCADAS9 x DC PCB accelerometers, 20,4 mV/(m/s2)9 x ICP PCB accelerometers, 10,2 mV/(m/s2)3 x Seismic PCB accelerometers, 1019,4 mV/(m/s2)1 x Triaxial PCB accelerometer, 10,2 mV/(m/s2)LMS Test.Lab 11A Turbine Testing software
3.2. Measurement Setup
Twenty three vibration measurement channels were recorded in total with eighteen on the hull structureand five on the superstructure as shown in Figure 2.
Figure 2: Measurement model indicating sensor location and measurement direction.
Vertical vibration (+Z) was measured on the port and starboard side of the ship hull at as close to equalincrements as was physically possible. This was done in order to investigate the horizontal (x-y) planefor vertical bending and torsional modes. Lateral vibration was measured at three points along the hullto investigate the longitudinal vertical (x-z) plane for transverse bending modes. Longitudinal vibration(+X) in the hull is not considered in the present work.
Vibration measurements in the superstructure include longitudinal (+X) measurements to investigatefore-aft bending, lateral (+Y) measurements to investigate transverse bending and vertical (+Z) measure-ments to investigate torsion. A measurement duration of an hour was selected based on the results fromRosenow [1]. This study concluded that measurement durations of one hour or more were necessary toobtain stochastic excitation. The global modes of the vessel are expected to lie below 10 Hz, but in orderto improve the resolution of the data, a sample frequency of 128 Hz was chosen.
3.3. Measurement Conditions
OMA was conducted on the S.A. Agulhas II while moored at East Pier in Cape Town harbour. Mooringlines were used to secure the vessel against large rubber tyres hanging from the quayside. The 1 hour runselected for the present analysis was between 00h00 and 01h00 on 28 February 2014. The offloading ofheavy cargo had been completed, and the ship had not been refuelled with polar diesel. The draft was 6,7m fore, 6,8 m midship and 6,9 m aft. The ship was on shore power and the wind had picked up duringthe night to 43 km/h. This measurement run was chosen due the lack of harmonic contamination fromthe engines and harbour generator as well as good wind and wave excitation.
4. RESULTS
4.1. Power Spectral Density
The power spectral density (PSD) in Figure 3 indicates the distribution of the signal power in the fre-quency domain. The PSDs of the acceleration measurements were calculated using the peak amplitudemode, Hanning window, 50 % overlap and 2048 NFFT points resulting in a frequency resolution of0,0625 Hz. The following observations are made:
1. There are distinct low frequency peaks at 1,9 Hz, 3,37 Hz and 4,66 Hz.
2. The amplitude of the frequency content under 5 Hz is the largest in the bow, closely followed bythat in the stern.
3. The amplitude in the frequency range between 5 Hz and 30 Hz is the largest in the stern.
4. The frequency content in the vertical (+Z) direction in the bridge is dominant in the frequencyrange from 30 Hz to 36 Hz.
0 64Hz
1.00e-9
0.01
Log
(m/s
2)2
/Hz
F PSD Point18:+Z
F PSD Point23:-Z
F PSD Point22:+Y
F PSD Point33:+Z
F PSD Point11:+Z
F PSD Point7:-Z
F PSD Point16:+X
1.9 3.37 4.66
Figure 3: Power Spectral Density (PSD) of the time signals.
4.2. Stabilization Diagrams
The structural dynamic characteristics of the vessel are identified using LMS Operational PolyMAX andARTeMIS CCSSI algorithms. An Eigenvalue analysis is used to plot the stabilization diagram from
which the modal parameters can be estimated. Figure 4 shows the stabilization diagram produced by theLMS Operational PolyMAX frequency domain algorithm. Stable poles with high confidence are seen toalign at 1,94 Hz and 3,37 Hz indicating stable physical modes. A third less stable pole at 4,72 Hz is alsoselected. Poles at higher frequencies were investigated but did not provide clear results.
0.00 16.0Linear
Hz
290e-12
781e-9
Am
plit
ude
( m/s
2)2
oo s s s o f f o
ds s s s f f f f f
od s s d f o f f o f
oo s s o f f o f f f o
oo s s f d d f f d f f f
do s s f d d o o f f f f f
oo s s f f f o f f d f f f
of s o s o d f o f f f f f ff
oo s o s f d d f f f f ff f f
oo s f s d d v d f d s f f f d o
fo os fs f d f f f d f o f f d o
of o s os d d f f f f s f d f f o o
ooo o s vf o v f o o f d f f o f f f f
of o f s sf f d o d f o f s s d f f s v f
oo o f s vo d o f f o ff f d f ff f f f
do o o s so d f f of o f f f f f f o f f f
od f o s s o s o d f o f o f d f ff f f o f
oo o so s f s f d o f ff d d o f f o d f o o o
oo o s o s f s f d o f ff d df d f f f f f f o
ss f s o o s f d f d o o ff of d f f f f f f f o
do o s o o s f s d d f o o d o d s d d f f d f oo o
so o s f o s s s d s f f o f d f d s s s d s f f f f f
ss o s f o s f s d f o d o o f d f s s s d d d f f f f f
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
71
8e-
9
290e-
12
(𝑚
/𝑠2)2
5 0 10 16 Frequency (Hz)
32
31
30
29
28
27
26
25
24
23
22
21
20
19
18
17
16
15
14
13
12
11
10
Figure 4: Operational PolyMAX stabilization diagram. (s) Stable pole with high confidence, (v) Some confidencein the Eigenvector, (d) Some confidence in damping, (f) Some confidence in the Eigenvalue, (o) Unstable pole.
The stabilization diagram from the ARTeMIS CCSSI time domain algorithm is presented in Figure 5.Three stable poles are identified at 1,93 Hz, 3,36 Hz and 4,67 Hz. The CCSSI stabilization diagramis clearer than that produced by Operational PolyMAX, as the third mode is also identified with highconfidence in CCSSI.
Frequency (Hz) 0 16
80
70
60
50
40
30
20
10
Figure 5: CCSSI stabilization diagram of estimated state space models. • Stable mode, NUnstable mode, �Noisemode.
The natural frequencies and damping ratios of the three selected modes using Operational PolyMAX andCCSSI as well as the percentage difference between the two estimates are presented in Table 3. CCSSIand Operational PolyMAX agree to within 1,2 % on the natural frequencies. The damping estimateshowever show less agreement, especially for mode 3 which differ by 59 %. The damping percentage
for all three selected modes is small due to the high stiffness of the structure which is reinforced for icenavigation.
Table 3: A Comparison of natural frequencies and damping ratio estimates using Operational PolyMAX andARTeMIS CCSSI as well as their percentage difference.
Frequency (Hz) Damping (%)Mode PolyMAX CCSSI Diff (%) PolyMAX CCSSI Diff (%)
1 1.935 1.934 0.052 0.651 0.571 -12.2892 3.367 3.363 0.119 0.969 1.075 10.9393 4.721 4.667 1.157 1.512 2.406 59.127
4.3. Mode Shapes
The operational mode shapes of the three selected poles are presented alongside the FE model predictionof STX Finland in Figure 6. Vertical vibration in the bow is selected as the reference as this resulted ina clear stabilization diagram. The modes are identified as the 2-node (first), 3-node (second) and 4-node(third) normal bending modes of the vessel. Due to the fairly symmetrical geometry of the structure, thethree modes show little complex mixing of modes, i.e. modes comprised of a combination of bendingand torsion.
(a) f = 1,94 Hz, 2-node vertical bending. (b) f = 2,60 Hz, 2-node vertical bending.
(c) f = 3,37 Hz, 3-node vertical bending (d) f = 4,28 Hz, 3-node vertical bending
(e) f = 4,72 Hz, 4-node vertical bending (f) f = 5,63 Hz, 4-node vertical bending
Figure 6: Mode shapes for the first three vertical bending modes. OMA models generated using LMS are on theleft, and FE models developed by STX Finland are on the right.
The FE model shows the nominal vectors, which do not have physical dimensions. The surroundingwater has been taken into account by the addition of mass to the mass matrix connected to the wetsurface of the shell. The method is based on green functions of pressure distribution, and describes thecase when the vessel is located in deep water with a draught of 7,7m [2].
The analytical FE natural frequencies are larger than the measured OMA natural frequencies by 34 %,
27 % and 19 % respectively. The FE model draught is 0,9 m deeper than the draught during OMAmeasurements. The effect of adding mass in the form of cargo and fuel is however expected to furtherdecrease the measured OMA natural frequencies, due to the relationship between the natural frequency(ωn), mass (m) and stiffness (k) matrices:
ωni =
√kimi
(1)
This implies that the stiffness of the elements used in the FE model may be inaccurate. The effects ofgreater hull surface area exposed to water and the mooring boundary conditions on the natural frequen-cies and damping however need to be further investigated. The absence of torsional modes from theOMA is thought to be due to the structure not being physically excited to measurable amplitudes in atorsional manner by waves and wind in the harbour mooring.
4.4. Modal Assurance Criterion (MAC) Matrix
The MAC matrix provides a quantitative comparison between mode shapes, were mode shapes are ex-pected to be independent of one another and comprised of orthogonal vectors. The MAC matrix for thethree modes identified using Operational PolyMAX is presented in Figure 7a, were it can be seen thatmode pair 1-2 as well as 2-3 have low MAC values and are therefore decoupled, while mode pair 1-3show a 69 % correlation. This is due to the physical similarity between the 1st and 3rd bending modeshapes and similar results were found by Orlowitz [3]. A cross MAC matrix can be computed to deter-mine the statistical similarity between the different algorithms as well as between OMA and FE results,and is proposed for future research.
(a) MAC - Isometric view.
(b) Complexity Plot - Mode 2.
Figure 7: MAC matrix and Complexity Plot.
4.5. Complexity Plots
The modes are validated by plotting the components of the Eigenvectors in the complex plane. Theresulting complexity plots for the second mode are presented in Figure 7b, with the real values on the x-axis and imaginary values on the y-axis. The complexity plots show that all the DOF are nearly in phasewithin each mode. This means that the Eigenvectors reach their respective maximums and minimums atthe same time, and that there is little complex mixing of modes. This confirms the proportional dampingmodel approximation for lightly damped structures which expects real valued mode shapes [4].
5. CONCLUSIONS
Operational Modal Analysis (OMA) is used to investigate the structural dynamic characteristics of thevessel. The LMS Operational PolyMAX frequency domain and ARTeMIS CCSSI time domain OMAtechniques are used to estimate the modal parameters. Three stable modes are identified at 1,94 Hz,3,37 Hz and 4,72 Hz and show agreement to within 1,2 % by both Operational PolyMAX and CCSSI.The damping estimates show less agreement, especially for mode 3 which differs by 59 %. The threemodes are identified as the 2-node (first), 3-node (second) and 4-node (third) normal bending modes.The MAC confirms three unique modes, with cross coupling between mode pair 1-3 due to the geomet-rical similarities. The complexity plots reveal real valued mode shapes which confirm the proportionaldamping model approximation for lightly damped structures and thus validate the results.
The natural frequencies predicted by the FE model are greater than those measured using OMA by 34 %,27 % and 19 % respectively. The FE calculations are based on a vessel draught of 7,7 m which is deeperthan the 6,8 m draught during OMA testing. The effect of adding mass to the structure is howeverexpected to further reduce the OMA natural frequencies. Investigations into the effect of boundary andoperation conditions is recommended for further insight into the modal parameters.
ACKNOWLEDGMENTS
The authors would like to thank The Department of Environmental Affairs, South Africa for allowing usto perform measurements on their vessel. Furthermore we acknowledge our project partners namely STXFinland, Aalto University, the University of Oulu, Aker Arctic, Rolls-Royce, DNV and Wartsila. Wewould also like to thank COMNAP for funding the on-going research and making this trip to Gijon, Spainpossible. We gratefully acknowledge the support of the National Research Foundation and Departmentof Science and Technology under the South African National Antarctic Programme for project funding.
REFERENCES
[1] Rosenow, S.-E. (2007) Identification of the dynamic behaviour of marine construction structures.Ph.D. thesis, Universitat Rostock.
[2] Luosma, J. (2013) S.A. Agulhas II FE-Model. STX Finland.
[3] Orlowitz, E. and Brandt, A. (2014) Modal test results of a ship under operational conditions. IMACXXXIII Conference and Exposition on Structural Dynamics, Orlando, Florida.
[4] Rainieri, C. and Fabbrocino, G. (2014) Operational Modal Analysis of Civil Engineering Struc-tures. Springer, New York.
STRUCTURAL VIBRATION ANALYSIS ON THE POLAR
SUPPLY AND RESEARCH VESSEL THE S.A. AGULHAS II IN
ANTARCTICA
Keith Soal1, Anriëtte Bekker
1, Jörg Bienert
2
1 Stellenbosch University, Stellenbosch, SOUTH AFRICA
2 Technische Hochschule Ingolstadt, Ingolstadt, GERMANY
ABSTRACT
The Polar Supply and Research Vessel the S.A. Agulhas II plays a key role in the South
African National Antarctic Program, and is relied upon for logistical and research support in
Antarctica and the Southern Ocean. The vessel is exposed to various ice and open water
conditions and encounters a number of different excitation mechanisms which are capable of
causing structural fatigue. To this end full scale measurements were conducted during a 78
day voyage from Cape Town to Antarctica during 2013/2014 to investigate the effect of
vibration on the vessel’s structure. Nineteen vibration measurement channels were recorded
on the hull structure and five on the superstructure in order to determine the global vibration
response of the vessel. The structure is found to be most affected by vertical vibration during
open water navigation, with a maximum peak velocity of 338 mm/s in 8 m swells.
Comparatively, structural vibration levels in open water are greater than those measured
during ice navigation. According to Germanischer Lloyd’s (2001) ship vibration guidelines,
structural fatigue as a result of vibration is found to reach levels where damage is possible in
the stern and were damage is probable in the bow. The vibration levels with potential to cause
fatigue damage were measured in 8 m swells during open water navigation. This calls for
further investigations into the effects of the duration at these exposures, materials of
construction, structural details in the affected areas, welding processes and environmental
conditions. It is recommended that further investigations be conducted into the effects of
hybrid ice-open water designs on vibration response.
INTRODUCTION
Antarctic research institutes and their scientists continue to rely on polar supply and research
vessels (PSRVs) to supply bases and serve as floating laboratories. Vessels operating in
Antarctica and the Southern Ocean are exposed to various ice and open water conditions and
encounter a number of different excitation mechanisms, which are capable of causing
structural fatigue.
Advances in the various fields of marine engineering have enabled modern ship designs
which are lighter in weight and offer increased propulsion power. These advances are
weighed against structural integrity and fatigue life (Orlowitz and Brandt, 2014). The
importance of dynamic structural analysis in the design phase of vessels is therefore more
relevant now than ever before. The reliability of these analyses is based on real engineering
data and the extrapolation of such data on reasonable assumptions.
POAC’15
Trondheim, Norway
Proceedings of the 23rd International Conference on
Port and Ocean Engineering under Arctic Conditions June 14-18, 2015
Trondheim, Norway
Full scale measurements have played an important role in understanding the dynamic
responses of ice going vessels (Nyseth et al., 2013). These measurements are compared to
relevant standards which provide guidelines for the measurement, evaluation and reporting of
structural vibration. The current lack of high quality data has been cited as one of the most
important factors limiting further understanding of the effects of various excitation
mechanisms on ship dynamic responses (Dinham-Peren and Dand, 2010).
The S.A. Agulhas II is a state-of-the-art PSRV, and the work horse of the South African
National Antarctic Program (SANAP). The vessel spends around 8 months a year at sea in
some of the harshest operating conditions on the planet, and was designed with an operational
lifetime of 30 years. After sea trials, delivery of the vessel from Finland to South Africa and
the shakedown cruise, the vessel experienced cracking in the hull structure at the rear of the
main cargo hold. While this could just be due to large bending moments in the midship and
the need for some bending allowance in the hull structure, it does warrant a thorough
investigation into the structural dynamic performance of the vessel, especially concerning
structural fatigue caused by large bending and torsional moments.
This paper details the initial investigation into structural dynamic fatigue based on full scale
measurements during a 78 day voyage from Cape Town to Antarctica in 2013/2014. The post
processing and evaluation methods are presented and the results are compared to
Germanischer Lloyd’s (2001) ship vibration guidelines (Asmussen et al., 2001). This work is
intended to be a precursor to further investigations as guided by the results of the structural
vibration analysis.
THE S.A. AGULHAS II
Full scale measurements were conducted on-board the PSRV S.A. Agulhas II, see Figure 1,
which was built by STX Finland at the Rauma Shipyard. The S.A. Agulhas II is designed to
carry cargo, passengers, bunker oil, helicopter fuel and is also equipped with laboratories, a
moon pool and drop keel to conduct scientific research in the Southern Ocean. The ship’s
main specifications are presented in Table 1.
Figure 1- The S.A. Agulhas II.
Table 1- The main specifications of the S.A. Agulhas II.
Length, bpp 121,8 m
Beam 21,7 m
Draught, design 7,65 m
Deadweight at design displacement 5000 t
Installed power 4 x Wärtsilä 6L32 3000 kW
Propulsion Diesel-electric 2 x 4500 kW
Speed, service 14 kn
MEASUREMENT EQUIPMENT
The data acquisition system (DAQ) and measurement equipment used is presented in Table 2.
The LMS SCADAS were configured in a master-slave setup which allows simultaneous
measurements controlled from one DAQ. The LMS SCADAS are equipped with a low-pass
anti-aliasing filter. Accelerometers were calibrated according to the South African Bureau of
Standards (SABS) by the National Metrology Institute of South Africa (NMISA).
Table 2- Measurement equipment.
Equipment
1 x 16 channel, LMS SCADAS
1 x 12 channel, LMS SCADAS
1 x 8 channel, LMS SCADAS
9 x DC PCB accelerometers, 20,4 mV/(m/ )
9 x ICP PCB accelerometers, 10,2 mV/(m/ )
3 x Seismic PCB accelerometers, 1019,4 mV/(m/ )
1 x Triaxial PCB accelerometers, 10,2 mV/(m/ )
LMS Test.Lab 11A Turbine Testing software
MEASUREMENT SETUP
Nineteen vibration measurement channels were recorded on the hull structure and five on the
superstructure as seen in Figure 2. The measurement locations were chosen as follows:
Measurements were conducted in the bow and stern to investigate the effect of hull
slamming in open water as well as ice breaking and reversing during ice navigation.
Measurements were conducted in the cargo hold and engine store room to determine
the effect of global bending modes on the midship. Cracks had occurred prior to the
current measurements in the cargo hold which is a further justification.
Measurements were conducted in the bridge in order to investigate the effect of
superstructure vibration on structural fatigue.
Vertical vibration (+Z) was measured on the port and starboard sides of the hull and at
as close to equal increments along its length, in order to investigate the vibration
response caused by normal bending and torsional modes.
Lateral vibration (+Y) was measured in the hull to determine the response of
transverse bending.
Longitudinal vibration (+X) was measured in the superstructure to investigate fore aft
bending.
Lateral vibration (+Y) was measured in the superstructure to investigate transverse
bending and vertical vibration (+Z) was measured to investigate torsion.
The measurement system was controlled by a single laptop mounted in the central
measurement unit (CMU) measurement rack. A fibre optic cable was routed through water
tight cable trays from the CMU to the steering gear room to enable synchronous
measurements using the master-slave setup. Accelerometers were mounted to girders,
transverse beams or longitudinal beams using super-glue. Rigid structural members were
chosen in order to measure the global ship vibration response.
Figure 2 – Acceleration measurement locations on the S.A. Agulhas II.
DESCRIPTION OF THE VOYAGE
The voyage started from Cape Town harbour on the 28th of November 2013 and lasted 78
days. Figure 3 shows the track of the S.A. Agulhas II and is followed by a brief description of
each leg. Please note that the numbers of the descriptions match those in Figure 3.
Figure 3 - GPS data of the 2013/2014 voyage to Antarctica.
1. The first leg saw the S.A. Agulhas II depart from Cape Town harbour in a South
Westerly direction on the Good Hope line, turning due South on the Greenwich
Meridian. Vibration measurements began on the 4th December 2013. Wave heights
averaged 3,8 m with a maximum of 8 m. The navigating officers measure the wind
speed and then estimate the wave height using the Beaufort scale.
2. The ship entered ice on the 7th December 2013. The ice began as pancake, brash and
small ice floes. Floe ice became thicker and more concentrated as the ship moved
deeper into the pack ice of the Weddell Sea. On the 9th December 2013 the ship
encountered thick pack ice with large ice ridges, and became stuck (beset). Ramming
techniques were used to break through thick ridges and ice floes. Large ridges and
thick ice floes resulted in the ship being beset, often for several hours, and ramming
numerous times over the next 13 days.
3. On the 22nd December 2013 the ship arrived at Atka Bay to begin offloading cargo
for the German Neumayer Station III. In order to reach the ice shelf the ship first
carved away the bay ice.
4. On the 24th December 2013 the ship departed for Penguin Bukta and arrived on 25th
December 2013. The ship offloaded cargo and fuel at Penguin Bukta and remained
pushed up against the ice shelf.
5. The ship departs on the 30th December 2013 for Southern Thule. Navigation
progresses well through pack ice with occasional bergy bits. The ice pack begins to
thin and open up, and the ship reaches Southern Thule on the 4th January 2014.
6. The ship departs on the 4th January 2014 for South Georgia. Shortly after departure
the ship enters open water, with an average wave height of 4,3 m and a maximum of
7nm. The ship arrives at Grytviken Bay, South Georgia on the 6th January 2014.
7. The ship departs from South Georgia on the 6th January 2014 to conduct 13 days of
whale research on the edge of the ice pack.
8. The ship enters the ice pack on the 23rd January 2014 and reaches Penguin Bukta on
the 24th January 2014 to begin flying passengers back from SANAE IV base.
9. On the 26th January 2014 the ship departs for Atka Bukta and arrives on the 27th
January 2014. The ship then carves bay ice until the 28th January 2014 to complete
back loading.
10. On the 31st January 2014 the ship sets sail for Cape Town. Ice navigation progresses
well and the ship enters open water on the 1st February 2014. Scientific stations see
the ship stop several times along the Good Hope line. The average wave height during
the 11 day return voyage is 4 m with a maximum of 7 m. The ship enters Cape Town
harbour on the 13th February 2014.
DATA PROCESSING
Structural vibration data is post-processed according to BS ISO 20283-2 (2008) which
provides guidelines for the measurement of vibration on ships. The raw acceleration data is
first converted from g to mm/ , and then integrated using the Matlab trapz.m function to
mm/s. It is then decimated from 2048 Hz to 256 Hz using decimate.m which first low-pass
filters the data with a cut-off frequency of 102,4 Hz before re-sampling. The signal is then
high-pass filtered to remove vibration amplitude in the rigid body bandwidth. BS ISO 20283-
2 (2008) specifies that data should be high-pass filtered above 2 Hz.
Figure 4 - Chebyshev high-pass filters.
Two high-pass filters were designed in order to effectively attenuate low frequency vibration
measured by the ICP and DC accelerometers respectively. The filters were designed using
Matlab's Filter Design and Analysis Tool. A Chebyshev high-pass filter with an order
Nn=n800, and a cut-off frequency = 1 Hz (Figure 4a and b) was used to filter the ICP data.
A higher order filter was required for the DC accelerometers which are able to measure low
frequency vibration. A Chebyshev high-pass filter with an order N = 1400, and a cut-off
frequency = 1.6 Hz (Figure 4c and d) was used to filter the DC data. The steepness and
complexity of the frequency response curve is determined by N, the filter order (Smith, 2007).
The Chebyshev finite impulse response (FIR) filters were selected due to their sharp drop off
and low ripple at 0 dB. The high filter order provides significant attenuation below the cut-off
frequency. While higher filter orders are more computationally expensive, they do not effect
the filter accuracy and it was decided that longer computational times was a necessary trade-
off. Structural vibration metrics were subsequently calculated which include peak velocity
values and frequency spectra.
RESULTS
Peak Vibration Velocity
The peak vibration velocity values are presented in Figure 5 and 6. The vertical dashed lines
indicate the events listed in the Description of the Voyage. The crosses indicate the maximum
peak vibration velocity values in the vertical, lateral and longitudinal directions. The
following sensors are plotted in Figure 5: Steering Gear Stb X, Y and Z, Bow Centre Y and
Stb Z, CMU Triaxial X, Y, Z, Cargo Hold Stb Z. Vibration in the Bridge Stb X, Y, Z is
presented in Figure 6.
Figure 5 - Structural Vibration. Longitudinal vibration (+X), Lateral vibration (+Y),
Vertical vibration (+Z)
0 50 100 150 200 250 300-2
-1.5
-1
-0.5
0
0.5
1
X: 230.8
Y: -1.849
Time (s)
Vib
ratio
n A
mp
litu
de
(g
)
The following observations are made:
1. The largest peak vibration value is 338,35 mm/s in the vertical (+Z) direction in the
steering gear room of the vessel in open water.
2. The largest maximum values occur in open water in the vertical (+Z) direction at all
the measurement locations.
3. Vibration amplitude is largest in the bow and stern, and decreases towards the middle
of the vessel, increasing vertically towards the bridge.
4. The peak structural vibration in the Bridge is roughly a third of that experienced in the
steering gear room.
Figure 6 - Structural vibration. Longitudinal vibration (+X), Lateral vibration (+Y),
Vertical vibration (+Z)
The raw acceleration time history of the largest peak vibration value is shown in Figure 7. It
can be seen that an impulse of -1,85 g was measured in the vertical (+Z) direction in the
steering gear room on the starboard side. The resulting acceleration time signal after
integrating to velocity, decimating and high pass filtering the signal is shown in Figure 8.
Figure 7 - Raw acceleration time history of the maximum peak velocity value.
0 50 100 150 200 250 300-300
-200
-100
0
100
200
300
400
X: 233.5
Y: 338.4
Time (s)
Vib
ratio
n A
mp
litu
de
(m
m/s
)
Frequency Spectra
The frequency spectra of the maximum peak values are presented in Figure 9. The PSDs are
shown in the left hand column and the FFTs in the right hand column. The FFTs are
compared to Germanischer Lloyd's (2001) guidelines for structural fatigue as a result of
vibration. Two limit curves define the lower region in which vibration is unlikely to cause
damage, the intermediate region where vibration may cause damage, and the upper region in
which damage is probable.
The PSDs provide an accurate estimate of the frequency content of the random signals since
they are based on the FFT of the autocorrelation function, which is a statistical indicator
(Inman, 2014). The PSDs are calculated in Matlab using the following input parameters:
Flattop window, 50 % overlap, block size of 4096 NFFT points, sample frequency of 256 Hz.
This results in a frequency resolution of 0,0625 Hz. Two distinct peaks at 2,06 Hz and
3,88nHz can be seen in the PSD plots at all the measurement locations.
The FFT plots are seen to contain peaks at the same frequencies. Results show that vibration
in the stern of the vessel reaches amplitudes at which vibration may cause damage, see Figure
9, and that vibration in the bow of the vessel reaches amplitudes at which damage is probable.
These vibration amplitudes occur at 2,06 Hz in both locations. This is identified as the 2-node
first bending mode of the vessel (Soal and Bekker, 2014). These findings highlight the need
for further investigation into the duration of possible fatigue exposure experienced by the
vessel each year, and the effect this will have on its expected service life of 30 years. Other
factors which also need to be investigated include the type of steel used in construction, the
structural details in critical areas, the welding processes, the production methods and
environmental conditions such as corrosion.
Figure 8 - Post-processed velocity time history of the maximum peak velocity value.
Figure 9 - Structural vibration. Longitudinal vibration (+X), Lateral vibration (+Y),
Vertical vibration (+Z)
CONCLUSION
The structure of the vessel is found to be most affected by vertical vibration during open
water navigation, with a maximum peak velocity of 338,35 mm/s in 8 m swells. Vibration
levels are larger in the bow and stern, decreasing towards the centre of the vessel and
increasing again as you move vertically to the bridge.
Structural fatigue as a result of vibration is found to reach the level where damage is possible
in the stern and were damage is probable in the bow according to Germanischer Lloyd's ship
vibration guidelines. The vibration levels with potential to cause fatigue damage were
measured in 8 m swells during open water navigation. This calls for further research into the
effects of the duration at these exposures, materials of construction, structural details in the
affected areas, welding processes and environmental conditions. The occurrence of cracks on
the ship hull in the cargo hold prior to these measurements provide justification for further
research into structural health monitoring and damage detection.
ACKNOWLEDGEMENTS
The authors would like to thank The Department of Environmental Affairs, South Africa for
allowing us to perform measurement on their vessel. Furthermore we acknowledge our project
partners namely STX Finland, Aalto University, the University of Oulu, Aker Arctic, Rolls-
Royce, DNV and Wärtsilä. We would also like to thank COMNAP for funding the on-going
research and enabling travel to Trondheim, Norway. We gratefully acknowledge the support
of the National Research Foundation and Department of Science and Technology under the
South African National Antarctic Programme for project funding.
REFERENCES
Asmussen, I., Menzel, W. and Mumm, H. (2001). Ship Vibration. Tech. Rep., GL
Technology, Germanischer Lloyd, Hamburg.
BS ISO 20283-2:2008 (2008). Mechanical Vibration - Measurement of Vibration on Ships-
Part 2: Measurement of Structural Vibration. International Organization for Standardization
(ISO).
Dinham-Peren, T.A. and Dand, I.W., 2010. THE NEED FOR FULL SCALE
MEASUREMENTS. In: William Froude Conference: Advances in Theoretical and Applied
Hydrodynamics - Past And Future, November. Portsmouth, UK.
Inman, D.J., 2014. Engineering Vibration. Fourth edition. Pearson, Essex, England.
Nyseth, H.v., Frederking, R. and Sand, B.r., 2013. Evaluation of Global Ice Load Impacts
Based on Real-Time Monitoring of Ship Motions. In: 22nd International Conference on Port
and Ocean Engineering Under Arctic Conditions (POAC’13). Espoo, Finland.
Orlowitz, E. and Brandt, A., 2014. Modal test results of a ship under operational conditions.
IMAC XXXIII Conference and Exposition on Structural Dynamics, Orlando, Florida.
Smith, J.O., 2007. Introduction to Digital Filters with Audio Applications. Center for
Computer Research in Music and Acoustics (CCRMA), Stanford University.
Soal, K.I., Bekker, A., 2014. Vibration Response of the Polar Supply and Research Vessel the
S.A. Agulhas II in Antarctica and the Southern Ocean. Master’s thesis. Stellenbosch
University, South Africa.
IOMAC Gijon (Spain) 1
Thank You