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, keithsoal@gmail.com.2 Prof, Technische Hochschule Ingolstadt, Joerg.Bienert@thi.de.3 Dr, Stellenbosch University, annieb@sun.ac.za.

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