Vibrations in an active

controlled hexapod
Wouter Diederen

Date: March 2020

Thesis committee: Prof. Dr. ir. A. Metrikine, TU Delft, Chairman
Ir. V. Vaniushkina TU Delft, Supervisor
Ir. J.J. Flohil Ampelmann B.V.
Ir. O. Kooy Ampelmann B.V.
Acknowledgement
I would like to express my sincere gratitude to Prof. Andrei Metrikine, chairman of the graduation committee
who gave me the opportunity to graduate under his supervision. The same applies for Valentina
Vaniushkina, who performed the role of daily supervisor. This project has been made possible by
Ampelmann. I would like to thank all the people from Ampelmann who helped me during my work. It was a
pleasure working there. I especially would like to thank Jasper Flohil and Otto Kooy for their supervision
and guidance. Without their support and input, I could not have completed this research successfully.

i
Abstract
Safe offshore access for people and cargo is a major challenge in the offshore industry. The Ampelmann
system is an active motion compensated system for six degrees of freedom. Creating a platform isolated
from the motions of the vessel making offshore access as easy as crossing the street. All the Ampelmann
system can be used for people transfer, while some of the systems can also be used as a cargo crane. The
basic system can be divided in two main systems: the hexapod and the transfer deck and gangway. Usually
the system is installed on a ship deck. However, in various cases the height of the Ampelmann system is
not sufficient to reach the landing point. An often-used solution is to place the system on a pedestal which
can be over 15 meters high. The Ampelmann system occasionally starts vibrating unexpectedly, especially
while the system is placed on a pedestal. These vibrations are believed to be caused by the
eigenfrequencies of the system and/or amplification caused by the motion control algorithm. In this research
an investigation in this phenomenon was done.

This investigation was done via an analysis of the eigenfrequencies of the active controlled hexapod. A
finite element method model was created to determine the eigenfrequencies of the system. This model was
made using MATLAB and the toolbox StaBIL 2.0, created by the university of Leuven. All the elements of
the system are modeled as beams except the hydraulic actuators. The properties of the elements which
represent the hydraulic actuators are calculated separately using a modelling study on stiffness
characteristics of hydraulic cylinder under multi-factors.

Possible causes for the unexpected vibrations have been investigated via measurements performed on
Ampelmann systems. Data which was readily available is analyzed. Based on this data three possible
causes have been determined. These are: the influence of the pedestal, residual motions due to limitations
of the Ampelmann system and vibrations in the bottom frame due to compensating for gangway motions.

To investigate the influence of the pedestal and the vibrations in the bottom frame two experiments have
been performed. The residual motions have been investigated via calculations based on data already
available. The amplitude of the response in the results from the first experiment performed to investigate
the vibrations in the bottom frame due to gangway motions is negligible over the entire test period. The
data from the calculations done to investigate the effect of the residual motions show no amplification. The
results of the experiment and the calculations lead to the conclusion that these do not cause unwanted
behavior.

The second experiment to determine the influence of the pedestal shows four peaks in the frequency
domain of both the signals. The first is directly caused by vessel motions. The other three have a cause
which is not directly related to vessel motions. The data from the sensor on the Ampelmann system, at the
top of the pedestal, does not contain a peak which is not present at the ship deck. From this it can be
concluded that the eigenfrequencies of the pedestal do not have a relevant influence. For one of these
peaks the amplitude of the graph related to the top of the pedestal is higher than the one corresponding to
the ship deck. A possible explanation for this phenomenon could be the eigenfrequencies and
corresponding eigenmodes of the ship deck. The pedestal functions as a leaver arm amplifying the rotations
related to eigenmodes of the ship deck. This may cause the unexpected vibrations.

Ampelmann’s motion control algorithm is complicated which makes it difficult to incorporate this algorithm
in the FEM model. An attempt is made to create a simplified 2D multi-body model including the algorithm.
At Ampelmann a model made by Temporary Work Design was available, consisting out of two masses,
which should have funtioned as the base for this model. During the investigation of the model an error was
discovered. The correction of this error resulted in an unstable system. A pole plot investigation is done.
However, no explanation for the instablity is found.

ii
Table of contents
1. INTRODUCTION 1

RESEARCH OBJECTIVE 3
THESIS OUTLINE 3

2. MOTION CONTROL ALGORITHM 4

CONTROL ENGINEERING 4
2.1.1 FEEDBACK 4
2.1.2 CONTROL 4
AMPELMANN’S MOTION CONTROL ALGORITHM 5
2.2.1 SIGNAL PROCESSING 5
2.2.2 INVERSE KINEMATICS 5
2.2.3 ACTUATORS CONTROLLER 6

3. EIGENFREQUENCIES AMPELMANN SYSTEM 8

BEAM ELEMENT 9
3.1.1 AXIAL STIFFNESS 9
3.1.2 BENDING MOMENTS 10
HYDRAULIC ACTUATOR 11
3.2.1 CYLINDER BARREL EXPANSION STIFFNESS 12
3.2.2 HYDRAULIC OIL STIFFNESS 12
FEM MODEL 13
3.3.1 A-TYPE 13
3.3.2 E-TYPE 15
CONCLUDING REMARKS 17

4. EXPERIMENTAL RESEARCH 18

EXPERIMENTAL FREQUENCY ANALYSIS 18

4.1.1 DISCRETE FOURIER TRANSFORM 18
4.1.2 SPECTRUM AVERAGING 19
4.1.3 FORWARD KINEMATICS OF A STEWART PLATFORM 20
PRE-PROCESSING 21
4.2.1 A-TYPE AND E-TYPE 21
POSSIBLE CAUSES OF UNEXPECTEDLY OCCURRING VIBRATIONS 25
4.3.1 INFLUENCE OF THE PEDESTAL 26
4.3.2 RESIDUAL MOTIONS 27
4.3.3 VIBRATIONS IN THE BOTTOM FRAME DUE TO COMPENSATING FOR GANGWAY MOTIONS 28

iii
 MEASUREMENTS ON THE AMPELMANN SYSTEM 29
4.4.1 INFLUENCE OF THE PEDESTAL 29
4.4.2 RESIDUAL MOTIONS 35
4.4.3 VIBRATIONS IN THE BOTTOM FRAME 37
CONCLUDING REMARKS 39

5. ANALYTICAL MODEL 40

MODEL BY TEMPORARY WORK DESIGN 40

DIFFERENTIATION ERROR 42
POLE PLOT INVESTIGATION 43
CONCLUDING REMARKS 44

6. CONCLUSION AND RECOMMENDATIONS 45

CONCLUSION 45
RECOMMENDATIONS 46

7. REFERENCES 47

8. APPENDIX 49

iv
List of figures
FIGURE 1-1: AMPELMANN SYSTEM ........................................................................................................................................1
FIGURE 1-2: SIMPLIFIED CONTROL LOOP .................................................................................................................................1
FIGURE 1-3: GENERIC A FRAME ............................................................................................................................................2
FIGURE 2-1: CLOSED LOOP (LEFT); OPEN LOOP (RIGHT) .............................................................................................................4
FIGURE 2-2: BOTTOM AND TOP FRAME COORDINATE SYSTEM .....................................................................................................6
FIGURE 2-3: AMPELMANN’S MOTION CONTROL ALGORITHM.......................................................................................................7
FIGURE 3-1: HEXAPOD, TRANSFER DECK AND GANGWAY (NODES AND ELEMENTS) ...........................................................................8
FIGURE 3-2: BEAM ELEMENT ................................................................................................................................................9
FIGURE 3-3: AXIAL STIFFNESS BEAM ELEMENT ........................................................................................................................10
FIGURE 3-4: BENDING STIFFNESS BEAM ELEMENT ...................................................................................................................10
FIGURE 3-5: HYDRAULIC ACTUATOR .....................................................................................................................................11
FIGURE 3-6: HYDRAULIC SYSTEM [7]; 1 CYLINDER BARREL, 2 HYDRAULIC OIL, 3 PISTON SEALING, 4 ROD SEALING, 5 PISTON ROD, 6 FLEXIBLE
HOSE, 7 METAL PIPE .................................................................................................................................................. 11
FIGURE 3-7: FEM MODEL A-TYPE SYSTEM ............................................................................................................................13
FIGURE 3-8: MODE SHAPES 1 TO 4 A-TYPE SYSTEM ................................................................................................................14
FIGURE 3-9: FEM MODEL E-TYPE SYSTEM.............................................................................................................................15
FIGURE 3-10: MODE SHAPES 1 TO 4 E-TYPE SYSTEM...............................................................................................................16
FIGURE 4-1: SUMMARY OF DISCRETE FOURIER TRANSFORM [11]...............................................................................................19
FIGURE 4-2: SEGMENT-BASED FREQUENCY DOMAIN AVERAGING [12] ........................................................................................20
FIGURE 4-3: LOCATION OF THE MOTION SENSORS ...................................................................................................................22
FIGURE 4-4: TRANSLATIONS A-TYPE SYSTEMS ........................................................................................................................22
FIGURE 4-5: ROTATIONS A-TYPE SYSTEMS .............................................................................................................................23
FIGURE 4-6: DIFFERENCE BETWEEN MRU 1 AND MRU 2 IN HEAVE DIRECTION FOR A-TYPE SYSTEMS ...............................................23
FIGURE 4-7: TRANSLATIONS E-TYPE SYSTEMS.........................................................................................................................24
FIGURE 4-8: ROTATIONS E-TYPE SYSTEMS .............................................................................................................................24
FIGURE 4-9: DIFFERENCE BETWEEN MRU 1 AND MRU 2 IN HEAVE DIRECTION FOR E-TYPE SYSTEMS ...............................................25
FIGURE 4-10: OVERVIEW AMPELMANN SYSTEM.....................................................................................................................25
FIGURE 4-11: CROSS SECTION OF AMPELMANN SYSTEM; VIBRATIONS CAUSED BY A PEDESTAL .........................................................26
FIGURE 4-12: CROSS SECTION OF AMPELMANN SYSTEM; RESIDUAL MOTIONS ..............................................................................27
FIGURE 4-13: CROSS SECTION OF AMPELMANN SYSTEM; VIBRATIONS IN THE BOTTOM FRAME .........................................................28
FIGURE 4-14: MRU LOCATIONS..........................................................................................................................................30
FIGURE 4-15: SHIP TO SHIP MRU .......................................................................................................................................30
FIGURE 4-16: LEAVER ARMS (DX & DZ)................................................................................................................................31
FIGURE 4-17: E-04 TRANSLATIONS (TIME DOMAIN) ...............................................................................................................32
FIGURE 4-18: E-04 ROTATIONS (TIME DOMAIN) ....................................................................................................................32
FIGURE 4-19: E-04 TRANSLATIONS (FREQUENCY DOMAIN).......................................................................................................33
FIGURE 4-20: E-04 ROTATIONS (FREQUENCY DOMAIN) ...........................................................................................................33
FIGURE 4-21: SCHEMATIC OVERVIEW SHIP DECK VIBRATIONS ....................................................................................................34
FIGURE 4-22: CYLINDER LENGTHS AND REFERENCE MOTIONS ....................................................................................................35
FIGURE 4-23: RESIDUAL MOTIONS (TIME AND FREQUENCY DOMAIN) ..........................................................................................36
FIGURE 4-24: OVERVIEW EXPERIMENT VIBRATIONS IN THE BOTTOM FRAME .................................................................................37
FIGURE 4-25: A-20 TRANSLATIONS AND ROTATIONS (ENTIRE SAMPLED TIME)..............................................................................38
FIGURE 4-26: A-20 TRANSLATIONS AND ROTATIONS (SAMPLED TIME SERIES DURING SINGLE BLOW WITH GANGWAY) ..........................38
FIGURE 5-1: OVERVIEW OF DESIRED RESULT ..........................................................................................................................40
FIGURE 5-2: TWD MODEL [15] ..........................................................................................................................................41
FIGURE 5-3: INTRODUCTION OF CYLINDER MODEL [15] ...........................................................................................................41

v
FIGURE 5-4: CONTROL LOOP IN ANALYTICAL MODEL ................................................................................................................42
FIGURE 5-5: POLE-ZERO MAP OF CLOSED LOOP, LEFT WITH ERROR & RIGHT WITHOUT ERROR .........................................................43
FIGURE 5-6: POLE ZERO MAPS; LEFT: PLANT, RIGHT: CONTROLLER .............................................................................................44
FIGURE 5-7: POLE ZERO MAPS; LEFT: LOWPASS FILTER FEED FORWARD, RIGHT: OPEN LOOP ............................................................44
FIGURE 8-1: STANDARD BLOCK DIAGRAM ELEMENTS [2] ..........................................................................................................49
FIGURE 8-2: TIME DOMAIN A-28 SYSTEM .............................................................................................................................51
FIGURE 8-3: TIME DOMAIN A-29 SYSTEM .............................................................................................................................51
FIGURE 8-4: TIME DOMAIN E-04 SYSTEM .............................................................................................................................52
FIGURE 8-5: TIME DOMAIN E-14 SYSTEM .............................................................................................................................52
FIGURE 8-6: A-28 WITHOUT PEDESTAL (UPPER), A-29 WITH PEDESTAL (LOWER)..........................................................................53
FIGURE 8-7: E-04 ON PEDESTAL (UPPER), OLYMPIC ORION, NO PICTURE WITH AMPELMANN SYSTEM AVAILABLE (LOWER) ...................53
FIGURE 8-8: MRU (OCTANS) SPECIFICATIONS .......................................................................................................................54
FIGURE 8-9: E-04 TRANSLATIONS DIFFERENCE (TIME DOMAIN) .................................................................................................55
FIGURE 8-10: E-04 ROTATIONS DIFFERENCE (TIME DOMAIN) ...................................................................................................55
FIGURE 8-11: E-04 TRANSLATIONS DIFFERENCE (FREQUENCY DOMAIN) ......................................................................................56
FIGURE 8-12: E-04 ROTATIONS DIFFERENCE (FREQUENCY DOMAIN) ..........................................................................................56
FIGURE 8-13: RESPONSE ANALYTICAL MODEL ........................................................................................................................57

vi
1. Introduction
Safe offshore access for people and cargo is a major challenge in the offshore industry. The Ampelmann
system is an active motion compensated system for six degrees of freedom. Creating a platform isolated
from the motions of the vessel making offshore access as easy as crossing the street. All the Ampelmann
system can be used for people transfer, while some of the systems can also be used as a cargo crane. The
basic system can be divided in two main systems: the hexapod and the transfer deck and gangway (Figure
1-1, left). Usually the system is installed on a ship deck. However, in various cases the height of the
Ampelmann system is not sufficient to reach the landing point. An often-used solution is to place the system
on a pedestal which can be over 15 meters high (Figure 1-1, right).

Transfer deck & gangway

Hexapod Pedestal

Figure 1-1: Ampelmann system

Hexapod
The hexapod is based on a Steward platform [1] consisting of a bottom frame, a base frame and six
hydraulic cylinders. The bottom frame is fixed to the vessel or pedestal. The transfer deck is connected to
the base frame. The hexapod or Steward platform is the system that compensates the vessel motions. The
control of the hexapod is based on the displacement and velocity measured by sensors located at the base
frame of the system. A simplified model of the control loop is given in Figure 1-2. In this figure Xref,global and
Xref,measured are the desired and measured location of the transfer deck and gangway and L cyl is the length
of the cylinders.

Figure 1-2: Simplified control loop

1
Transfer deck & gangway
Personnel accessing an offshore structure from a ship via an Ampelmann system commence their crossing
on the transfer deck. The system starts up after which the people transfer begins. Personnel walk over the
gangway to the offshore structure. The gangway can maneuver in three degrees of freedom, basically
resembling the boom of a crane.

Pedestal
The pedestals, which are used by Ampelmann to increase the possible height of the landing point, are
mostly individually designed. At the moment, Ampelmann is in the process of designing a generic A-frame
(GAF). This will be a modular based pedestal (Figure 1-3).

Figure 1-3: Generic A frame

A-type & E-type

Ampelmann employs three types of systems. The A-type, E-type and L-type. The L-type does not use a
hexapod so no investigation regarding this type will be done in this research. The A-type is an offshore
access system for performing safe transfer of personnel in sea states with a maximum significant wave
height of 3 meters. The system weights roughly 40 tons and has a gangway length of 20 to 25 meters. The
E-type is designed to safely transfer offshore personnel in rough sea conditions. The gangway system is
based on the same technology as the A-type, only 1.5 times larger. As a result of its increased size, the E-
type is capable of compensating sea states up to 4.5 meters significant wave height. The systems weights
roughly 100 tons and has a gangway length of 25 to 30 meters.

2
 Research objective
The Ampelmann system occasionally starts vibrating unexpectedly, especially while the system is placed
on a pedestal. These vibrations are believed to be caused by the eigenfrequencies of the system and/or
amplification caused by the motion control algorithm. At the moment, the knowledge present at Ampelmann
about these vibrations is mostly based on field experience. In this research an investigation in this
phenomenon will be done. To perform this the following research goal is formulated:

The objective of this thesis is performing an analysis on the eigenfrequencies of an active controlled
hexapod and explain the unexpectedly occurring vibrations in the Ampelmann system.

To reach the research goal a set of sub goals are formulated:

1. Design a model which can be used to calculate the eigenfrequencies and corresponding
eigenmodes of an Ampelmann system.
2. Investigate whether motion compensation caused by vibrating of the pedestal can cause the
unexpected vibrations in an Ampelmann system.
3. Determine the frequency content of the residual motions and compare them to the
eigenfrequencies of the gangway which might cause the unexpectedly occurring vibrations in an
Ampelmann system.
4. Investigate whether vibrations in the bottom frame, due to compensating for gangway motions, can
cause the unexpected vibrations in an Ampelmann system
5. Create a simplified multibody model incorporating the motion control algorithm which can explain
the unexpectedly occurring vibrations in an Ampelmann system.

The first sub goal, about the eigenfrequencies of the system, helps to reach the first part of the research
goal. The second, third and fourth aim to answer the second part of the research goal. These come down
to investigating whatever happens below the hexapod, what happens due to limitations of the hexapod and
what happens above the hexapod. The fifth sub goal aims to combine all the data into a single model and
explain the vibrations.

Thesis outline
This thesis consists out of a total of six chapters. The first step in this work is a brief discussion about control
engineering and the motion control algorithm used by Ampelmann in chapter 2. Next, in chapter 3, a finite
element method model is created to determine the eigenfrequencies of the system and sub-systems. In
chapter 4 some experimental research is done to identify and explain possible causes for the unexpectedly
occurring vibrations. In chapter 5 a simplified 2D multi-body model is created. However, due to an error,
which will be explained later, in a template model which was supposed to be used, a small research
regarding this mistake is done. Chapter 6 summarizes the conclusions of this research and adds
recommendations.

3
2. Motion control algorithm
The Ampelmann motion control algorithm (Figure 2-3) calculates and controls the required lengths of the
hydraulic actuators such that the transfer deck experiences no rotations and translations. The goal of this
chapter is to get a basic understanding of motion control engineering and Ampelmann’s algorithm. First
some general motion control engineering theory is given. Next the control algorithm is discussed. The
algorithm consists of three parts: Signal processing, kinematics and the controller. These three parts will
be discussed briefly.

Control engineering
Motion control engineering focuses on moving a load from one place to another by precisely controlling the
position, velocity and the acceleration of the load under defined operating conditions. The basic principles
of motion control can be divided in feedback and control. Feedback refers to the situation where two or
more dynamic systems are connected and influence each other’s behavior over time. Control is the design
of component of an engineering feedback system to achieve a desired behavior [2].

2.1.1 Feedback
A dynamical system is a system whose behavior changes over time. Feedback is about two or more
connected dynamic systems such that they influence each other’s behavior and are thus strongly coupled.
Simple causal reasoning about a feedback system is difficult because the first system influences the second
and vice versa. A consequence of this is that the behavior of a feedback system is often counterintuitive.
This makes it necessary to investigate the entire system. Two often used terms when referring to such
systems are open loop and closed loop systems (Figure 2-1). A system is closed loop when the systems
are connected in a circle and open loop when this circle is broken. Feedback has potential disadvantages:
It can create dynamic instabilities or even runaway behavior. A second drawback is that feedback can
introduce unwanted sensor noise into a system requiring careful signal processing and filtering of the sensor
signal [2].

Figure 2-1: Closed loop (left); Open loop (right)

2.1.2 Control
In the area of motion control engineering the term control is defined as the use of algorithms and feedback
in engineering systems. A modern controller senses the operation of a system, compares it to a desired
behavior (𝑒 = 𝑟 − 𝑦), and then computes corrective actions based on a model of the system’s response to
external inputs and actuates the system for a desired change. This loop of sensing, computation and
actuation is a central concept in control engineering [2].

Proportional Integral Derivative control

Using only an on-off control often causes oscillations because the system overreacts. A small error causes
a “full throttle” response in the system. This effect is avoided using proportional controller. The characteristic
of the controller is proportional to a control error for small errors (equation 2.1). In this equation kp is the
controller gain and 𝑒𝑚𝑖𝑛 = 𝑢𝑚𝑖𝑛 ⁄𝑘𝑝 and 𝑒𝑚𝑎𝑥 = 𝑢𝑚𝑎𝑥 ⁄𝑘𝑝 is the proportional band where the behavior of the
controller is linear.

𝑢𝑚𝑎𝑥 𝑖𝑓 𝑒 ≥ 𝑒𝑚𝑎𝑥
𝑢 = { 𝑘𝑝 𝑒 𝑖𝑓 𝑒𝑚𝑖𝑛 < 𝑒 < 𝑒𝑚𝑎𝑥 (2.1)
𝑢𝑚𝑖𝑛 𝑖𝑓 𝑒 ≤ 𝑒𝑚𝑖𝑛

4
A proportional controller is a huge improvement compared to an on-off controller. However, a proportional
controller has the drawback that some level of control signal is required for the system to maintain a desired
value, meaning 𝑒 ≠ 0. This can be avoided by making the control action proportional to the integral of the
error (equation 2.2), which is called an integral controller. This controller has zero steady state error [2].
The problem however is that there may not always be a steady state because the system is for example
oscillating.
𝑡
𝑢(𝑡) = 𝑘𝑖 ∫ 𝑒(𝜏) 𝑑𝜏 (2.2)
0

An additional refinement is to provide the controller with a predictive ability by using a prediction of the error.
The simplest form is using a linear extrapolation (equation 2.3) predicting the error Td units of time ahead.

𝑑𝑒(𝑡)
𝑒(𝑡 + 𝑇𝑑 ) = 𝑒(𝑡) + 𝑇𝑑 (2.3)
𝑑𝑡
Combining these three actions, proportional, integral and derivative control results in a PID controller which
is mathematically expressed in equation 2.4.
𝑡
𝑑𝑒(𝑡)
𝑢(𝑡) = 𝑘𝑝 𝑒(𝑡) + 𝑘𝑖 ∫ 𝑒(𝜏) 𝑑𝜏 + 𝑘𝑑 (2.2)
0 𝑑𝑡

Ampelmann’s motion control algorithm

The orientation of the bottom frame is acquired from a 6 degree of freedom motion sensor located at the
center of the bottom frame. Using this data, the necessary lengths of the hydraulic actuators are calculated
using inverse kinematics. These lengths are sent to a controller which controls the length of the cylinder
using a feed forward and multiple feedback loops.

2.2.1 Signal processing

The sensors measure, in 6 degrees of freedom, the motions of the vessel. These contain a high frequency
part for which the Ampelmann system should not compensate. This part of the signal is filtered using a low
pass filter. The S-curve is used to slowly start up the system. Instead of going to full compensation instantly,
the process is smoothened over a period of time. The main goal of this part of the motion control algorithm
is creating a signal in which all unwanted parts are removed resulting in a clean signal which can be used
for motion compensating.

2.2.2 Inverse kinematics

The pose of a Steward platform can be defined by the position and orientation of the top and bottom frame
with respect to each other [3]. The lengths of the 6 actuators can be determined with these orientations
using inverse kinematics. The orientation of the bottom frame is measured by a six degree of freedom
motion sensor. This sensor is located at the center of the bottom frame while the top frame is assumed to
be horizontal and motionless. Two local coordinate systems (Figure 2-2) are introduced one at center of
the top frame (Ot) and one at the center of the bottom frame (O b). The coordinates of the bottom joints are
expressed in the coordinates system of the top frame using a rotation matrix. The required lengths of the
hydraulic actuators can now be determined.

5
 Figure 2-2: Bottom and top frame coordinate system

2.2.3 Actuators controller

The Ampelmann controller is relatively complicated, consisting out of a saturation block, rate limiter, lookup
tables, S-curves etcetera. For details see appendix A. However, the general idea of the controller is the
following: The input of the controller is the error between the reference lengths and the actual lengths of
the actuators and the error between the reference velocity and actual velocity of the actuator. These errors
are translated into a certain pressure. This pressure results in the output. The output is the valve position
in the hydraulic actuators.

6
Figure 2-3: Ampelmann’s motion control algorithm

7
3. Eigenfrequencies Ampelmann system
To mimic the unexpectedly occurring vibrations, two computer models are made. The first model will be
used to determine the eigenfrequencies of the system. This model is made using MATLAB and the toolbox
StaBIL 2.0 [4], created by the university of Leuven resulting in a finite element method model. The finite
element method model will be used to determine eigenfrequencies and eigenmodes. The FEM model is
created using the toolbox StaBIL 2.0. The toolbox consists out of a set of MATLAB functions which perform
a matrix method based discrete-element idealization. StaBIL 2.0 is based on [5] & [6].

A discrete-element model of the Ampelmann system will be created. In Figure 3-1 an example of the
hexapod, transfer deck and gangway is given. The system is assumed to exist out of a set of beams coupled
by rigidly connected joints. The interaction forces between the various elements are represented by joint
forces. These joint forces are axial forces, shear forces, bending moments and torques. For each element
a local stiffness and mass matrix is composed which will be transferred to the global coordinates system
using a matrix. For elements which are long compared to their cross section dimension the elastic
characteristics can be determined accurately [5]. The pedestal and transfer deck and gangway are all
assembled using beam elements. The hexapod is also assembled using beam elements. However, the
stiffness and mass of elements which represent the hydraulic actuators are calculated separately using a
modelling study on stiffness characteristics of hydraulic cylinder under multi-factors [7].

Figure 3-1: Hexapod, transfer deck and gangway (nodes and elements)

8
 Beam element
Beam elements are assumed to be straight members of uniform cross section capable of resisting axial
forces, bending moments about its two principal axes in plane of its cross section and a twisting moment
about its centroidal axis. The location and positive direction of the degrees of freedom are given in Figure
3-2. The position of the beam element in space is specified by the location of the p-end of the beam and
the direction of the local x-axis. The stiffness and mass matrix for a uniform beam element are derived
directly from the differential equations for beam displacements in engineering beam theory [5] & [6]. The
entire stiffness matrix is given in appendix B. The two most dominant stiffness properties, axial and bending,
are worked out below. For the other stiffness parameters reference is made to [5].

Figure 3-2: Beam element

3.1.1 Axial stiffness

The differential equation for axial displacement is:

𝑑𝑢1
𝐹1 = − 𝐸𝐴 (3.1)
𝑑𝑥

This equation can be integrated and by assuming the following boundary conditions: The left end of the
beam at 𝑥 = 0 has a displacement u1 and the right end at 𝑥 = 𝑙 has zero displacement (Figure 3-3). This
results in the following solution:

𝐸𝐴
𝐹1 = 𝑢 (3.2)
𝑙 1

From equation 3.2 and force equilibrium in the x direction, the axial stiffnesses can be computed:

𝐸𝐴
𝑘1,1 = −𝑘7,1 = 𝑘7,7 = (3.3)
𝑙

9
 Figure 3-3: Axial stiffness beam element

3.1.2 Bending moments

The differential equation for deflection of the beam is:

𝑑 2 𝑢2
𝐸𝐼𝑧 = 𝐹2 𝑥 − 𝐹6 (3.4)
𝑑𝑥 2

The following boundary conditions are assumed:

𝑢2 = 𝑢8 = 0
𝑑𝑢8 𝐹2 (3.5)
=− 𝑎𝑡 𝑥 = 𝑙
𝑑𝑥 𝐺𝐴𝑠

Equation 3.4 can be integrated and by applying the boundary conditions, equilibrium and symmetry, the
bending stiffness can be determined:

(4 + 𝛷)𝐸𝐼𝑧
𝑘6,6 = 𝑘12,12 =
(1 + 𝛷)𝑙
(3.6)
12𝐸𝐼
𝛷=
𝐺𝐴𝑠 𝑙2

The bending stiffness in the direction of u5 and u11 can be determined in a similar way. Using the moment
of inertia (I) and the effective shear area (As) with respect to the y-axis.

Figure 3-4: Bending stiffness beam element

10
 Hydraulic actuator
Each hydraulic actuator is assumed to exist out of two elements (Figure 3-5). One represents the cylinder
and one represents the piston. For all the degrees of freedom except the axial directions both the cylinder
and the piston are modeled using the beam theory as explained in paragraph 3.1. However, the axial
stiffness of a hydraulic actuator is significantly affected by the characteristics of, for example the bulk
modulus of the hydraulic oil. The determination of the axial hydraulic actuator stiffness is based on [7].

Figure 3-5: Hydraulic actuator

The main factors affecting the hydraulic actuator stiffness (Figure 3-6) are the hydraulic oil stiffness [Ko],
the piston rod axial stiffness [Kr], the cylinder barrel expansion stiffness [Kp], the flexible hose expansion
stiffness [Kh] and the sealing ring deformation stiffness [Ks]. The total stiffness can be determined as the
sum of the reciprocals of the all the stiffnesses [7]. The level of influence of each stiffness according to [7]
is: the hydraulic oil stiffness is about 80 %, the expansion deformation of the cylinder barrel is about 10 %
and the axial deformation of the piston rod is about 6 % of the total. The other factors are smaller than 3 %
and are neglected because of their small contribution. The cylinder actuator stiffness can be calculated
using equation 3.7. The piston rod in general is a solid cylindrical steel rod. The axial stiffness is calculated
based on beam theory, as given in paragraph 3.1.

1 1 1 1
= + + (3.7)
𝐾 𝐾𝑜 𝐾𝑟 𝐾𝑐

Figure 3-6: Hydraulic system [7]; 1 cylinder barrel, 2 hydraulic oil, 3 piston sealing, 4 rod sealing, 5 piston rod, 6
flexible hose, 7 metal pipe

11
3.2.1 Cylinder barrel expansion stiffness
The cylinder barrel expansion stiffness is produced by the movement of the piston as a result of the radial
cylinder expansion caused by a pressure change [ΔP]. The radial deformation [ΔD] due to a pressure
change can be computed using equation 3.8, with Do as the outer barrel diameter Di as the inner barrel
diameter, Eb is Young’s modulus and υb as the Poisson ratio.

𝐷∆𝑃 𝐷𝑜2 + 𝐷𝑖2 𝐷𝑜2 + 𝐷𝑖2

∆𝐷 = ( + 𝜐𝑏 ) ; 𝜆𝑐 = (3.8)
𝐸𝑏 𝐷𝑜2 − 𝐷𝑖2 𝐷𝑜2 − 𝐷𝑖2

By calculating the volumetric change as a result of the radial deformation caused by the pressure change
and using the spring stiffness equation 𝐾𝑠𝑝 = 𝑑𝑓 ⁄𝑑𝑥, the cylinder barrel expansion stiffness is given by
equation 3.9. In this equation A is the cross section of the cylinder, L is the length of the cylinder and V is
the volume of the cylinder.

∆𝑃𝐴2 𝐸𝑏 𝐴 1
𝐾𝑐 = = ∗ (3.9)
∆𝑉 2𝐿 𝜆𝑐 + 𝜐𝑏

3.2.2 Hydraulic oil stiffness

The hydraulic oil stiffness can be computed using the following equations:

𝐴2
𝐾𝑜 = 𝐸𝑜 (3.10)
𝑉
The oil bulk modulus [E0] is influenced by the amount of air in the fluid and the pressure of the fluid. The oil
bulk modulus can be described using equation 3.11 [8]. In this equation, E0’ is the oil bulk modulus without
air, pa is the atmospheric pressure, α is the relative air content in oil under atmospheric pressure and n is
the isentropic coefficient (n = 1.4).

𝑝𝑎 1⁄𝑛
1 + 𝛼 (𝑝 + 𝑝)
𝑎
𝐸𝑜 = 𝐸𝑜′ ⁄
(3.11)
𝑝𝑎1 𝑛
1 + 𝛼 𝐸0
𝑛(𝑝𝑎 + 𝑝)(𝑛+1)⁄𝑛

12
 FEM model
3.3.1 A-type
In Figure 3-7 the FEM model of the A-type on a pedestal is plotted. The pedestal consists out of two GAF
modules. To get some confidence the mass of the model is calculated and compared to masses from real
systems. The mass is calculated using the volumes and densities of the elements. This results in a
difference of less than 10 % for the A-type system when compared to the actual system, which is heavier.
The object only contains structural members. The difference of 10 % can be explained by some
miscellaneous like pipework and a stairway. In Figure 3-8 the first four eigenmodes of the system according
are plotted and in Table 3-1 the corresponding eigenfrequencies are given.

Figure 3-7: FEM model A-type system

13
Figure 3-8: Mode shapes 1 to 4 A-type system

Table 3-1: Eigenfrequencies A-type system

Mode Eigenfrequency
A-type [Hz]
1 0.54
2 2.25
3 3.19
4 3.95

14
3.3.2 E-type
In Figure 3-9 the FEM model of the E-type on a pedestal is plotted. The pedestal consists out of two GEF
[9] modules. To get some confidence the mass of the model is calculated and compared to masses from
real systems. The mass is calculated using the volumes and densities of the elements. This results in a
difference of less than 15 % for the E-type system with the real system being heavier. In Figure 3-10 the
first four eigenmodes of the system according to the FEM model are plotted and in Table 3-2 the
corresponding eigenfrequencies are given.

Figure 3-9: FEM model E-type system

15
Figure 3-10: Mode shapes 1 to 4 E-type system

Table 3-2: Eigenfrequencies E-type system

Mode Eigenfrequency
E-type [Hz]
1 0.44
2 1.73
3 2.18
4 2.70

16
 Concluding remarks
A discrete-element model of both the hexapod systems are created to determine the eigenfrequencies of
the systems. The models are created using MATLAB and the toolbox StaBIL 2.0. All the elements of the
system are modeled as beams except the hydraulic actuators. The properties of elements which represent
the hydraulic actuators are calculated separately using a modelling study on stiffness characteristics of
hydraulic cylinder under multi-factors [7]. The first four eigenfrequencies of both types of system, based on
the model, are given in Table 3-3. Analyzing the eigenfrequencies and the eigenmodes of both the systems,
the first two eigenfrequencies seem to be dominated by gangway vibrations. These eigenfrequencies
correspond eigenfrequencies calculated in an earlier theoretical study done at Ampelmann about fatigue
life of a gangway [10].
Table 3-3: Eigenfrequencies Ampelmann systems

Mode Eigenfrequency Eigenfrequency

A-type [Hz] E-type [Hz]
1 0.54 0.44
2 2.25 1.73
3 3.19 2.18
4 3.95 2.70

17
4. Experimental research
To identify possible causes for the unexpected vibrations, measurements are performed on Ampelmann
systems. The goal of this chapter is performing an investigation into the unexpectedly occurring vibrations.
First experimental theory needed for this investigation is discussed, next data which is already available is
analyzed. Based on this analysis, possible causes are determined. Next some calculations and two sets of
experiments are performed investigating the possible causes.

Experimental frequency analysis

In general, frequency analysis in the field of vibration analysis is based on the discrete Fourier transform
(DFT). This is a method used to transform measured samples into a frequency spectrum. The discrete
Fourier transform is defined as [11]:
𝑁−1

𝑋(𝑘) = ∑ 𝑥(𝑛) ∗ 𝑒 −𝑗2𝜋𝑘𝑛⁄𝑁 (4.1)

𝑛=0

The inverse discrete Fourier transform (IDFT) is defined as:

𝑁−1
1
𝑥(𝑛) = ∑ 𝑋(𝑘) ∗ 𝑒 𝑗2𝜋𝑘𝑛⁄𝑁 (4.2)
𝑁
𝑛=0

When the following sampled signal is assumed: 𝑥(𝑛) = 𝑥(𝑛∆𝑡) and N samples, which is called the
blocksize, have been collected. N is usually taken to be an integer power of two (2𝑝 ) to use and optimize
the fast Fourier transform algorithm (FFT). This is an algorithm which computes the DFT in a faster way
then using equations 4.1 directly.

4.1.1 Discrete Fourier transform

In Figure 4-1 a summary of the DFT is given. This summary is based on [11]. First a continuous signal (A.1)
is transferred to the frequency domain (B.1) using the regular Fourier transform. The next step is to sample
the time signal (A.2 and A.3). This is equivalent to multiplying the continuous signal by an ideal train of
pulses with an unit value at each sampling instant and otherwise zero. In the frequency domain this
operation corresponds to the convolution with the Fourier transform, which in this case is a train of pulses
at multiples of the sampling frequency (B.2 and B.3). Measurements are only performed during a finite
period in time, which is a multiplication of the continuous signal with a rectangular window (A.4 and A.5). In
the frequency domain this is equivalent to the convolution with a sinc function: 𝑠𝑖𝑛𝑐(𝑥) = 𝑠𝑖𝑛(𝑥)⁄𝑥 (B.4 and
B.5). The result of this truncation is uncertainty in the frequency domain, which can be seen by the ripple
of the spectrum in (B.5). The final step is carried out in the frequency domain (B.6 and B.7). The DFT only
calculates the spectrum at discrete frequencies, i.e. it is a multiplication of the spectrum with a train of
pulses with a frequency increment of ∆𝑓 = 1⁄𝑇 . In the time domain (A.6 and A.7) this step is equivalent to
the convolution of a train of pulses with a separation of T.

18
4.1.2 Spectrum averaging
Many measurement signals contain random noise, either because the signal is random, or because it is
periodic or transient but contains contaminating noise. When this is the case, spectra are often averaged
frequency by frequency to reduce the random error of the spectrum estimate. The entire time signal is
divided into M segments (Figure 4-2). Of each segment the DFT is calculated. The squared magnitude
value for each frequency is averaged. In case of periodic of transient signals typically 3 – 10 segments are
necessary. Overlapping processing means using the same time sample more than once, so the final value
will contain more DFT results from the same data. This process gives a better result because the time
window used before the DFT calculation removes some information at the ends, where the window
approaches zero. The amount of overlap that should be used depends on the time window. With the
Hanning window, which is used, an overlap of 50% is usually seen as optimal [12].

Figure 4-1: Summary of discrete Fourier transform [11]

19
 Figure 4-2: Segment-based frequency domain averaging [12]

4.1.3 Forward kinematics of a Stewart platform

The steward platform mechanism is a parallel kinematic system that can be used as the basis for controlled
motion in six degrees of freedom. The mechanism consists out of a stationary platform and a mobile
platform connected via, in the case of Ampelmann, six hydraulic actuators. The desired position and
orientation of the mobile platform is achieved by combining the lengths of the six struts, transforming the
six transitional degrees of freedom into three positional and three orientational ones. The lengths of the
actuators cannot be changed independently, but only in a way that the hexapod construction allows [1].
The main difficulty with parallel manipulators is the complexity of controlling their movement. The inverse
kinematics (see paragraph 2.2.2) can be defined as finding the actuator lengths needed to position the
mobile platform in a desired position. The solution to this problem is not complex and can be computed in
a short time. The forward kinematics of a parallel manipulator is finding the position and orientation of the
mobile platform when the actuator lengths are known. This problem has no known closed form solution
[13].

The forward kinematics of a hexapod system can be formulated mathematically in several ways. Every
representation of the problem has its advantages and disadvantages which become clear when a different
optimization algorithm is applied. The forward kinematics given below are based on [13]. The most common
approach, which is also the one used by Ampelmann, uses three positional coordinates of the center of the
mobile platform (t) and three angles to define its orientation. The hexapod geometry is defined with six
vectors for the base platform (bi) and six vectors for mobile platform(pi), which define the six joint
coordinates on each platform. The actuator vector (li) is expressed in equation 4.3, where R is a rotation
matrix calculated from the rotation angles.

𝑙𝑖 = −𝑏𝑖 + 𝑡 + 𝑅 ∗ 𝑝𝑖 (4.3)

For an arbitrary solution to the forward kinematics problem the error can be expressed as the sum of the
squares of the difference between the calculated lengths and the actual values (see equation 4.4), D is the
distance between each vector pair. This results in an optimization function which relates all the unknows.
This function is non-linear due to the trigonometric function in the rotation matrix. However, the function is
derivable and is the most common representation of the forward kinematics problem.
6

𝑓 = ∑(𝐷 (𝑏𝑖 , 𝑡 + 𝑅 ∗ 𝑝𝑖 )2 − 𝑙𝑖2 )2 (4.4)

𝑖=1

20
 Pre-processing
The Ampelmann system contains a lot of sensors. The data they provide is continuously stored, for a period
of two weeks, and available through the Ampelmann system data network (ASDN). The Ampelmann system
uses two motion reference units (MRUs), where one functions as a redundancy for the other. The two
sensors are located on the bottom frame (Figure 4-3). Also, the length of the cylinders is measured over
time. The Ampelmann system is not able to compensate for all the motions resulting in some residual
motions. In the figures discussed in this paragraph for clarity only part of the measured signal, namely one
window as explained in paragraph 4.1, is shown in the time series.

4.2.1 A-type and E-type

In Figure 4-4 and Figure 4-5 the normalized translations and rotations for two A-type Ampelmann systems
are plotted in the frequency domain. Both systems are installed on identical vessels. However, the A-28 is
installed on the ship deck of the Esvagt Froude, while the A-29 is placed on a pedestal on the Esvagt
Faraday (see appendix D). In Figure 4-7 and Figure 4-8 the normalized translations and rotations for two
E-type Ampelmann systems are plotted in the frequency domain. One of the systems is placed on a
pedestal (E-04) on the Siem Barracuda while the other is attached to the ship deck (E-14) on the Olympic
Orion (see appendix D). The vessel on which the systems are located differ from each other, so the
comparison is not as good as for the A-type system. For both of the systems the motions in the time domain
are plotted in appendix C.

For all the systems, in the frequency domain a JONSWAP spectrum can be distinguished between roughly
0 Hz and 0.75 Hz as a result of the vessel motions. The energy in the frequency range between 0.75 Hz
and 6 Hz for the systems placed on a pedestal (A-29 and E-04) is higher and more distributed than for the
system placed on the deck (A-28 and E-14). Indicating motion not directly related to vessel motions. For
the system located on the deck four clear peaks are visible at 2.3 Hz, 2.6 Hz, 4.8 Hz and 5.1 Hz.

In Figure 4-6 and in Figure 4-9 the difference between the two MRUs for the two A-type systems and E-
type system in heave are plotted in the time and frequency domain. Most of this difference is located in the
lowest part of the spectrum (0 Hz – 0.03 Hz). Some clear peaks for the system attached to the ship deck
again are visible in the relative higher frequency range (1 Hz – 6 Hz) at the same frequencies as in the
heave direction.

21
MRU 1

MRU 2
Bottom frame

Figure 4-3: Location of the motion sensors

Figure 4-4: Translations A-type systems

22
 Figure 4-5: Rotations A-type systems

Figure 4-6: Difference between MRU 1 and MRU 2 in heave direction for A-type systems

23
Figure 4-7: Translations E-type systems

Figure 4-8: Rotations E-type systems

24
 Figure 4-9: Difference between MRU 1 and MRU 2 in heave direction for E-type systems

Possible causes of unexpectedly occurring vibrations

In Figure 4-10 a schematic overview of an Ampelmann system is given. Based on paragraph 4.2 three
possible causes are determined. These possible causes come down to investigating the influence of the
parts above (gangway) the hexapod, the influence of parts of the system bellow (pedestal) the hexapod
and the working limits of the Ampelmann system resulting in residual motions.

Figure 4-10: Overview Ampelmann system

25
4.3.1 Influence of the pedestal
In Figure 4-11 a schematic cross section of an Ampelmann system on a pedestal is given. When the
pedestal is excited and starts vibrating the system will try to compensate for these. However, these extra
motions are not directly related to vessel motions. In Figure 4-4, Figure 4-5, Figure 4-7 and Figure 4-8 the
six degrees of freedom of an A-type and E-type Ampelmann system, placed on a pedestal and attached to
the ship deck, are plotted in the frequency domain. In all the degrees of freedom the influence of the
pedestal is clearly visible. More energy is contained in the relative higher frequencies (0.75 Hz – 6 Hz).
These motions are measured by the MRUs which results in the system trying to compensate for these
motions. This might result in a vicious circle resulting in unintended vibrations in the systems. The following
will be investigated:

Compensating for motions caused by vibrations in the pedestal can cause the unexpected vibrations in an
Ampelmann system.

Figure 4-11: Cross section of Ampelmann system; vibrations caused by a pedestal

26
4.3.2 Residual motions
Due to the limitations of the Ampelmann system some residual motions exist (Figure 4-12), which means
the transfer deck and gangway are not fully compensated. For example: the cylinders of the Ampelmann
systems have a limited range. The cylinders must be prevented from running in to their buffers. Because
of this the cylinders are never controlled to more than 80 % of the maximum rake. Secondly when the
cylinders are over 50 % of their rake, the system starts to partially compensate for the measured motions.
Also, the frequency filter in the motion control algorithm (Figure 2-3) can cause lag. These cause residual
motions. When peaks in the frequency domain of these motions correlate to the eigenfrequencies of the
transfer deck and gangway, vibrations can occur. The following will be investigated:

The frequency content of the residual motions corresponds to the eigenfrequencies of the gangway which
can cause the unexpected vibrations in an Ampelmann system.

Figure 4-12: Cross section of Ampelmann system; residual motions

27
4.3.3 Vibrations in the bottom frame due to compensating for gangway motions
In Figure 4-13 a schematic cross section of an Ampelmann system is given. The pushing and pulling of the
hydraulic cylinders against their gimbals results in a response in the bottom frame. As a result, the bottom
frame will deform resulting in motions in the MRUs. In Figure 4-6 and Figure 4-9 the difference between the
two MRUs in heave direction is plotted. In these figures the fluctuations in the time domain might indicate
vibrations in the bottom frame. When the Ampelmann system starts compensating for motions with the
same frequency as the eigenfrequency of the bottom frame, large motions might arise. These motions are
measured by the MRUs which leads to the system trying to compensate for these motions. This might result
in a vicious circle introducing unintended vibrations in the systems. The following will be investigated:

Vibrations in the bottom frame due to compensating for gangway motions can cause the unexpected
vibrations in an Ampelmann system.

Figure 4-13: Cross section of Ampelmann system; vibrations in the bottom frame

28
 Measurements on the Ampelmann system
An experiment can be classified based on the input or loading using the following: Is the input dynamic or
static? Is the input controllable and whether the inputs are measurable. An experimental procedure can be
roughly divided into a number of steps [14]. These steps are:

1. Selection of inputs, their locations and their means of measurement.

2. Selection of outputs, their locations and their means of measurement.
3. Gathering and transmission of signals to a recorder/logger.
4. Conversion of the signal to a storable form.
5. Data storage

The sensors which will be used for measuring during the different tests are all iXblue Octans (look for
technical specifications in appendix E). This is the same MRU which is used on the Ampelmann systems.
The Octans is a gyrocompass and motion sensor measuring displacements and accelerations in six
degrees of freedom. The Octans is a fiber optic gyroscope. In these types of gyroscope, pulses of light are
sent through windings of fiberglass, both clockwise and counterclockwise. When the sensor rotates, a tiny
difference can be measured in the arrival time of the light pulses, when comparing the clockwise pulses
with the counterclockwise pulses.

All signals measured on mechanical systems are defined continuous in time. When they are recorded, they
are measured in time discrete signals with a sampling frequency. The sampling frequency of the sensor is
set at 50 Hz. According to the Nyquist frequency, the sampled signal, which is continue, can only be
uniquely represented by discrete samples if it is sampled using a frequency larger than twice the highest
frequency in the analogue signal. This means that the lowest frequency to be investigated is 25 Hz. This is
well above the frequencies of the vibrations which are the subject of investigation.

4.4.1 Influence of the pedestal

To investigate the possible influence of the pedestal on the bottom frame the following test is performed: A
ship to ship MRU (Figure 4-15) are placed on the ship deck. The data from the ship to ship MRU and the
MRUs which are already on the system are compared. From this data the influence of the pedestal can be
determined. The measurements are taken from an E-type (E-04) system. This system is placed on the Siem
Baracuda (appendix D). The measurements are performed during regular operations of the system.

During this experiment the input will be the vessel motions at the base of the Ampelmann system or the
pedestal. These motions are uncontrollable but measurable. The output will be the system responses. The
measurements are done using the MRUs of the Ampelmann system and a ship to ship MRU. The gathering
of the data is done using CPU of the Ampelmann system. This makes sure all the signals are synchronized
time wise and will automatically convert the data to usable format and is stored.

29
 Figure 4-14: MRU locations

Figure 4-15: Ship to ship MRU

30
Results
In Figure 4-17 the translations and in Figure 4-18 the rotations in the time domain of the two MRUs are
plotted. The rotations for both the MRUs are roughly the same (for the difference see appendix F). The
trend and the amplitude are reasonably the same. However, this is not the case for the translations. This is
because the motions measured by the MRUs are translated to the center of gravity of the Ampelmann
system. This is done using leaver arm as indicated in Figure 4-16 by dX and dZ. The leaver arm of the
MRU, which is fitted on the system, is precisely known. Determining the exact leaver arm for the ship to
ship Octans proofed to be difficult. Because the leaver arm for the ship to ship MRU are large, the
translations for the most part are determined by the rotations of the MRU. When the leaver arm are not
precisely correct, the translations will contain an error.

In Figure 4-19 translations and in Figure 4-20 the rotations in the frequency domain are plotted. The energy
of the motions measured by the regular MRU, which includes the pedestal, contains more energy in the
relative higher frequencies when compared to the ship to ship MRU. This is the same as in paragraph 4.2.
The peaks in the frequency graphs are roughly located at: 2,8 Hz, 5,4 Hz and 8,1 Hz for all the degrees of
freedom. This is because the data measured by both the MRUs is measured with respect to a fixed
coordinate system. As a consequence, the motions of the pedestal have an influence in multiple directions
in the MRU’s local coordinate system. Another consequence of this dependency is that Figure 4-19 and
Figure 4-20 are similar.

Four peaks can be identified in both the signals. The first is directly caused by vessel motions so a
JONSWAP spectrum. The other three have a cause which is not directly related to vessel motions. The
second and third peak have roughly the same amplitude for both the data sets. However, the fourth peak
of the system MRU Octans 1 has a much larger amplitude compared to the amplitude of the ship to ship
Octans. Around the frequency of this peak, 8,1 Hz, more energy is present at the top of the pedestal than
at the bottom. For this frequency the pedestal seems to amplify the energy present at the ship deck.

Figure 4-16: Leaver arms (dX & dZ)

31
Figure 4-17: E-04 Translations (time domain)

Figure 4-18: E-04 Rotations (time domain)

32
Figure 4-19: E-04 translations (frequency domain)

Figure 4-20: E-04 Rotations (frequency domain)

33
Possible cause
As in the previous paragraph is stated, the pedestal seems to amplify energy present at the ship deck. A
possible explanation for this phenomenon could be found in the eigenfrequencies and corresponding
eigenmodes of the ship deck. In Figure 4-21 a schematic drawing showing this characteristic is given. The
ship deck bends according to a certain mode shape. The pedestal functions as an arm increasing the
amplitude of the motion of the Ampelmann system on top. The system will try to compensate for these
motions. The compensation for this motion will occur with the same frequency as the eigenfrequency of the
ship deck adding more energy at this frequency to the system. This might result in a vicious circle resulting
in unintended vibrations in the systems.

The different stiffeners in a ship deck of a crane vessel close to a heavy-duty crane are in general much
heavier than the stiffeners below the main deck to cope with the forces introduced by the crane. In general,
the Ampelmann system is placed on the main deck of a vessel not specifically engineered to cope with the
system. An Ampelmann system introduced roughly the same dynamic forces as a crane. This supports the
idea that vibrations in the ship deck might cause the unexpected vibrations.

Figure 4-21: Schematic overview ship deck vibrations

34
4.4.2 Residual motions
To investigate the possible influence of the residual motions and if they can cause the vibrations, these
residual motions are calculated. For these calculations the forward kinematics as explained in paragraph
4.1.3 are used. The forward kinematics require as input the motions at the base of the hexapod, which is
the Octans data, and the corresponding cylinder lengths (Figure 4-22). For this data again the Ampelmann
system data network (ASDN) is used. To make sure sufficient residual motions are present during the
sampled time, data is used while the vessel containing the Ampelmann system was in the most severe sea
state available.

Figure 4-22: Cylinder lengths and reference motions

Results
In Figure 4-23 the residual motions in the time and frequency domain are given. In Table 4-1 the first five
theoretical eigenfrequencies of a typical Ampelmann gangway are given based on a fatigue analysis done
at Ampelmann [10]. The frequencies given in Table 4-1 coincide with the inverted peaks in the frequency
graph. Meaning the motion control algorithm and Ampelmann system is good at compensating for vibrations
with the eigenfrequency of the gangway. In the graph no amplification of energy can be distinguished.
Because of this it can be concluded that the residual motions do not cause the unexpected vibrations.

35
Figure 4-23: Residual motions (time and frequency domain)

Table 4-1: Theoretical eigenfrequencies gangway [10]

MODE NUMBER EIGENFREQUENCY

1st 0,45

2nd 1,28

3rd 1,64

4th 2,19

5th 3,02

36
4.4.3 Vibrations in the bottom frame
To investigate the possible influence of the gangway on the bottom frame/MRUs the following test is
performed. The gangway on an A-type Ampelmann system is given an excitation using the hydraulic
cylinders normally used for luffing. This resembles the effect of hitting a beam with a hammer. The gangway
will start vibrating in a decaying manner with its damped eigenfrequency. The hydraulic actuators will “feel”
these motions and transfer them into the bottom frame. For the experiment the A-20 will be used. During
the tests the system in placed unbolted on the factory floor.

During this experiment the input will be a blow with the gangway resulting in a decaying vibration of the
gangway. This blow is uncontrollable and unmeasurable. The output will be the system responses. The
responses in the bottom frame are measured using the MRUs of the Ampelmann system which are located
on the bottom frame. The gathering of the data will be done using CPU’s of the Ampelmann system. This
makes sure all the signals are synchronized time wise and will automatically convert the data to usable
format and is stored.

An overview of the experiment is given in Figure 4-24. During the experiment multiple blows are given with
the gangway using a different amount of extension of the gangway and different orientations of the
gangway. The lengths of the gangway are gangway completely retracted, gangway halfway extended and
fully extended. The first orientation is with the gangway perpendicular to the two MRUs. The second
orientation is while the gangway is in line with the two MRUs. For the second orientation due to limited
space the test can only be performed with a retracted gangway. This sequence will be performed twice.
The first time while the system is in the neutral position, meaning the motion control algorithm is not
activated. The second time will be performed while system is actively compensating and the motion control
algorithm is activated.

Figure 4-24: Overview experiment vibrations in the bottom frame

37
Results
In Figure 4-25 the translations and rotations of both the MRUs for the entire sampled time is plotted. In
Figure 4-26 the translations and rotations of both the MRUs during a single blow of the gangway is plotted.
From the plots the conclusion can be drawn that the gangway is not able to cause vibrations in the bottom
frame. For all six degrees of freedom the amplitude of the motions is negligible over the entire test period.
Since no responses haven been detected in the bottom frame during this test the conclusion is drawn that
what happens on the top of the hexapod cannot cause vibrations in the bottom frame.

Figure 4-25: A-20 Translations and rotations (entire sampled time)

Figure 4-26: A-20 Translations and rotations (sampled time series during single blow with gangway)

38
 Concluding remarks
To identify possible causes for the unexpected vibrations, measurements are performed on Ampelmann
systems. Data which was already available at Ampelmann is analyzed and based on this data three
possible causes are determined. These are:

• the influence of the pedestal,

• residual motions due to limitations of the Ampelmann system
• vibrations in the bottom frame due to compensating for gangway motions.

To investigate the influence of the pedestal and the vibrations in the bottom frame two experiments are
performed. The residual motions are investigated via calculations based on data readily available.

The result of the experiment, which is performed to investigate the vibrations in the bottom frame due to
gangway motions, show that for all six degrees of freedom the amplitude of the response in the bottom
frame is negligible over the entire test period. Using these results the conclusion is drawn that the gangway
is not able to cause the unexpectedly occurring vibrations. The results from the calculations done to
investigate the effect of the residual motions show no amplification concluding that the residual motions do
not cause the unwanted behavior.

The experiment to determine the influence of the pedestal shows four peaks in the frequency domain of
both the signals. The first is directly caused by vessel motions. The other three have a cause which is not
directly related to vessel motions. The data from the sensor on the Ampelmann system, at the top of the
pedestal, does not contain a peak which is not present at the ship deck. From this it can be concluded that
the eigenfrequencies of the pedestal do not have an influence. For one of these peaks in both of the data
sets, namely the fourth peak at 8,1 Hz more energy at the top of the pedestal is present than at the ship
deck. At this frequency the pedestal seems to amplify the energy present at the ship deck. A possible
explanation for this phenomenon could be the eigenfrequencies and corresponding eigenmodes of the ship
deck. The pedestal functions as a leaver arm amplifying the rotations related to eigenmodes of the ship
deck. This may cause the unexpected vibrations.

39
5. Analytical model
Ampelmann’s motion control algorithm is complicated which which makes it difficult to incorporate this
algorithm in the FEM model. Because of this there is a need for a second model. This model is a simplified
2D multi-body model. The degrees of freedom will be based on the direction of the most dominant
eigenfrequency and corresponding eigenmode determined using the FEM model. Using the simplified
multibody model, the equations of motion will be derived where the motion control algorithm will be included.
In Figure 5-1 an overview of the desired result is given. The system is represented by three masses: The
pedestal + Bottom frame, Top frame and the gangway. The pedestal is assumed to be rigidly connected to
the ship deck. At Ampelmann an analytical model (Figure 5-2), made by Temporary Work Design (TWD),
is available which will be used as a start.

Figure 5-1: Overview of desired result

Model by Temporary Work Design

Temporary work designs (TWD) has produced a model for Ampelmann (Figure 5-2) [15]. This model is a
dual lumped mass model consisting out of two masses which represent the compensated part of the mass
and the uncompensated part of the mass of the system. The equations of motion of this model have been
determined by TWD and are given in equation 5.1.

𝐼1 𝜃̈1 + 𝐵𝜃 𝜃̇1 + 𝐾𝜃 𝜃1 = −𝐹1 𝑑ℎ + 𝑇1

𝐼2 𝜃̈2 = −𝑇1 (5.1)
𝑀2 𝑦̈ 2 = −𝐹1
F1 and T1 represent the actuator forces and the reaction forces between the two masses. However, because
the controller of Ampelmann is based on velocity the controller cannot be added in this way. To overcome
this a cylinder model is introduced (Figure 5-3). A velocity input is given to the cylinder, which results in
velocity input 𝐿̇𝜃 & 𝐿̇𝑦 , in other words force is created by moving one end of the spring. The stiffness of the
cylinders is assumed to be infinitely high, which means the position of the cart is equal to the position of the
compensated mass. The new equations of motion are given in equation 5.2 [15].

40
 𝐼1 𝜃̈1 + 𝐵𝜃 𝜃̇1 + 𝐾𝜃 𝜃1 = −𝐹1 𝑑ℎ + 𝑇1
𝜃̇2 = 𝐿̇𝜃 + 𝜃̇1
𝑦̇ 2 = 𝐿̇𝑦 − 𝜃̇1 𝑑ℎ (5.2)
𝐼2 𝜃̈2 = −𝑇1
𝑀2 𝑦̈ 2 = −𝐹1

Figure 5-2: TWD model [15]

Figure 5-3: Introduction of cylinder model [15]

41
In Figure 5-4 an overview of the control loop is given. The plant represents the transfer function from cylinder
velocity to MRU position. AC represents the actuator dynamics, which in practice means a delay. For more
details about the control loop reference is made to [15] & [16]. Using this control loop and the analytical
model a state space model is created.

Figure 5-4: Control loop in analytical model

Differentiation error
As per definition the transfer function is the output divided by the input: 𝐺(𝑠) = 𝑦(𝑠)⁄𝑢(𝑠) [17]. In equation
5.3 to 5.6 the transfer functions are worked out.

Laplace transform of equations of motions.

(𝐼1 𝑠 2 + 𝛽𝜃 𝑠 + 𝐾𝜃 ) θ1 (𝑠) = −𝑑ℎ 𝐹1 (𝑠) + 𝑇1 (𝑠)

𝑠 𝜃2 (𝑠) = 𝑠 𝐿𝜃 (𝑠) + 𝑠 𝜃1 (𝑠)
𝑠 𝑌2 (𝑠) = 𝑠 𝐿𝑦 (𝑠) − 𝑑ℎ 𝑠 𝜃1 (𝑠) (5.3)
𝐼2 𝑠 2 𝜃2 (𝑠) = −𝑇1 (𝑠)
𝑀2 𝑠 2 𝑌2 (𝑠) = −𝐹1 (𝑠)
Differentiating θ2 and Y2 and rewriting.

(𝐼1 𝑠 2 + 𝛽𝜃 𝑠 + 𝐾𝜃 ) θ1 (𝑠) = −𝑑ℎ 𝐹1 (𝑠) + 𝑇1 (𝑠)

𝐼2 𝑠 2 (𝐿𝜃 (𝑠) + 𝜃1 (𝑠)) = −𝑇1 (𝑠) (5.4)
2
𝑀2 𝑠 (𝐿𝑦 (𝑠) − 𝑑ℎ 𝜃1 (𝑠)) = −𝐹1 (𝑠)

Rewriting.
(𝐼1 𝑠 2 + 𝛽𝜃 𝑠 + 𝐾𝜃 ) θ1 (𝑠) = 𝑑ℎ 𝑀2 𝑠 2 (𝐿𝑦 (𝑠) − 𝑑ℎ 𝜃1 (𝑠)) − 𝐼2 𝑠 2 (𝐿𝜃 (𝑠) + 𝜃1 (𝑠))
(5.5)
((𝐼1 + 𝐼2 + 𝑑ℎ2 𝑀2 ) 𝑠 2 + 𝛽𝜃 𝑠 + 𝐾𝜃 ) θ1 (𝑠) = 𝑑ℎ 𝑀2 𝑠 2 𝐿𝑦 (𝑠) − 𝐼2 𝑠 2 𝐿𝜃 (𝑠)

42
 𝜃1 −𝐼2 𝑠 2
𝑃𝜃 = =
𝐿̇𝜃 (𝐼1 + 𝐼2 + 𝑑ℎ2 𝑀2 ) 𝑠 2 + 𝛽𝜃 𝑠 + 𝐾𝜃
(5.6)
𝜃1 𝑑ℎ 𝑀2 𝑠 2
𝑃𝑦 = =
𝐿̇𝑦 (𝐼1 + 𝐼2 + 𝑑ℎ2 𝑀2 ) 𝑠 2 + 𝛽𝜃 𝑠 + 𝐾𝜃
In the model the second step, the differentiation of the θ2 and Y2, has not been done. In practice this means
that instead of the acceleration, the velocity is entered. This results in the transfer functions given in
equations 5.7. When the correct transfer functions are used, the real value for one pole and a zero becomes
positive (Figure 5-5), indicating an unstable control loop. In appendix F the response of the system is given.
This response goes to infinity, as is to be expected with an unstable control loop.

𝜃1 −𝐼2 𝑠
𝑃𝜃 = =
𝐿̇𝜃 (𝐼1 + 𝐼2 + 𝑑ℎ 𝑀2 ) 𝑠 2 + 𝛽𝜃 𝑠 + 𝐾𝜃
2
(5.7)
𝜃1 𝑑ℎ 𝑀2 𝑠
𝑃𝑦 = =
̇𝐿𝑦 (𝐼1 + 𝐼2 + 𝑑ℎ2 𝑀2 ) 𝑠 2 + 𝛽𝜃 𝑠 + 𝐾𝜃

Figure 5-5: Pole-Zero map of closed loop, Left with error & right without error

Pole plot investigation

In Figure 5-6 the pole plot of the plant and the controller are given. In Figure 5-7 the pole plots of the
lowpass feed forward are given. The plots of the plant, controller and the lowpass feed forward all are stable
since there are no poles with a positive real value. However, the pole plot of the open loop system does
have poles with positve real values and thus is unstable.

43
 Figure 5-6: Pole zero maps; Left: plant, Right: controller

Figure 5-7: Pole zero maps; Left: lowpass filter feed forward, Right: open loop

Concluding remarks
Ampelmann’s motion control algorithm is complicated which makes it difficult to incorporate this algorithm
in the FEM model. An attempt is made to create a simplified 2D multi-body model including the algorithm.
TWD has produced a model, consisting out of two masses, which should have funtioned as the base for
this model. During the investigation of the model an error was discovered. The correction of this error
resulted in an unstable system. A pole plot investigation is done. However, no explanation for the instablity
is found.

44
6. Conclusion and recommendations
Conclusion
The Ampelmann system occasionally starts vibrating unexpectedly, especially while the system is placed
on a pedestal. These vibrations are believed to be caused by the eigenfrequencies of the system and/or
amplifications caused by the motion control algorithm. In this thesis an investigation has been done into
this phenomenon. A discrete-element model of the Ampelmann system was created to calculate the
eigenfrequencies of the system. This model has been created using MATLAB and the toolbox StaBIL 2.0.
The first four eigenfrequencies, calculated using the model, are given in Table 6-1.
Table 6-1: Eigenfrequencies A-type and E-type systems

Mode Eigenfrequency Eigenfrequency

A-type [Hz] E-type [Hz]
1 0.54 0.44
2 2.25 1.73
3 3.19 2.18
4 3.95 2.70

To identify possible causes for the unexpected vibrations data, which was readily available at Ampelmann
through the Ampelmann system data network, has been analyzed. Using this, three possible causes have
been determined. These causes are: the influence of the pedestal, residual motions due to limitations of
the Ampelmann system and vibrations in the bottom frame due to compensating for gangway motions. To
investigate these three cases, two sets of experiments and a calculation based on already available data
have been done (see paragraph 4.4).

The experiment, which has been performed to investigate the vibrations in the bottom frame due to
compensating for gangway motions, shows that the response in the bottom frame is negligible. Using this
result, it was concluded that the gangway is not able to cause the unexpectedly occurring vibrations. The
results from the calculations done to investigate the effect of the residual motions showed no amplification
of energy. From this it was concluded that the residual motions do not cause the unwanted behavior.

The experiment done to investigate the influence of the pedestal shows that amplification of energy at a
certain frequency, namely 8,1 Hz, is present. The pedestal could be the cause of the problem. However,
instead of “internal” vibrations due to eigenfrequencies (as explained in paragraph 4.3.1), the amplification
seems to have a different origin. The pedestal functions as a leaver arm amplifying the rotations related to
eigenfrequencies and corresponding eigenmodes of the ship deck. This may cause the unexpected
vibrations.

An attempt has been made to create a simplified multibody model incorporating the motion control
algorithm. This model was supposed to be an extension to a model made by technical work design available
at Ampelmann. However, an error in the model has been discovered. The correction of this error caused
the model to become unstable. Instead of expanding the model, a pole plot investigation for all the sub
parts of the model has been done. However, no explanation for the instablity was found.

The objective of this thesis was creating a dynamic model to calculate the eigenfrequencies of the
Ampelmann system and explain the unexpectedly occurring vibrations in the Ampelmann system. The
eigenfrequencies have been determined using the created MATLAB model. Three possible causes for the
unexpectedly occurring vibrations have been investigated. One of these three, the stiffness of the ship deck
(paragraph 4.4), was found to be a plausible cause.

45
 Recommendations
The general consensus at Ampelmann was that the unexpectedly occurring vibrations are caused by
eigenfrequencies and corresponding eigenmodes of the pedestal. In this research an indication was found
that the origin of the problem is not the pedestal itself, but instead the pedestal functions as an amplifier.
Further research is necessary. For example, a desk study could be performed where ship decks, on which
the phenomenon occurred, are compared to ship decks on which nothing happened. This research might
result in a correlation between deck stiffness and the chance of having the unexpectedly occurring
vibrations.

In this research an attempt was made to expand a simplified multibody model of the Ampelmann system.
However, the correction of a discovered error in the model resulted in the system having positive real poles,
meaning there is an instability in the system. Instead of expanding the model a pole plot investigation is
done trying to discover the cause of this real positive pole. The cause of this pole has not been found, so
further research is necessary. The model is now based on a linear system of equations. It is reccomanded
to investage wheater these are appliclable for the Ampelmann sytem. The added value of building a
relatively simple computer model, in which the motion control algorithm is incorporated, could be quite
significant. What could be investigated within this model is:

• How does Ampelmann’s algorithm handle vibrations introduced to the base of the system or to the
base of the pedestal?
• What is the effect of the height of the pedestal?
• Are there other situations in which the controller becomes unstable?

At the moment Ampelmann is designing the generic A frame, which is a modular pedestal for A-type
systems. During the design a lot of research has gone into defining the desired eigenfrequencies of this
frame hoping to prevent the unexpectedly occurring vibrations. Based on this thesis, the eigenfrequencies
of the pedestal do not play a significant role in preventing these. Instead, the focus should be on the
connection between the vessel and the Ampelmann system.

Ampelmann is designing a new gangway system capable of lifting up to 5 tons, where previous systems
where capable of lifting up to 1 ton. The different stiffeners in a ship deck of a crane vessel close to a heavy-
duty crane are in general much heavier than the general stiffeners below the main deck. These heavier
stiffeners are implemented to handle the forces introduced by the crane. To cope with these higher loads
of the larger Ampelmann systems, it is recommended that extra reinforcements are added to vessels on
which these systems will be installed. It should be noted that these reinforcements are not be limited to the
main deck but should reach deeper into the ship spanning multiple decks.

In this research in almost all the graphs in the frequency domain peaks are found at 2.3 Hz and 2.6 Hz. The
cause of this peaks could not be identified. What could cause these peaks is for example: the internal
workings of the sensor or the cabinet of the sensor. Further research into the cause of this peaks is
recommended.

46
7. References

[1] D. Stewart, "A Platform with Six Degrees of Freedom," Proceedings of the Institution of Mechanical
Engineers, pp. 371 - 386, 1965.

[2] K. J. Aström and R. M. Murray, Feedback systems, New Jersey: Princeton University Press, 2012.

[3] A. v. Leer, "Motion Control Enhancements for the Ampelmann System," Master's Thesis, TU Delft,
2012.

[4] D. Dooms, M. Jansen, G. d. Roeck, G. Degrande, G. Lombeart, M. Schevenels and S. Francois,

StaBIL: A Finite Element Toolbox For MATLAB, Leuven: Katholieke Universiteit Leuven, 2010.

[5] J. S. Przemieniecki, "Stiffness Properties of Structural Elements," in Theory of Matrix Structural

Analysis, McGraw-Hill, 1968, pp. 61 - 128.

[6] T. Yokoyama, "Vibrations of a Hanging Timoshenko Beam Under Gravity," Journal of Sound and
Vibrations , vol. 141, no. 2, pp. 245-258, 1990.

[7] H. Feng, Q. Du, Y. Huang and Y. Chi, "Modelling Study on Stiffness Characteristics of Hydraulic
Cylinder under Multi-Factors," Journal of Mechanical Engineers, vol. 7, no. 8, pp. 447 - 456, 2017.

[8] A. Bureček, L. Hružík and M. Vašina , "Determination of Undissolved Air Content in Oil by Means of
a Compression Method," Journal of Mechanical Engineering, vol. 61, no. 7-8, pp. 477-485, 2015.

[9] Ampelmann, "P18.030.03 AM - E-TYPE GEF XL," Ampelmann, Delft, 2019.

[10] Y. C. Chung, "Dynamic Analysis of the Ampelmann G25 Gangway," TU Delft, Delft, 2016.

[11] A. Brandt, "Experimental Frequency Analysis," in Noise and Vibration analysis: Signal Analysis and
Experimental Procedures, Wiley, 2013, pp. 207-250.

[12] A. Brandt, "Spectrum and Correlation Estimates Using the DFT," in Noise and Vibration Analysis:
Signal Analysis and Experimental Procedures, Wiley, 2011, pp. 251-295.

[13] D. Jakobovic, "Forward Kinematics of a Steward Platform Mechanism," Faculty of Electrical

Engineering and Computing; University of Zagreb, Zagreb, 2002.

[14] F. N. Catbas, T. Kijewski-Correa and A. E. Aktan, Structural Identification of Constructed Systems

Approaches, Methods and Technologies for Effective Practice of Structural Identification, Reston:
American Society of Civil Engineers, 2013.

[15] Temporary Work Design, Frequency analysis, Delft: TWD, 2019.

[16] Temporary Work Design, "TWD NL 2019 030C 100 REV 0 Dynamic model of A-type Ampelmann with
pedestal-TF," TWD, Delft, 2019.

[17] S. Skogestad and I. Postlethwaite, "Multivariable Feedback Control Analysis and Design," Wiley,
2007.

47
[18] J. N. Reddy, Introduction to Nonlinear Finite Element Analysis, New York: Oxford University Press,
2004.

[19] B. Friedland, CONTROL SYSTEM DESIGN An Introduction to State-Space Method, New York: Dover
Publications, 2005.

48
8. Appendix
A. Standard block diagram elements

Figure 8-1: Standard block diagram elements [2]

49
B. Beam stiffness matrix
𝐸𝐴
𝑙
12𝐸𝐼𝑧
0
𝑙 3 (1 + 𝛷𝑦 )
12𝐸𝐼𝑦
0 0
𝑙 3 (1 + 𝛷𝑧 )
𝐺𝐽
0 0 0 𝑠𝑦𝑚𝑚𝑒𝑡𝑟𝑖𝑐
𝑙
−6𝐸𝐼𝑦 (4 + 𝛷𝑧 )𝐸𝐼𝑦
0 0 0
𝑙 2 (1 + 𝛷𝑦 ) 𝑙(1 + 𝛷𝑧 )
6𝐸𝐼𝑧 (4 + 𝛷𝑦 )𝐸𝐼𝑧
0 0 0 0
𝑙 2 (1 + 𝛷𝑦 ) 𝑙(1 + 𝛷𝑦 )
−𝐸𝐴 𝐸𝐴
0 0 0 0 0
𝑙 𝑙
𝐾= −12𝐸𝐼𝑧 −6𝐸𝐼𝑧 12𝐸𝐼𝑧
0 0 0 0 0
𝑙 3 (1 + 𝛷𝑦 ) 𝑙 2 (1 + 𝛷𝑦 ) 𝑙 3 (1 + 𝛷𝑦 )
−12𝐸𝐼𝑦 6𝐸𝐼𝑦 12𝐸𝐼𝑦
0 0 0 0 0 0
𝑙 3 (1 + 𝛷𝑧 ) 𝑙 2 (1 + 𝛷𝑦 ) 𝑙 3 (1 + 𝛷𝑧 )
−𝐺𝐽 𝐺𝐽
0 0 0 0 0 0 0 0
𝐴 𝑙
−6𝐸𝐼𝑦 (2 − 𝛷𝑧 )𝐸𝐼𝑦 6𝐸𝐼𝑦 (4 + 𝛷𝑧 )𝐸𝐼𝑦
0 0 0 0 0 0 0
𝑙 2 (1 + 𝛷𝑦 ) (1 + 𝛷𝑧 ) 𝑙 2 (1 + 𝛷𝑦 ) 𝑙(1 + 𝛷𝑧 )
6𝐸𝐼𝑧 (2 − 𝛷𝑦 )𝐸𝐼𝑧 −6𝐸𝐼𝑧 (4 + 𝛷𝑦 )𝐸𝐼𝑧
0 0 0 0 0 0 0 0
[ 𝑙 2 (1 + 𝛷𝑦 ) (1 + 𝛷𝑦 ) 𝑙 2 (1 + 𝛷𝑦 ) 𝑙(1 + 𝛷𝑦 ) ]

12𝐸𝐼𝑧
𝛷𝑦 =
𝐺𝐴𝑠𝑦 𝑙 2
12𝐸𝐼𝑦
𝛷𝑧 =
𝐺𝐴𝑠𝑧 𝑙 2

50
C. Time domain A-28, A-29, E-04 & E-14

Figure 8-2: Time domain A-28 system

Figure 8-3: Time domain A-29 system

51
Figure 8-4: Time domain E-04 system

Figure 8-5: Time domain E-14 system

52
D. Vessels on which the different systems are installed

A-28

A-29

Figure 8-6: A-28 without pedestal (upper), A-29 with pedestal (lower)

E-04

Figure 8-7: E-04 on pedestal (upper), Olympic Orion, no picture with Ampelmann system available (lower)

53
E. MRU (Octans) specifications

Figure 8-8: MRU (Octans) specifications

54
F. Octans & ship to ship data E-04 (difference)

Figure 8-9: E-04 Translations difference (time domain)

Figure 8-10: E-04 Rotations difference (time domain)

55
Figure 8-11: E-04 translations difference (frequency domain)

Figure 8-12: E-04 Rotations difference (frequency domain)

56
G. Excitation analytical model

Figure 8-13: Response analytical model

57
58