Ocean Engineering 137 (2017) 99109

Contents lists available at ScienceDirect

Ocean Engineering
journal homepage: www.elsevier.com/locate/oceaneng

Propulsion control strategies for ship emergency manoeuvres MARK

M. Altosole , M. Martelli
Polytechnic School of Genoa University, DITEN, Via allOpera Pia 11 A, 16145 Genova, Italy

A R T I C L E I N F O A BS T RAC T

Keywords: This work deals with the propulsion control aspects relating to some of the most critical emergency manoeuvres
Ship propulsion control of a ship: slam start and crash stop. In these particular situations a very important role is played by the
Simulation based design automation system that has to manage the whole propulsive chain (i.e. main engine, mechanical transmission
Slam start and propeller) in a safe and ecient way. With regard to this, a simulation based design methodology is adopted
Crash stop
to develop and test new control schemes for ship propulsion. The proposed control layout is applicable to any
Engine
Propeller pitch
type of propulsion systems equipped with controllable pitch propellers, since it is mainly based on the automatic
adjustment of the propeller pitch. Thus the desired performance requirements are met through adaptive control
strategies able to address the complex issues of slam start, crash stop and similar stressful manoeuvres. The
adaptivity of the automation process to several critical propulsive conditions reduces signicantly the number of
the control parameters to be estimated, as recently demonstrated by the automation design of a new twin-screw
ship. For this application, the comparison between simulation results and sea trials data is nally shown for
validation design purposes.

1. Introduction
Current marine propulsion systems are notable for their high
performance and exibility, dicult features to be achieved without
the development of an ecient control system. The large power
available to the propellers entails a careful management of the
propulsion machinery, in every propulsive condition. Especially during
critical manoeuvres (e.g. slam start, crash stop, tight turning circles,
), the automation designers have to develop proper control strategies
and set a lot of parameters, normally based on their experience, in
several ship propulsion modes. The identication of the proper solution
among all the possible combinations is rather dicult and time
consuming. In light of this, special control functions, characterized
by an adaptive behaviour, could be very useful because they would
reduce the number of variables to be assessed.
Time-domain simulation can be used to evaluate the eectiveness
of these algorithms, especially for the representation of ship critical
manoeuvres that could dangerously stress the whole propulsion system
(Altosole et al., 2012). Many simulation studies on ship manoeuvr-
ability can be found in the scientic literature but it is not easy to
represent every manoeuvre by a single numerical model. An attempt to
develop a comprehensive unied versatile mathematical model, suita-
ble for the all types of manoeuvres in still water, has been recently
undertaken by Sutulo and Guedes Soares (2015): in their study, the
comparisons with full-scale data (available for the full helm turning
tests, crash stop tests and bow thruster turning of a shuttle LNG
carrier) show reliable simulation results, practically without special
tuning of the model.
On the contrary to what happens for manoeuvrability issues, the
machinery behaviour during demanding manoeuvres is rather less
investigated in the available literature. The most stressful conditions
for the prime mover, propeller and mechanical transmission can be
experienced during the ship emergency manoeuvring, as in the case of
slam start and crash stop. As well known, the rst one is traditionally
referred to the vessel acceleration from zero speed to the full power
condition, while the second one is the ship stopping, as quickly as
possible, starting from the maximum speed of the vessel. Especially
during crash stop, usually performed to avoid collisions, the main
engine and propeller are subjected to severe stress and loading. From
this point of view, relevant scientic papers exist on the propeller
structural safety in this particular emergency situation. In the available
literature, the prediction of loads on the propeller during crash stop is
usually performed by using propeller series charts or, more specically,
CFD analyses, in order to evaluate the structural safety of the blades. In
Hur et al. (2011), propeller torque values are estimated by Wageningen
B series characteristics and compared with sea trial data, in spite of the
dierence of the propeller blade shape and ow steadiness between
series and recordings. Wageningen results during crash stop provide
slightly higher torque values but the discrepancies can be considered
marginal in evaluation of the structural safety. However, series charts

Corresponding author.
E-mail address: marco.altosole@unige.it (M. Altosole).

http://dx.doi.org/10.1016/j.oceaneng.2017.03.053
Received 26 November 2015; Received in revised form 15 February 2017; Accepted 25 March 2017
0029-8018/ 2017 Published by Elsevier Ltd.
M. Altosole, M. Martelli
Nomenclature
A yoke area
B oil bulk modulus
Cip oil leakage coecient
C Coriolis centripetal matrix
e error
n rotational speed
n rotational speed corresponding to the maximum lever
position
nc commanded rotational speed
poil oil pressure
t time
KD derivative gain
KI integral gain
KP proportional gain
Ib moment of inertia of the propeller blade
Ip polar inertia
M ship inertia matrix
qi oil volumetric ow
QE engine torque
cannot predict the pressure distribution on the propeller blade; hence
CFD simulation is suggested to represent the hydrodynamic load acting
on the propeller. About this, Black and Swithenbank (2009) have
examined the water ow velocities, measured during a crash-back test
of a model propeller: experimental data, concerning mean and extreme
loading conditions, have been used in a strip theory approach to
develop loads for nite element analysis, in order to assess stresses on
the individual blades.
Structural safety of the several propulsion components (propeller,
shaft line, thrust-bearing and main engine) can be ensured also by
adopting proper control strategies to avoid overloading during critical
manoeuvres. From this point of view, numerical simulation can be used
to evaluate the eectiveness of original control devices, as reported by
Yabuki and Yoshimura (2010) and Wirz (2012). In the rst mentioned
work, a simulation study is described to evaluate the ship-handling
method during the stopping manoeuvre: a turning moment is applied
to the ship by the maximum rudder angle steering prior to the
reversing operation of the Controllable Pitch Propeller (CPP). The
simulation analysis conrms that CPP ships can be suciently
controlled by the proposed method.
The combination of a slow-speed two-stroke diesel engine and a
Fixed Pitch Propeller (FPP) is rather disadvantageous from the point of
view of the crash stop performance. Consequently, Wirz (2012)
proposes and simulates a novel method of applying a braking torque
to the FPP by means of water injection into the engine cylinders during
unred operation.
As regards the characteristics and requirements of a good control
propulsion system, able to prevent overloads and possible failures of
the ship machinery, very little was written so far. Banning et al. (1997)
introduce a tracking control system aimed at saving fuel and optimizing
eciency. A propulsion control procedure, based on several controllers
switching between them to cover the whole operating range of the
vessel, is proposed by Lopez et al. (2010).
However, except for occasional works, the control issue in marine
literature is often addressed from the manoeuvring point of view, such
as in the path-following problems (Skjetne et al., 2005; Fossen, 2011)
and unmanned surface vehicles (Breivik et al., 2008) but a proper
investigation of the engines and actuators behaviour is a key point for
the correct evaluation of every ship manoeuvre (Martelli, 2015).
In line with the latter consideration, the present article deals with
the development of some proper control algorithms for the ship
handling during slam start and crash stop. A rst theoretical applica-
100
Ocean Engineering 137 (2017) 99109
QElim engine torque limit
Qf shaft friction torque
Qhyd hydraulic torque
QP propeller torque
QS shaft torque
QSlim shaft torque limit
Q torque due to the forces interaction between blade and
bearing of the propeller
Si automation signal value
v ship speed vector
Vc chamber volume
xp piston position
H forces and moments acting on ship hull
P forces and moments acting on ship hull
R forces and moments acting on ship rudder
propeller pitch
c commanded propeller pitch
r propeller pitch reduction value
propeller pitch tolerance
rpm revolutions per minute tolerance
tion was considered during the control scheme design of an innovative
propulsion system (Altosole et al., 2012) characterized by a very
exible use of the dierent engine types onboard (gas turbine and
electric motor). Simulation results are shown and discussed, in order to
justify the choice of the proposed special control functions. Similar
simulation approaches to represent crash stop dynamics are adopted
by Krger and Haack (2004) and Schoop-Zipfel et al. (2012). In both
works, the authors address the negative inuence of the wind milling
eect (i.e. propeller driven by the water) on the whole propulsion
system during the ship stopping but do not provide any details
concerning the control procedure to manage this critical situation.
The automation logics presented in this study are suitable for CPP
propulsion systems, since they are characterized by an adaptive
behaviour due to the automatic computation of the propeller pitch
reference. By this way, it is possible to drive a wide range of possible
heavy accelerations and stopping manoeuvres, as demonstrated by the
full-scale validation shown at the end of the article.
2. Overall control system and design methodology
In this section, the main characteristics of a possible control system
layout are described for a generic marine propulsion system. Prime
movers can be diesel engines, gas turbines or even electric motors. The
controller architecture is based on machinery regulation and protection
functions. The purpose of the main propulsive regulation is to provide
the proper revolutions signal to the engine, in order to keep the
commanded propeller speed value achieving the desired velocity of the
vessel. Protection logics aim at preserving propulsion machinery from
overloads (i.e. over-torque, over-speed, over-temperature, etc). All
three types of engines above mentioned are usually controlled in terms
of their rotational speed. In fact the overall controller calculates the
signal reference to the engine governor over time, based on PID
algorithms applied to revolutions error between commanded and
actual values. This control signal is nally transformed by the governor
into the required mechanical torque, by means of acting on the fuel
ow in the cases of diesel engines and gas turbines, or on the frequency
in the case of electric motors.
For the sake of clarity, the main control features are resumed
through the control scheme of Fig. 1, referring to the single shaftline of
a generic propulsion system equipped with Controllable Pitch
Propellers (CPPs).
M. Altosole, M. Martelli Ocean Engineering 137 (2017) 99109

Fig. 1. Overall propulsion control scheme.

The overall controller has to manage propeller pitch as well as of the engine torque QE :
prime movers dynamics.
e2(t ) = QElim QE (t ) (3)
The left side of Fig. 1 shows the control signal of the bridge lever
position, split into two components, acting on the propeller revolutions
and pitch angle respectively. The two signals are modulated over time 3) Shaft torque protection: i.e. Eq. (1) applied to the error between the
by ramps. In detail, the ramp function generates a signal that begins to shaft torque limit QSlim (the maximum load torque of the shaftline)
increase, or decrease, at a constant rate (i.e. the slope of the ramp) until and the shaft actual torque QS :
the target value is reached. The ramp function is usually required to e3(t ) = QSlim QS (t ) (4)
dampen the dynamic eects on the plant due to the input signals
variation. A control function reducing the propeller pitch reference is
Next, the modulated signals are converted into reference values for introduced in case of shaft line over torque. Therefore, the nal value
propeller revolutions and pitch angle, via the combinator laws (i.e. the of the propeller pitch setpoint can be expressed as the dierence
steady state numerical correspondence of the lever position with between the propeller pitch c (commanded by the combinator) and the
propeller rotational speed and pitch) illustrated in Fig. 1 by the speed correction r :
and pitch combinator tables. Thus, the two propeller setpoints are (t ) = c(t )r (t ) (5)
compared with the eld data to calculate the error signals, to which PID
algorithms are applied to obtain the two target values, respectively for The pitch reduction is calculated by a PID algorithm too:
the engine governor and propeller pitch change mechanism. t de3(t )
Referring to the well-known Eq. (1), the calculation of the PID r (t ) = KPe3(t ) + KI 0 e3(t )dt +KD
dt (6)
output Si over the time t is based on the error between setpoint and
feedback, through the action of the proportional gain KP, the integral where:
gain KI and the derivative gain KD. 0 0
t de (t ) r

Si(t ) = KPei(t ) + KI 0 ei(t )dt +KD i r (t ) = r & r >0

dt
On the top of Fig. 1 it is shown that the nal setpoint to the engine
governor is calculated as the minimum value among three signals Si,
achieved through the following i actions (i=13):
1) Shaft speed regulation by means of Eq. (1), applied to the error
e1between the commanded rotational speed nc and the actual shaft
speed n :
e1(t ) = nc(t )n(t )
2) Engine torque protection: i.e. the signal calculated by Eq. (1), on the
basis of the error between the engine torque limit QElim (the
maximum continuous torque of the engine) and the actual value
(1)
r (7)
The conditions represented by Eq. (7) are provided by the satura-
tion block illustrated on the bottom of Fig. 1. In such a way, the c value
can be only reduced and not increased. The pitch reduction can never
be greater than a limit value , decided by the automation designer
(anyway is relatively small to avoid losing excessive performance in
terms of ship speed).
In the case of a twin-screw ship, the pitch reduction for over torque
(2) protection can be made on the two shaftlines independently: this is why
during tight turning circles, the two shaft lines dynamics can be quite
dierent in terms of required power and torque.
In accordance with the control scheme of Fig. 1, the numerical
values of the several control parameters (i.e. PID gains) are chosen so
that overload protection can be rstly achieved by reducing the

101
M. Altosole, M. Martelli
propeller pitch and secondly by decreasing the fuel ow rate (if the rst
action is not sucient), resulting into a rotational speed reference
reduction. Proportional, integral and derivative gains are generically
named in the text as KP, KI and KD, assuming dierent numerical
values in all the above equations according to their specic task. The
proper tuning of such gains is one of the main activities of the
automation designer. This task, usually time consuming and expensive,
can be facilitated by the use of numerical simulation during the
preliminary design phase, although the behaviour of the real hardware
on shipboard could be quite dierent from that one simulated.
In accordance with the scheme of Fig. 2, the Central Processing
Units (CPUs, i.e. the control units) work in real time, exchanging data
with the eld via their Input/Output boards (named IO boards in
Fig. 2). The automation designer should be sure that the performance
foreseen by simulation is maintained in a real ship environment. To
this end, it is necessary to limit, as far as possible, most of the
dierences between CPUs behaviour and numerical simulation. The
main discrepancies could be due to: the cyclic time of the CPUs, the
dierent time delay in exchanging data among controllers and the
native functions that can be implemented. Further dierences could be
represented by the presence of many functionalities usually not
implemented in the ship numerical model (but that interact with the
propulsion control) and the thousands of signals that the automation
has to monitor on the real system.
To cancel the inuence of these kinds of dierences on the design
process, the Real Time Hardware in the Loop (RT HIL) method can be
adopted.
The RT HIL simulation, born primarily in the automotive industry,
is becoming a standard part of the design process for the propulsion
controller of some recent naval vessels, as fully described by Altosole
et al. (2007) and Altosole et al. (2010). According to this design
approach, the ship propulsion system is simulated by a numerical code
that is directly linked to the CPU. In detail, the numerical model is
compiled as an executable le, able to perform simulation in real time
and installed on the same computer of the real controller. The
exchanging of data between controller and propulsion model is possible
by using the OPC (i.e. object linking and embedding for process
control) protocol. The executable le is an OPC client and reads the
command input of the controller that acts as an OPC server. All the
parameters exchanged via OPC can be logged to a le, but they are also
available to the operator by means of a graphical panel. By this way it is
possible to have a comprehensive view of all the working parameters
and the designer can have a realistic feedback about the hardware
behaviour, before its installation on board. Traditionally, the test and
tuning of the real controller is made onboard during the ship delivery
period. These trials are time consuming and very expensive, also
because they require the full availability of the vessel. By using RT
HIL simulation, the physical availability of the ship is not needed,
therefore time and costs can be signicantly reduced. For the recent
applications followed by the authors, the automation designer has
selected the numerical values of the main control parameters by using a
trial and error approach.
Fig. 2. Example of a ship propulsion controller architecture.
102
Ocean Engineering 137 (2017) 99109
3. Control design in emergency manoeuvres
A propulsion controller, as represented in Fig. 1, is characterized by
many variables to be estimated. In detail, for a telegraph consisting of
twenty one possible steps of the lever (i.e. the zero thrust position, ten
positions for the ahead running of the vessel and ten for the astern
condition), the automation designer has to set the numerical values of
the following control parameters:
twenty one values of the shaft speed, corresponding to other twenty
one pitch angles of the propeller, to be inserted into the pertinent
combinator tables;
ten values of the ramp slopes for the ship acceleration and other ten
for the deceleration condition, referring to revolutions setpoint
calculation;
twenty slope rates for the ramp referred to the propeller pitch
setpoint;
proportional, integral and derivative values of the three PID algo-
rithms to calculate the engine setpoint;
proportional, integral and derivative values of the PID algorithm for
the possible pitch setpoint reduction.
In addition, at least two propulsion conditions exist for a ship:
cruising and manoeuvring, each one represented by a dierent
combinator table. Therefore most of the previous control parameters
should be set twice, increasing both complexity and time of the
controller design phase.
Moreover, the numerical tuning carried out for standard and soft
manoeuvres is often not suitable for emergency propulsive conditions
as slam start and crash stop.
The combinator table, together with the pertinent ramps slopes
acting on each signal, can signicantly aect the dynamic behaviour of
the whole propulsion system during heavy accelerations and decelera-
tions. During this kind of transients, the reference values of propeller
revolutions and pitch are commanded by a combinator that is usually
designed for the desired steady state performance. Consequently, its
numerical values cannot be optimized for non-standard operating
conditions.
Without a dedicated control logic for the most critical manoeuvres,
the propulsion performance can be largely reduced and the desired
requirements (e.g. time to reach the maximum ship speed during slam
start or stopping space to perform crash stop) may not be met. This
could be due to the intervention of several protections of the engine
governor, responsible for a sudden cut o of the fuel supply in case of
engine over-torque or over-speed. As an example, a similar situation
will be further detailed in the paragraph 4.1, to highlight the need of
the control optimization process during a typical slam start.
Therefore, to improve the control system behaviour also in o-
design conditions, original control functions are introduced and
analysed in the following. They are planned to manage heavy accel-
eration and decelerations of the ship in a safe and ecient way,
through a systematic design process able to avoid a further dierent
tuning of all the control parameters, previously listed. The core of the
proposed control scheme is represented by the automatic computation
of the propeller pitch setpoint on the basis of the actual propeller
speed. Thus, a proper combinator table, representing the relationship
between pitch and revolutions for all the several lever positions, is no
longer required.
By preventing the engine over-torque as main objective, it is
assumed that other possible kinds of mechanical overloads (e.g.
overloads for shaftline and thrust-bearing) will probably be avoided
(or at least reduced).
In the next sections, the denition of slam start and crash stop is
given from the point of view of the automation system, together with
the description of the pertinent control logics.
In general, the automatic computation over time of the propeller
M. Altosole, M. Martelli
Fig. 3. Flow chart of the slam start control logic.
pitch setpoint is based on the introduction of two distinct parameters
regarding propeller revolutions: a rotational speed tolerance (in the
following named rpm ) to be used for demanding ship accelerations
(e.g. slam start) and a reference constant threshold acting during crash
stop.
3.1. Slam start control process
In this work, slam start condition means every manoeuvre for
which the bridge lever is suddenly moved to its maximum position
(100%), starting from a step value lower than 70%. On this general
basis, slam start manoeuvre can start also when the vessel is moving
astern: in this case, both propeller pitch and speed are managed
according to a proper combinator law, up to the zero thrust condition
of the propeller. On the contrary, if the pitch value at the beginning of
the manoeuvre is greater than the one corresponding to the zero thrust
(0), a PID algorithm calculates the propeller pitch reference, while the
commanded rotational speed linearly assumes its maximum value via a
ramp function.
Fig. 4. Calculation process of the propeller pitch setpoint variation during slam start.
103
Ocean Engineering 137 (2017) 99109
Fig. 5. Example of propeller pitch dynamics during a slam start.
Fig. 6. Example of propeller speed dynamics during a slam start.
The pitch reference computation represents the main core of the
control logic during the slam start of the ship. The purpose is to obtain
the automatic calculation of the pitch reference value during each
heavy acceleration of the vessel, without providing new dedicated
combinator tables and ramps slope rates. This is why it would be very
dicult to foresee a proper unique combinator law for any possible
critical acceleration.
The ow chart of Fig. 3 summarizes the whole control process,
where the computation of the increasing of the pitch setpoint (i.e. the
pitch variation to be added to the initial value) is detailed in Fig. 4.
The proposed control function calculates the propeller pitch
reference up to its design value, making sure that the actual propeller
speed (in Fig. 4 named rpm feedback, i.e. revolutions per minute)
follows its increasing setpoint in a smooth and linear way.
Referring to Fig. 4, the saturation up to zero for negative errors
(between setpoint and actual revolutions) is responsible for the
activation of the pitch freezing. This condition is studied to be
M. Altosole, M. Martelli Ocean Engineering 137 (2017) 99109

activated when the CPP actual speed is too far from its setpoint, in
order to avoid that any increase of the pitch value results in a sudden
and signicant drop of CPP revolutions, with a consequent drop in ship
performance.
The meaning of pitch freezing action during a slam start is
graphically explained by Fig. 5. The stopping of the pitch setpoint is
evident in the whole period in which the propeller speed is too far from
its reference value, as shown in Fig. 6.
The proposed control strategy allows the increase of the pitch
setpoint up to the desired value when the following mathematical
condition is met:
n (t )n(t )rpm<0 (8)
where n(t ) is the actual speed of the propeller depending on time t and
n (t ) is the corresponding nal setpoint corresponding to the maximum
lever position (depending on time t because it is modulated by the
ramp over time). When the dierence between reference and actual
value is lower than rpm , the pitch setpoint grows up to its nal value,
otherwise a zero signal (produced by the saturation eect) feeds the
PID algorithm and consequently the pitch reference increase is stopped
(freezing action).
Numerical simulation allows to tune the value of the speed
tolerance rpm .
of the whole propulsion plant dynamics, inuenced by the ship motions
in six degrees of freedom (i.e. surge, sway, yaw, roll, heave and pitch).
The full description of the motion equations can be found in Martelli
et al. (2014b), where the ship model validation by specic manoeuvring
sea trials (i.e. turning circles, zig zag, etc.) is also discussed. A good
correlation between the experimental and simulated values has been
experienced. The ship dynamics is deduced by applying the forces and
moments equations for a rigid body. Hull, propeller and rudder forces
are considered while environmental forces are not taken into account.
The added masses are evaluated by means of regression formulas.
Given the importance of the control aspects, particular attention has
been paid to the numerical modelling of the engine (Altosole et al.,
2009, 2014) and the CPP mechanism behaviour (Altosole et al., 2012;
Martelli et al., 2014a). Each component of the engine is represented by
steady-state performance maps and thermodynamic equations. The
dynamics of the engine is inuenced by its governor behaviour, that in
this study is simulated in accordance with the indications available
from the manufacturer.
The simulation process can be summarized by solving the following
main dierential equations:
Mv + C (v)v = H + P + R (9)
2Ipn = QE + QP + Qf (10)

3.2. Crash stop control process Ib = Qhyd + Qs + Q (11)

As in slam start condition, the denition of crash stop can be Vc

p = q Cippoil Axp
extended to several possible critical manoeuvres from the automation B oil i (12)
point of view. In the present proposal, crash stop means every
Eq. (9), representing the ship motions in six degrees of freedom, is
condition in which the bridge lever is moved from any forward position
higher than 20%, up to an astern step at least equal to 100%. This
assumption signicantly increases the special cases to be analysed by
the automation designer, complicating the control optimization process
during the design stage. In order to reduce the parameters range to be
set for any possible case, the propeller pitch is automatically adapted to
the engine revolutions variation, in accordance with the control logic
summarized through the ow chart of Fig. 7.
As long as the propeller pitch values are greater than pitch
corresponding to zero thrust (0), the pitch setpoint is calculated by
the proper function described in Fig. 8, based on a PID algorithm
acting on the error between revolutions feedback and a constant
threshold, selected by the automation designer. The PID result is
added to the initial value corresponding to combinator pitch angle at
the beginning of the crash stop manoeuvre.
The saturation block illustrated in Fig. 8 allows the pitch freezing if
the propeller rotational speed increases too much due to a quick
reduction of the pitch angle.
At the same time, the shaft dynamics is managed acting on the
throttle valve of the engine: if the actual rotational speed is greater than
the considered threshold, the engine is immediately commanded to the
idle condition; otherwise it is driven by its governor on the basis of the
usual and standard speed regulation.
On the contrary, for pitch values lower than 0 (i.e. values
corresponding to the astern condition of the vessel), the calculation
of both pitch and shaft speed reference values is managed according to
the combinator law. In light of the latter, revolutions threshold and
combinator values should be chosen high enough to ensure a good
astern thrust of the propeller.

4. Simulation results and shipboard application

The algorithms previously illustrated have been recently proposed

within the control design process of a twin-screw ship propulsion
system, equipped with CPPs. In particular, this new automatic control
system was implemented into a propulsion simulation code, developed
in Matlab-Simulink environment, that was used for the representation Fig. 7. Flow chart of the crash stop control logic.

104
M. Altosole, M. Martelli
Fig. 8. Calculation process of the pitch angle setpoint during crash stop.
Fig. 9. Ship speed simulation during slam start.
Fig. 10. Shaft speed simulation during slam start.
expressed in vector form for sake of compactness (the whole equations
system, including hydrodynamic terms, is fully described in Martelli
et al., 2014b); v is the ship velocities vector; M is the inertia matrix
including added masses; C is the Coriolis centripetal matrix; H , P and
R represent forces and moments acting on hull, propeller and rudder.
Eq. (10) describes the shaft dynamics for the computation of the
propeller rotational speed n , depending on the engine torque QE , the
propeller torque QP and the shaft friction torque Qf ; Ip is the polar
inertia of the rotating masses involved in the shaftline dynamics.
Propeller thrust and torque are simulated by means of the open
water diagrams. Under this assumption, the propeller dynamics is only
modelled as a consequence of the pitch change mechanism behaviour,
represented by Eqs. (11) and (12). In particular, Eq. (11) shows the
motion of the propeller blade around its spindle axis in terms of
angular acceleration , depending on the blade loads (i.e. the hydraulic
torque Qhyd acting on the pin slot, the spindle torque Qs and the torque
105
Ocean Engineering 137 (2017) 99109
Fig. 11. Propeller pitch simulation during slam start.
Fig. 12. Propeller thrust simulation during slam start.
Q that is due to the interaction forces between propeller blade and
blade bearing). Ib is the moment of inertia of the blade. In Eq. (12),
representing the oil pressure poil that is required to turn or to hold the
blade position, qi is the inlet volumetric ow, Cip is the oil leakage
coecient, A is the yoke area, xp is the piston position, B is the oil bulk
modulus and Vc is the chamber volume of the actuator.
4.1. Simulation analysis
For the same vessel, the authors carried out a sensitivity analysis by
simulation (Altosole et al., 2012) that is shown again in this section to
analyse the eects of the proposed control logics on the propulsion
system performance.
Fig. 916 deal with a slam start manoeuvre, while Figs. 1723
illustrate a crash stop. The considered variables are ship and shaft
speeds, propeller pitch and thrust, automation signal driving the
M. Altosole, M. Martelli
Fig. 13. Engine signal simulation during slam start.
Fig. 14. Engine torque simulation during slam start.
Fig. 15. Ship speed comparison between standard and optimum control (slam start).
engine governor and the corresponding brake torque. All results are
reported as percentages of their maximum values: in particular, the
considered maximum torque and thrust correspond to the continuous
engine torque limit and to the thrust-bearing limit.
The simulated slam start consists in a fast acceleration from zero
speed up to full ahead. Three dierent simulations have been
performed, each one with a distinct value of the speed tolerance (i.e.
rpm1 < rpm2 < rpm3). The increase of the parameter rpm leads to
a positive eect on the ship acceleration, as it is evident in Fig. 9.
Figs. 10 and 11 show the propeller revolutions and pitch, following
106
Ocean Engineering 137 (2017) 99109
Fig. 16. Engine behaviour comparison between standard and optimum control (slam
start).
Fig. 17. Ship speed simulation during crash stop.
Fig. 18. Ship stopping space simulation.
their corresponding automation setpoints. From Fig. 10 it is possible to
derive the slope rate of the ramp considered for the speed reference.
The pitch signal is greatly inuenced by rpm , given that the freezing
action is more signicant at low values of rpm . As rpm increases,
thrust and torque are closer to their maximum limit values (Figs. 12
and 14). Especially engine torque reaches the 100% in correspondence
of the highest rpm value.
In conclusion, a too high rpm produces a good ship acceleration
but it signicantly increases the risk of the engine over-torque. In this
M. Altosole, M. Martelli
Fig. 19. Propeller speed simulation during crash stop.
Fig. 20. Propeller pitch simulation during crash stop.
Fig. 21. Propeller thrust simulation during crash stop.
light the speed tolerance value rpm2 could represent a valid compro-
mise to optimise the slam start manoeuvre both from the points of view
of the ship performance and engine behaviour.
To underline the eectiveness of the proposed control algorithm,
the same kind of slam start, but performed in accordance with the
traditional control scheme of Fig. 1, is simulated in terms of ship speed
(Fig. 15) and working points over the engine load diagram (Fig. 16). In
the two gures, the propulsion performance curves, achieved by the
optimization corresponding to rpm2 (dash line) and by the standard
control (solid line), are compared. Through the comparison, it is
107
Ocean Engineering 137 (2017) 99109
Fig. 22. Engine signal simulation during crash stop.
Fig. 23. Engine torque simulation during crash stop.
Fig. 24. Engine signal comparison (crash stop).
possible to note that the ship acceleration time is very similar (i.e.
about 140 s to reach the maximum speed of the vessel) but the
standard control scheme is responsible for a worse behaviour of the
engine. In fact, during the last part of the manoeuvre, the working
points correspond to the maximum power limit of the engine, denoting
a saturation of the throttle valve actuator (in Fig. 16, the dash dot line
represents the maximum output power and speed - and consequently
the maximum torque available from the engine governor). As known,
the saturation on a control system mainly means poor performance
and/or even instability (Levine, 2000). Moreover, the engine is forced
M. Altosole, M. Martelli Ocean Engineering 137 (2017) 99109

Fig. 25. Propeller pitch angle comparison (crash stop). Fig. 28. Propeller thrust comparison (crash stop).

Fig. 26. Engine speed comparison (crash stop). Fig. 29. Ship speed comparison (crash stop).

Fig. 19 and Fig. 20 report the propeller behaviour in terms of

rotational speed and pitch. The latter is greatly inuenced by the
revolutions threshold. In detail, the pitch freezing action is evident at
the beginning of the manoeuvre and around the zero thrust pitch angle
(0). The rst freezing is due to the actual propeller speed that is
higher than threshold value, while the second action is due to the
transition to a dierent pitch setpoint calculation (i.e. the combinator is
considered when the actual pitch angle is lower than 0).
In general, the selection of a greater threshold tends to improve the
ship stopping performance but also makes the automation signal less
stable during the engine acceleration (Fig. 22). In addition, Figs. 21
and 23 show that the threshold variation does not signicantly aect
the maximum values of the propeller thrust and engine torque.

4.2. Results validation by sea trials data

Fig. 27. Engine torque comparison (crash stop).
The control logics developed during the simulation based design
have been implemented in the automation software on shipboard.
to work in a high fuel consumption zone. Based on these considera-
During sea trials, it was possible to achieve the crash stop performance
tions, the proposed control strategy during heavy ship accelerations
illustrated from Fig. 24 to Fig. 29. In the same gures, also simulation
results certainly better than the standard control procedure.
results are reported in order to evaluate the reliability of the design
The same kind of simulation analysis is carried out for the crash
methodology. All the numerical values are given in dimensionless form
stop manoeuvre. Three simulations have been performed, each one
for condential reasons. In the considered case, the ship stopping is
characterized by a dierent value of the revolutions threshold (thresh-
performed moving the bridge lever from 90% up to its maximum astern
old1 < threshold2 < threshold3). Increasing the threshold value, the
position.
ship stopping is faster (Fig. 17). Fig. 18 shows that all three considered
The general dynamics is well predicted although some dierences
thresholds allow to achieve ship stopping distances below 100%,
between simulation and experimental data are present.
representing the desired requirement for the specic application.
It should be considered that the nal tuning of some important

108
M. Altosole, M. Martelli
control parameters (i.e. combinator laws, ramps slope and PID gains)
has never been available from the automation provider. Thus the
present simulation is performed assuming the same numerical identi-
cation carried out during the control design stage and previously
discussed (only the threshold value of shaft revolutions was conrmed
to be equal to threshold2). In this light, the comparison is certainly
aected by the not exact correspondence between simulation and
actual input data, however other interesting considerations can be
made. In particular, Fig. 24 shows an early starting of the engine in
comparison with simulation, due to a dierent calculation of the
propeller pitch setpoint. The latter, in the real system, is characterized
by a faster decreasing (Fig. 25): therefore the zero thrust pitch and
consequently the engine starting are reached about six seconds earlier
than simulation. Since the pitch setpoint is calculated through the error
between a constant threshold of the propeller revolutions and the
actual rotational speed, the time discrepancy between the two pitch
behaviours is essentially due to a dierent representation of the
shaftline dynamics by simulation. Experimental data of Fig. 26 show
an earlier revolutions drop that could mean an imperfect modelling of
the propeller load to the engine. In this regard, it is proper to underline
that the propeller thrust and torque are simulated through the open
water characteristics, so the shaft dynamics equation is solved by using
a steady state representation of the propeller torque. This approach is
usually reliable to simulate ship soft manoeuvres but it may lose
precision for fast decelerations and accelerations of the vessel. An even
better agreement between simulation and sea trials could be achieved
considering a dynamic model for the propeller performance descrip-
tion.
The nal oscillation of the engine speed (solid line of Fig. 26) is due
to the instability of the engine signal, and therefore of the engine
torque, in the last part of the manoeuvre (Fig. 24 and Fig. 27). It may
be a consequence of the interaction between the overall controller and
the engine governor. Simulation does not reproduce this behaviour,
probably because of a not very accurate modelling of the engine
governor (e.g. dierent numerical values of the PID gains between
reality and simulation).
The dierent behaviour of the simulated propeller pitch and speed
aect also the thrust, especially during the rst period of the
manoeuvre. In detail, the simulation result of Fig. 28 reports a lower
thrust of the propeller from the beginning up to fty seconds and a
higher negative magnitude at the end. This nal overestimation leads
to a minor stopping time of the vessel in comparison with sea trials
data, as evident in Fig. 29.
5. Conclusions
The article provides general control schemes to drive CPP propul-
sion systems during standard and emergency manoeuvres of a vessel.
The proposed algorithms are evaluated through numerical simulation,
providing a realistic feedback before the installation on board of the
real control system. An overview of the several control parameters, to
be set by the automation designer, is given in reference to a generic
propulsion plant: simulation shows that their numerical tuning, carried
out for standard and soft manoeuvres, can hardly meet high and safe
performance requirements also during the most critical transient
conditions. Therefore, in order to avoid a new and demanding
evaluation of the several parameters, special control strategies are
proposed for the most stressful manoeuvres (slam start and crash
stop). They are essentially based on the introduction of a reference
rotational speed, to be used for the automatic calculation of the
propeller pitch setpoint. Compared to a standard control procedure,
the two algorithms, developed for fast accelerations and ship stopping,
lead to a reduction of the parameters number to be estimated around
109
Ocean Engineering 137 (2017) 99109
80%. Moreover, the adaptive control can be extended to other
propulsive conditions (e.g. accelerations during demanding turning
circles) further reducing the design activity.
The eorts, due to the simulator development and to the controller
optimization, are largely balanced by the savings obtained in time and
costs to meet the main requirements of eciency and safety. The good
eectiveness of the method is demonstrated by the comparison
between simulation results and sea trials data for a specic application:
the positive response suggests that both the model based design
procedure and the presented control algorithms can be applied
successfully to other new projects.
Acknowledgements
The authors wish to thank SEASTEMA S.p.A. (Fincantieri
Company) and Italian Navy for the support received during the
research activity and the sea trials campaign.
References
Altosole, M., Bagnasco A., Figari, M., Maoletti, L., 2007. Design and test of the
propulsion control of the aircraft carrier Cavour using real-time hardware in the
loop simulation. In: Proceedings of the EuroSiw 2007. June 2007, Genoa, Italy.
Altosole, M., Benvenuto, G., Campora, U., Figari, M., 2009. Real-time simulation of a
COGAG naval ship propulsion system. Proc. Inst. Mech. Eng. Part M: J. Marit.
Environ. 223 (1), 4762.
Altosole, M., Dubbioso, G., Figari, M., Michetti, S., Millerani Trapani, A., Viviani, M.,
2010. Simulation of the dynamic behaviour of a CODLAG propulsion plant. In:
Proceeding of Warship 2010 Conference. London, UK.
Altosole, M., Figari, M., Martelli, M., Orr, G., 2012. Propulsion control optimisation for
emergency manoeuvres of naval vessels. In: Proceedings of the 11th International
Naval Engineering Conference and Exhibition, INEC 2012. May 2012, Edinburgh,
UK.
Altosole, M., Campora, U., Martelli, M., Figari, M., 2014. Performance decay analysis of a
marine gas turbine propulsion system. J. Ship Res. 58 (3), 117129.
Banning, R., Johnson, M.A., Grimble, M.J., 1997. Advanced control design for marine
diesel engine propulsion systems. J. Dyn. Syst.-T ASME 119 (2), 167174.
Black, S., Swithenbank, S., 2009. Analysis of crashback forces compared with
experimental results. In: Proceedings of the 1th International Symposium on Marine
Propulsors (SMP09). June 2009, Trondheim, Norway.
Breivik, M., Hovstein, V.E., Fossen, T.I., 2008. Straight-line target tracking for
unmanned surface vehicles. Model Ident. Control 29 (4), 131149.
Fossen, T.I., 2011. Nonlinear maneuvering theory and path-following control. Mar.
Technol. Eng. 11 (2011), 445460.
Hur, J.W., Lee, H., Chang, B.J., 2011. Propeller loads of large commercial vessels at crash
stop. In: Proceedings of the 2nd International Symposium on Marine Propulsors
(SMP11). June 2011, Hamburg, Germany.
Krger, S., Haack, T., 2004. Design of propulsion control systems based on the
simulation of nautical manoeuvres. In: Proceedings of the 9th Symposium on
Practical Design of Ships and Other Floating Structures. Luebeck-Travemuende,
Germany.
Levine, W.S., 2000. Control System Fundamentals. CRC Press LLC, USA.
Lopez, M.J., Garcia, L., Lorenzo, J., Consegliere, A., 2010. Design methodology based on
H control theory for marine propulsion system with bumpless transfer function.
WSEAS Trans. Syst. 9 (3), 253262.
Martelli, M., 2015. Marine Propulsion Simulation: Methods and Results. Walter de
Gruyter GmbH & Co KG, Berlin, Collection: engineering.
Martelli, M., Figari, M., Altosole, M., Vignolo, S., 2014a. Controllable pitch propeller
actuating mechanism, modelling and simulation. Proc. Inst. Mech. Eng. Part M: J.
Marit. Environ. 228 (1), 2943.
Martelli, M., Viviani, M., Altosole, M., Figari, M., Vignolo, S., 2014b. Numerical
modelling of propulsion, control and ship motions in 6 degrees of freedom. Proc.
Inst. Mech. Eng. Part M: J. Marit. Environ. 228 (4), 373397.
Schoop-Zipfel, J., Abdel-Maksoud, M., Tigges, K., 2012. Stopping simulations and sea
trial measurements of a ship with diesel-electric propulsion and controllable pitch
propellers. In: Proceedings of International Conference on Marine Simulation and
Ship Manoeuvrability (MARSIM 2012). April 2012, Singapore.
Skjetne, R., Fossen, T.I., Kokotovi, P.V., 2005. Adaptive maneuvering, with experiments,
for a model ship in a marine control laboratory. Automatica 41 (2), 289298.
Sutulo, S., Guedes Soares, C., 2015. Development of a core mathematical model for
arbitrary manoeuvres of a shuttle tanker. Appl. Ocean Res. 51, 293308.
Wirz, F., 2012. Optimisation of the crash-stop manoeuvre of vessels employing slow-
speed two-stroke engines and xed pitch propellers. J. Mar. Eng. Technol. 11 (1),
3543, (9).
Yabuki, H., Yoshimura, Y., 2010. On the control of CPP ships by steering during in-
harbour ship-handling. Trans. Nav. 4 (2), 157162.