Beitrag Benedict

Combination of fast-time simulation and automatic assessment for optimising tuning of mathematical simulator ship models

Prof. Dr.-Eng. Knud Benedict, Dr.-Eng. Michael Baldauf,
Dipl.-Eng. Matthias Kirchhoff, Dipl.-Eng. Walter Köpnick, Dipl.-Eng. René Eyrich
Hochschule Wismar - University of Technology, Business and Design; Department of Maritime Studies


  1. Introduction
  2. Current state - process in ship handling simulators and optimisation procedures
  3. Description of new tools for fast and effective simulation and analysis of results
    1. Simulation module SIMOPT
    2. Simulation analysis module SIMDAT
  4. Parameter - optimisation for ship model files
  5. Manoeuvre - optimisation
  6. Conclusion
  7. References
  8. Author's biographies


Abstract: Worldwide there is a need for generating ship models for ship handling simulator to fulfil the needs for training of ships personnel and for research projects dealing with new ship types and port or waterway layouts. The process of modelling dynamic ships characteristics is very complex and can be done in several ways. In simulation centres mostly databases of existing vessels are used as a start to verify the dynamic parameters in order to achieve the best coincidence with the manoeuvring data of the given vessel. This method however is very time consuming because of the many manoeuvres which have to be done to measure the simulation results and to compare with the real ships data.
To improve the effectiveness of this work a fast time simulation algorithm was programmed in order to simulate a set of predefined manoeuvres suitable to modelling purposes. Specific parameters of the ships mathematical model can be varied to get series of results. The results will be assessed to find out the characteristic manoeuvring data. Some work was done to get some insight into the type of parameters relevant for the tuning process.


1. Introduction

The role of simulators in the education and training of seafarers has become more important over the last decades. Simulators are now used for the purpose of improving knowledge and assessing competencies according to the STCW convention. In parallel they are used as research tools for waterway investigations and fairway design, studies for manoeuvring capabilities of new ship types and for many other applications.
For these purposes many types of ships have to be modelled in the simulator. This means there is a great need for fast and effective modelling/tuning processes e.g. for clients from shipping companies and for research projects/contracts. If this modelling process is done manually by conventional tuning methods in the real Ship Handling Simulator (SHS) then there are some disadvantages: there is high time consumption for this processes, up to one month or longer, because manoeuvre simulation is done in real time; even by using the simulator in "fast mode" which is up to ten times faster it is still to slow. Commonly there are no effective tools for supporting the modelling process, e.g. graphical comparison with analysis options. Moreover using the simulator for tuning of models generally means expensive occupation of simulator resources.
In order to avoid these problems a PC-based simulation software was developed at Maritime Simulation Centre Warnemünde (MSCW) with the same ships dynamic capabilities as the Ship Handling Simulator (SHS), except for some environmental impact as for instance shallow water, current or banking effects which are not considered for saving computation time.
In parallel this PC tool can be effectively used for optimising reference manoeuvres which have to be defined to compare the results of students with an optimal track e.g. for the best option for person over board manoeuvres.
In the following paper the concept of the software and some of the features and results will be shown to demonstrate the principle and the application.


2. Current state - process in ship handling simulators and optimisation procedures

The Maritime Simulation Centre Warnemünde at Wismar University, Department of Maritime Studies in Rostock-Warnemünde accommodates six simulators embracing a common network and comprised of four ship-handling bridge systems with differing levels of equipment, a ship's engine system and a VTS simulation facility. The interaction of many of the simulators can be interfaced either mutually or to form a big scenario comprising all simulators (Benedict 2000 [2]).
The Ship handling Simulator (SHS) comprises four bridges: Bridge 1 consists of a fully integrated replica bridge assembly projector-based 360° visual display, Bridge 2 has a similar 257° visual display system which can be specifically used for manoeuvring a ship from bridge wing, the remaining two bridges 3 and 4 are used mainly as radar cabins, each being additionally equipped with 120° visual display screens. A lab with four stations for computer-based Instructorless Training (ILT) completes the setup for effective ship handling training.
In general there are the following processes in the simulator operation, shown in Figure 1, where the left part describes specifically the training process:

  • It starts with the Preparation of Simulation, consisting of preparation of the students (e.g. by Voyage-/Manoeuvre-Planning), and as prerequisites scenarios have to be developed and ships to be modelled.
  • The Manoeuvre-Simulation is done for training with the Ship Handling Simulator as Real-Time-Simulation.
  • Assessment/Optimisation are based on the comparison of training results with training objectives, mostly done by individual debriefing with simulator replay. New tools for semi-automatic assessment by software modules can be used (e.g. by means of the SIMDAT program, see Benedict et al. 2002 [1] and 2003 [5]).
  • Output/Results are given by recording files, tables or score lists.
  • Presentation of Results can be done via bridge "Life Replay" or as beamer presentation in birds eye view on ECDIS or visual observation of the scenario.

> Within these processes there are several sources for optimisation in the operation of the simulators, specifically connected to the assessment process:

  • Training Optimisation plays the most important role, both for qualifying the training process and to improve the students itself as a result of that process (Figure 1 left part, green colour).
  • Manoeuvre - Optimisation is needed during the voyage planning respectively for the assessment process when reference manoeuvres are used to compare the students result with optimal variants of a manoeuvre (Figure 1 centre part, blue colour).
  • Parameter - Optimisation is a method to be applied when creating a new or changing an existing simulator ship model to adjust the parameters and coefficients of the math model in a way that the manoeuvring characteristics of the simulator ship are in coincidence with the results of the real vessel (Figure 1 right part, red colour).

Conventionally all of these tasks were executed by using the ship handling simulator. However, because of the high time consumption in that real time simulation it was decided to develop specific software tools to support this work for the instructors by PC based modules.
As a matter of principle the data format of the ship parameter files and the format of the result files were adjusted via interfaces in that way that an exchange of ship files and result files can be used both for the standalone PC programs and the real ship handling simulator.
In the following chapters several examples are given in order to explain these modules in more detail.
Some results will be presented from initial studies which were made in order to:

  • Gain experience for effect of ship model parameters for ship math model tuning,
  • Gain experience for effect of manoeuvring parameters for ship manoeuvre tuning for optimal/reference manoeuvres.

Figure 1: Processes in ship handling simulator: Training Optimisation for improving students (left, green), Manoeuvre optimisation creating reference manoeuvres (centre, blue) and Parameter optimisation for generating Ship model files (right, red colour)


3. Description of new tools for fast and effective simulation and analysis of results

3.1. Simulation module SIMOPT

The PC-based simulation software was developed at MSCW with the same ships dynamic capabilities as the Ship Handling Simulator (SHS) system, except for some environmental impact as for instance shallow water, current or banking effects which are not considered for saving computation time. The following equation of motion was used as math model for the ships dynamic:

The ships hull forces are normally represented by polynomials based on dimensionless parameters, for instance in the equation for transverse force Y and yaw moment N given as the sum of terms with linear components Nr, Nv, Yr and Yv and additional non-linear terms.
The programming was done in MATLAB and C++.
The Advantage and Capabilities of this software is:

  • The Math Model reveals same simulation results as SHS,
  • It is remarkably faster than "SHS real time simulation " the ratio is up to 1/100,
  • The steering of simulator vessels is done by specific manoeuvre-control settings/commands for standard procedures and individual manoeuvres.

The following figures show some examples of the SIMOPT interface:

Figure 2: SIMOPT Interface Elements - Data Overview and Simulation process monitoring: Ship Data (left), Coefficients, e.g. Clarke (centre) and Manoeuvre Commands (top right) as well as Manoeuvre Optimisation criteria

If a ship is loaded the ships main data are displayed (or can be entered for a new ship) in the left part of Figure 2. The hull coefficients are displayed in the centre. Manoeuvres can be selected from the right top menu. Several options can be chosen from the top menu Figure 3 in order to calculate the hull data and other parameters based on methods published by CLARKE 1983 [6], 1997 [7] or OLTMANN 2003 [8].

Figure 3: SIMOPT Interface Elements - Top Menus: Detailed Selection of Simulation and Analysis Elements from several menus

Figure 4: SIMOPT-Optimising Ship Model Parameters and Manoeuvres by Parameter-Series

Manoeuvres can be selected from the right top menu. Simulations can be done either as single run or as simulation series following the principle to be seen in interface Figure 4 for selection of up to 3 Parameter series to be simulated in parallel or sequential for:

  • Simulation parameters, e.g. Manoeuvre series (here 8 rudder angles)
  • Ship Parameters (L, B, T, or others)
  • Hull/force parameters Clarke coefficient, e.g. Nur
  • Environmental data, e.g. wind force

The example in that figure represents a series of 8 ruder angle variations, 6 parameter changes of Hull yaw moment coefficient Nur and 5 different wind force conditions - that means in total 8x6x5=240 simulation runs, given in the bottom line!
During the simulation run the monitoring of simulation process is clearly indicated e.g. by "coloured bars" in the respective windows Figure 2 at current manoeuvre element.


3.2. Simulation analysis module SIMDAT

The specific new "Offline assessment tool" SIMDAT was originally designed at the MSCW to supply the instructor with semiautomatic assessment of the recorded exercise data (BENEDICT et al. 2003 [5]). The tool allows for a detailed evaluation of the trainees results, e.g.:

  • by analysing the plotted parameters or more complex data (e.g. the risk levels for collision avoidance situation) during the exercise or
  • by comparing the ships track steered by the students with reference tracks.

The concept of data evaluation and assessment tool was to evaluate a variety of different manoeuvres and exercise elements with one common interface. During the evaluation all measurement data are analysed automatically according to selected criteria. Time- and limit- dependent violations are shown in the central window as well as penalty points according to an exercise specific algorithm are given. Apart from the evaluation of students training result the tools were used even in waterway investigation (BENEDICT et al. 2004 [9]).

For the purpose of simulator ships parameter tuning and optimisation of manoeuvres this SIMDAT tool was extended:

  • The Data for the manoeuvring characteristics can now be automatically retrieved for all manoeuvres used for simulator ships tuning
  • Enhanced Graphic tools are available for displaying various types of results

In the upper graphic of Figure 5 the complete track history of a simulation run is shown. A slider on the right hand side of the graphic allows for the timely and detailed analysis of periods during the simulation. The track can be presented in x/y co-ordinates or in geographical co-ordinates.
The lower graphic displays a number of ships data measured during the simulation. This includes Rudder angle, speed or course information of all ships. All graphics can be zoomed so that details of the exercise become visible and the graphical data shown on the surface can be saved and exported for further use.
Depending on the simulated manoeuvre types several special evaluation algorithms are used to produce the results for the manoeuvre as shown in resulting graphs and tables of the particular evaluation.


 

Figure 5: SIMDAT-Simulation result analysis in plots: Main interface and results for turning circle series varying rudder angles - Tracks and plots of time histories

Additionally to the different graphical presentations specific overviews on the results are provided when series of manoeuvres have been simulated. This figure shows a comparison of simulation series results for turning circle with respect to Transfer, Advance, Diameter, Final Speed and Final ROT. It can be given in tables or in diagrams. Moreover the complete set of ships manoeuvring characteristics can be retrieved as a basis for the simulator ships manoeuvring documents as in Figure 6.

 

Figure 6: SIMDAT-Final Graphics of manoeuvring results (tracks of vessel -left; time history of speed - right) for Crash-Stop Tests from 5 different initial speed rates ahead to full astern (CV 7500 TEU)


4. Parameter - optimisation for ship model files

The objective of the parameter optimisation process is to find suitable simulator ship model files which can be used in the simulator to represent the real ships dynamic (see Figure 1, right side). Starting from the ships main data, Basic Ship data files will be generated using simple methods (e.g. according to CLARKE 1983 [6]), to have a first estimation of the dynamic behaviour. By means of the SIMOPT program the fast time simulation produces various results of manoeuvring characteristics which are retrieved by SIMDAT and compared with the manoeuvring characteristics of the real vessel. By changing the Model-Parameters the manoeuvring performance of Simulator Ship Model is improved. The final goal is to achieve a ship file with optimised model-parameters to be applicable for training & research in SHS. The biggest problem is that there are up to 200 parameters and the effects and tendencies of the changes are not very clear; some changes may even have effects which counteract the results of the others. Therefore it is very important to know about the parameters which have a clear impact on the manoeuvring characteristics.
Two examples are given to indicate the effect of tuning:

 

Figure 7: Model tuning (1) Parameter series for changing Moment of Inertia Turning circle tracks and speed and extract of characteristic manoeuvring data

(1) Effect of Hull Parameters:
As an example for varying one of the Hull parameters here the variation of ships moment of inertia Iz is given; this parameter is expressed as kzz² in the database with

For the demonstration a Parameter-Series of turning circles with Hard Rudder to Starboard was simulated varying the value of kzz² (which is initially 0.2) between 0.1 and 0.2 in steps of 0.01. The result in Figure 7 shows a clear effect on the advance of the turning circle whereas the diameter and the speed loss did not change.

(2) Effect of Rudder Parameters As another example the variation of the rudder "race factor" k is given; this parameter is expressed as k in the database. When uR is the inflow speed to the rudder it is affected by the thrust T.

The race factor k can be used to tune this effect. Starting from the initial value k=0.5 it was changed in steps of 0.1 up to 0.8. This simulation series shows clearly in Figure 8 that the radius of the turning circles and the speed was reduced with increasing k.

 

Figure 8: Model tuning (2) Model Parameter series as Example for changing the Race factor k in turning circles to port

The knowledge of those effects can be used to effectively tune simulator ships to have manoeuvring characteristics as real ships. In Figure 9 a comparison is shown from several phases and variants of tuning steps using size of rudder area and the race factor to achieve suitable coincidence between simulation and the real ship.

Figure 9: Comparison of results for turning circles for two simulator models (CV 1600 TEU -left; Multi-Purpose-Emergency Response Vessel with two Azimuth Propellers -right): Blue basic parameter data set (Clarke Estimation); Red- optimised data set by variation of ships rudder area and race factor; Green original trail data

These are the first results in using those tools. Future intentions for improving the model tuning process are:

  • Simulate series of manoeuvres with variations of selected parameter values to get some more experience on the effect of different, important parameter changes
  • Establish a method as "optimisation procedure" to find out the set of parameters in an "automatic way" which fits best with the original manoeuvring data from ship manoeuvring trails.

The experience with model tuning so far have shown some reasons to be very careful: Some physical characteristics for the parameters have to be taken into account in order not to find parameters out of reasonable limits. There can be an excellent coincidence with one manoeuvre, but for others it is not suitable at all. For that reason plots of parameter curves are suitable to see whether the plot of selected forces versus most important parameters indicates a common nature force or moment.

 

Figure 10 shows a graphical analysis of forces specifically for the components of transverse force Y versus drift angle for -180° > > +180° at r=0 and =0 for n=94 RPM. Because the rate of turn and rudder angle are zero in that example, only the components due to the linear term Yuv and the quadratic term Yvv are to be seen here, the others are zero.

Another type of Force analysis is dealing with the time history of forces and moments during manoeuvres: From those plots as Figure 11 one can learn which part of the polynomial expression of forces have impact on the dynamic of the process to find out the most important ones for the tuning.

Figure 10: Components of transverse force Y versus drift angle (at r=0 and =0 for n=94 RPM)

Figure 11: Graphical analysis of forces: Plot of Components of transverse force Y versus time during manoeuvre Zig-Zag-Test 10°/10°


5. Manoeuvre - optimisation

The objective of the manoeuvre optimisation process is to find suitable procedures which can be used in the simulator or in reality with the real ships (see Figure 1, centre).
In the beginning there are standard files for manoeuvre control settings for the specific manoeuvres. By means of the SIMOPT program the fast time simulation produces various results of manoeuvres which are retrieved by SIMDAT and compared with Manoeuvre-Quality Criteria. By changing the Manoeuvring-Parameters (e.g. rudder, RPM…) the manoeuvring performance can be improved. The final goal is to achieve an Optimised Manoeuvre control setting for training & research in SHS. The biggest problem is that there are many options possible and the effect of the changes of the manoeuvring parameters is not very clear; some changes may even have effects which counteracts the results of the others. Results have to be applicable as reference manoeuvre for training & assessment e.g. in SHS.
An example is given below to indicate the need and the effect of manoeuvring optimisation by means of an Emergency Return Manoeuvre.

The STCW Code emphasises a thorough knowledge and ability to apply the procedures of search and rescue operations. In Figure 12 an example of an emergency return manoeuvre is given, well known as the "Scharnow-Turn" (see BENEDICT et al. 1986 [3] and [4]).
The main aim of this person over board manoeuvre is to return the vessel to the original track by the shortest route and with minimum loss of time. In practice the vessel initially follows the turning circle, and after shifting the rudder by a course change of about 240?, finally turns to counter rudder and amidships, the vessel then swings back to the opposite course at a certain measurable distance from the original track, respectively at a certain distance from the reference manoeuvre.
The problem is how to get the "Optimal reference manoeuvre" because the heading change of 240? is an average only and can differ among ships between 225 up to 260° in the same way as for the Williamson Turn which can vary from 25° to 80° instead of the standard average value of 60°.

Figure 12: Reference outline for the Scharnow-Turn

Figure 13: "Scharnow-Turn" Optimisation: Series for different heading changes 220°, 230°, 240°,250° and 260° for counter rudder

Using the SIMOPT and the SIMDAT programs there are two ways to come to the optimal result:

  • The 1st Option is to simulate series of manoeuvres using standard "Scharnow-Turn" manoeuvring commands in automated simulation series: This method can be seen in Figure 13 where several heading changes were used as parameter to vary final result of distance between the initial track and return track.
  • The 2nd Option is to start with a Standard "Scharnow-Turn" manoeuvring command series for automated simulation (centre right) together with optimisation procedure:

An optimising algorithm is used to find suitable heading change for counter rudder as parameter to achieve smallest distance (limit=10m) between initial track and return track on opposite heading (limit=2°). The Optimal track is indicated by yellow colour in Figure 14. The main parameters of the optimised manoeuvre procedure are given in Table 1.

Figure 14: SIMOPT to optimise Emergency Return Manoeuvre: Optimisation procedure

Table 1: SIMOPT to optimise Emergency Return Manoeuvre: Extract & Display of Optimised Manoeuvre Data by SIMDAT-manoeuvring details for Optimised track


6. Conclusion

For optimising the tuning of Simulator Ship Models two software modules SIMOPT and SIMDAT were designed combining fast-time simulation and automatic assessment of the simulation data.
The following advantages can be seen: The Math Model of the PC based version reveals the same results as full mission ship handling simulator, but it is remarkably faster than real time simulation (up to 1/100). Steering of vessels is done automatically by series of manoeuvre-control procedures giving commands for standard procedures and individual manoeuvres. The data for ships manoeuvring characteristics are automatically retrieved. Sophisticated graphic tools allow for displaying results as track or parameter curves together with numerical data presentation. Evaluation tools allow for comparison of different manoeuvres from different sources (e.g. SHS, yards measurement trials). The software tools developed at MSCW give extra support to the simulator instructor allowing for faster evaluation of simulation results both for developing simulator ship models and reference manoeuvre procedures; besides they are used even in waterway investigation.


7. References

[1]Baldauf, M., Benedict, K., Böcker, Th., Felsenstein, C., Herzig, M. "Computer-based evaluation of ship handling simulator exercise results", INSLC, San Francisco, USA, 2002
[2]Benedict, K. "Integrated Operation of Bridge-, Engine Room- and VTS-Simulators in the Maritime Simulation Centre Warnemünde", Conference on Simulation CAORF/JSACC 2000, New York, 3 7 July 2000, Proceedings Vol. 1
[3]Benedict, K.; Hilgert, H. "Rückführung des Schiffes bei Mann-über-Bord-Unfällen" (Returning the ship after man-over-board accidents), Part 1 HANSA, Hamburg, 1986
[4]Benedict, K.; Hilgert, H. "Optimising man-overboard manoeuvres", 15th Conference of Bulgarian Ship Hydrodynamic, Centre Varna, Proceedings Vol. 1, 1986
[5]Benedict, K., Baldauf, M., Felsenstein, C., M. Kirchhoff, M. "COMPUTER-BASED SUPPORT FOR THE EVALUATION OF SHIP HANDLING SIMULATOR EXERCISE RESULTS", MARSIM International Conference on Marine Simulation and Ship Manoeuvrability, Kanazawa, Japan, August 25th - 28th, 2003
[6]Clarke, D., Gedling, P., Hine, G., "The Application of Manoeuvring Criteria in Hull Design Using Linear Theory", Transactions of the RINA, London, pp. 45 68, 1983
[7]Clarke, D., Horn, J.R., "Estimation of Hydrodynamic Derivatives" Proceedings of the 11th Ship Control Systems Symposium, Southampton, U. K., Vol. 3, pp. 275 289, 1997
[8]Oltmann, P., "Identification of Hydrodynamic Damping Derivatives - a Pragmatic Approach", International Conference on Marine Simulation and Ship Manoeuvrability, Kanazawa, Japan, August 25th - 28th, 2003
[9]Benedict, K., Baldauf, M., Herberg, S., Kirchhoff, M., Felsenstein, C., Dettmann, T.: "EXAMPLE FOR INLAND WATERWAY DESIGN INVESTIGATIONS WITH WIND IMPACT IN SHIPHANDLING SIMULATOR AND COMPUTER-BASED ASSESSMENT OF THE RESULTS", IMSF International Simulator Forum Annual Conference 13th 19th Sept 2004 Antwerp, Belgium


8. Author's biographies

K. Benedict achieved his PhD's in Ship Dynamics and on Advisory Systems for Ship Operation. He is Professor/Senior Lecturer for Ship's theory and Vessel Traffic Technology and the Head of MSCW.
M. Baldauf obtained his Ph.D. in Safety Engineering and is presently employed as chief coordinator for research. He is involved in several projects dealing e.g. with AIS.
M. Kirchhoff took his diploma in the field of automation and control engineering. He is working on the development of the computer-based optimization and evaluation tools.
W. Koepnick is an engineer for applied mechanics and is currently working on methods of modelling dynamic ship behaviour.
R. Eyrich graduated as a mathematician and is currently working on methods of modelling dynamic ship behaviour.