Expert Systems with Applications 27 (2004) 493499 www.elsevier.

com/locate/eswa

The application of an expert system for simulation investigations in the aided design of ship power systems automation
Ryszard Arendt*
Faculty of Electrical and Control Engineering, Technical University of Gdansk, ul. Narutowicza 11/12, 80-952 Gdansk, Poland

Abstract A structure and function of information system for aided design of ship power subsystem automation is presented. More detailed a function of evaluation of design solutions based on simulation investigations is described. Data describing a designed power subsystem is introduced in interactive mode to an expert system, then production rules evaluate correctness to the substance and form of the design. In the case when there are no mistakes, the expert system creates simulation model of designed subsystem and calls program MATLAB-SIMULINK . After a simulation session the program environment returns to the expert system. In the paper a knowledge representation of designed ship power subsystem, and a choice of component model structures of subsystems, which enable an application of production rules for simulation models creation and design correctness evaluation are shown. Also an algorithm of simulation models creation of designed subsystems and an exemplary session of simulation investigations of ship propeller subsystem are presented. q 2004 Elsevier Ltd. All rights reserved.
Keywords: Expert system; Simulation; Design; Ship power system; Dynamic

1. Introduction The design of ship systems automation includes a range of important systems among which we can distinguish: power, freezing, air-conditioning, auxiliary installation, navigation, rolling stabilizer, steam production, and loading systems (Bertram, 1999; Kowalski, Arendt, Meler-Kapcia, & Zielinnski, 2001; Lee & Lee, 1999; Wu, Guo, Chen, & Chen, 1999). A great complexity of the systems, a diversity of the applied technical solutions as well as the strict requirements of classication societies slow down the design process and the evaluation of the proposed solutions. A complex multi-functional information system has been developed that aids ship systems automation design (Arendt, 2002; Kowalski, Arendt, Meler-Kapcia, & Zielinski, 2001). For this purpose a series of research studies was carried out that concerned mainly: investigation and description of the design process of power systems automation, development of databases and knowledge bases connected with the process. The specicity of ship systems automation design was dened as well as the designers tasks during particular stages of the design process (preliminary design, commission project, technical project) with the documentation specied
* Tel.: 48-583-472-157; fax: 48-583-472-487. E-mail address: [email protected] 0957-4174/$ - see front matter q 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.eswa.2004.05.011

by classication societies, ship owners and shipyards. The equipment database was developed together with component elements and ship systems automation and the requirements of classication societies. The database, created in MS Access contains information on automation objects and systems of existing ships as well as catalogue information concerning control elements. The aided design of ship systems automation is largely based on the use of information on automation design of already built ships. Serving this purpose, the algorithm was developed which determines similarity of a ship being designed to particular ships on the basis of the information included in the database. To create the knowledge system for aided design of ship systems automation, a shell expert system Exsys Developer was used. The program is characterized by rule-oriented knowledge representation various methods of inference, the possible use of fuzzy logic and co-operation with databases and other programs. A stimulation investigation subsystem was created that co-operates with the database (Arendt, 2002; Kowalski, Arendt, Meler-Kapcia, & Zielinski, 2001). An use of mathematical models of elements, units and ship systems may constitute a source of deep knowledge for an expert system. The application of simulation investigations during the design stage enables: evaluation of efciency and correctness of the assumed design solutions,

494

R. Arendt / Expert Systems with Applications 27 (2004) 493499

the choice of parameters for control and adjustment systems, the evaluation of static and dynamic properties of the designed systems, as well as the development of the methods of system diagnosis. The application developed in the expert system Exsys Developer V.7.0, Multilogic Inc. aids the design process of control devices of ship power systems and performs two key functions: enables collection of the data describing the structure and elements of the designed subsystem and evaluation of the correctness to the substance and form of the design, creates M-le of SIMULINK and enables a series of simulation investigations in order to evaluate the static and dynamic properties of the designed subsystem. The implementation of simulation investigations in the expert system required systematization and a series of issues had to be resolved, which included: choice of model structures of system component elements that would ensure close typology of a simulation model in regard to project diagram, which facilitates the development of the knowledge rules creating simulation models of the system (Arendt, 1998); dening criteria and developing methods of model parameters selection that would ensure reliability of modeling of certain types of real objects (Arendt, 2001); developing mathematical model libraries of system component elements (Arendt, 1998, 2001, 2003); normalization of system evaluation criteria in classication societies regulations. In this paper the following issues are included: knowledge representation for the designed power subsystem, rules evaluating correctness of the completed design as well as the algorithms of simulation models creation of the designed subsystems. 2. The structure of the expert system Basing on a detailed analysis of the scope of activities performed in particular stages of the ship power systems design, the expert system was assumed to carry out two main functions (Kowalski, Arendt, Meler-Kapcia, & Zielinski, 2001): preliminary design encompassing bidding project and commission project, main design encompassing technical project. In each of these tasks we can distinguish some basic functions, such as: Gathering input information on the designed ship and the equipment selected for installation in the engine room.

Fig. 1. The general structure of system with the database.

The information should be stored in the database of multi-access mode. The system can check their completeness. Searching databases for a completed design in order to nd identical or similar solutions of control devices. Choice of solution on the basis of up-dated system base and automation components base. The system shows either the existing solutions of the completed designs or the lack of solutions. Turning to a user (a designer) for a decision in case of the lack of ready-made solutions or in case of possible alternative solutions (ambiguous solutions). Displaying on a screen (as schemata, description or index) the achieved fragmentary solutions (for example solutions concerning particular objects or ship power subsystems) for a user to accept. Taking into account the requirements of a given classication society and international conventions during the inference process (searching for solutions). Generating the outcome design documentation in a form of a technical description of automation, lists and other text les, schemata, graphic projects as well as data (databases) to be used by other systems. Data preparation and conducting simulation investigations of the designed control devices.

The realization of the functions listed above required a complex structure of the expert system, whose elements are presented as a block diagram in Fig. 1.

3. Knowledge representation of a designed ship power subsystem Decision-making on the design aided by an expert system consists of choosing (assigning) the values for (to) a chosen set of variables. A set of variables with possible variants of values describes a knowledge of the designed system necessary for creating models and conducting simulation investigations in the program environment of MATLAB /SIMULINK .

R. Arendt / Expert Systems with Applications 27 (2004) 493499

495

For each of the designed subsystems, the following data are dened: (a) component elements: general types of elements such as engine, a coupling, a generator; particular types of elements e.g. Diesel engine ZV 40/48 Zgoda-Sulzer; the number of elements for a given general type. (b) structures: indication of an initial element connections of an initial element with an intermediate element; indication of subsequent intermediate elements; further connections of intermediate elements; indication of nal elements and their connections. In the recently developed application of the expert system the ship power subsystems designer can enter data in the dialog mode. The program asks for the following data: the number of general type component elements of subsystem; the proper names of the used elements; the sequence of elements transforming energy, starting with the initial element until the nal one. A nite set of component elements of the designed ship power subsystems was assumed. One can use in maximum: 4 diesel engines (DIESEL1 DIESEL4), 4 shafts (SHAFT1 SHAFT4), 4 couplings (COUPLING1 COUPLING4), 4 control pitch propeller screws (CONTRPROP1 CONTRPROP4), 4 constant pitch propeller screws (CONSTPROP1 CONSTPROP4), 3 summing torque gears (SUMMGEAR1 SUMMGEAR3), 2 distributing torque gears (DISTRGEAR1 DISTRGEAR2), 2 shaft generators (SHAFTGENER1 SHAFTGENER2).

Table 1 Frame with collected data describing the designed propeller subsystem of a ship Element DIESEL1 DIESEL2 DIESEL3 DIESEL4 SHAFT1 SHAFT2 SHAFT3 SHAFT4 COUPLING1 COUPLING2 COUPLING3 COUPLING4 CONTRPROP1 CONTRPROP2 CONTRPROP3 CONTRPROP4 CONSTPROP1 CONSTPROP2 CONSTPROP3 CONSTPROP4 SUMMGEAR1 SUMMGEAR2 SUMMGEAR3 DISTRGEAR1A DISTRGEAR1B DISTRGEAR2A DISTRGEAR2B SHAFTGENER1 SHAFTGENER2 Name PC3V PC3V Connections COUPLING1 COUPLING2

W1

CONSTPROP1

SP1 SP1

SUMMGEAR1 SUMMGEAR1

S2

PRZ1

SHAFT1

as to generate and convert electric energyelectric power subsystems. The system should be analyzed as a coherent unit because the power link chain may include, for example: Diesel enginemechanical energy source, synchronous generatormechanical/electrical energy converter, induction motorelectrical/mechanical energy converter, propeller screwthe receiver of mechanical energy.

The collected data are set up in a frame (le structur.frm). In Table 1 there is an example of the data that describe the propeller subsystem consisting of two middle-speed Diesel engines, working through couplings, a summing torque gear, a shaft for a constant pitch propeller screw. The data collected in the frame are displayed on screen in order to be accepted by the designer as correct.

4. The choice of component element model structures of a ship power subsystem The task of ship power subsystems is to generate and convert mechanical energypropeller subsystems as well

The analysis of ship power systems is complicated by a great variety of the employed devices: the ones generating, transmitting and receiving mechanical energy, mechanical energy converters, electric energy receivers as well as electric energy converters. The applied technical solutions allow for various congurations of the employed elements, which involves the amount and variety of the employed elements together with the possible topology (Arendt, 1998). The ship power systems design requires qualitative and quantitative analysis in static and dynamic state of a system

496

R. Arendt / Expert Systems with Applications 27 (2004) 493499

Fig. 2. An exemplary structure of a propeller unit, the following symbols are accepted: 1, middle speed Diesel engine; 2, a exible coupling; 3, a friction clutch; 4, a gear; 5, a constant pitch propeller screw.

and, in this case, simulation investigations are found to be very useful. The goal is to organize the component elements model structures is such a way that they could be directly connected to each other according to the assumed conguration of the ship power system. The automation of creation of ship power system simulation models based on schemata and descriptions of a design requires maintenance of topological consistence between a designing scheme and the developed structure of simulation model, which is not always attainable since some essential state variables can be confounding in nature. The creation of a power system simulation model must take into account a cause and effect relation between the subsequent component elements of a system. Among the elements models the state variables are exchanged that ensure a correct cause and effect description. The choice of structures of component element models mainly consists in dening the set of state variables conveyed to the other models. Dening as state variable the torque and angle paths in a model of unit propeller seems to be very convenient. During simulation investigation of propeller units it is simple to dene torsional stress of shafts of the component elements and the conditions exceeding the accepted values. The speed and angle acceleration values result from the angle path changes in the time and are easy to obtain with the use of differential blocks. In Figs. 2 and 3 a diagram of an exemplary propeller unit is shown as well as the structure of its simulation model. The choice of state variables: voltage amplitude, current amplitude, frequency, and phase shift in power subsystems models enables the monitoring of all important parameters in electric circuits. In Figs. 4 and 5 an exemplary structure of a tubular rudder and the structure of its simulation model are presented. The given analyses show that it is possible to match the structures of a simulation model of propeller unit component elements that ensure the topological closeness between simulation models and system design schemata.

Fig. 4. An exemplary structure of a tubular rudder; the following symbols are accepted: 1, middle speed Diesel engine; 5, a control pitch propeller screw; 6, generator; 7, induction motor.

The only difference appears in the number of the conveyed data (signals), which results from the different level of abstraction in the designers modelthe design schema and the simulation model. The structures of a model of a ship electrical power system show smaller topological consistence; however, simple production rules were provided to create serial and parallel circuit models of energy receivers. The obtained structures of simulation models enable the algorithmization of simulation model production procedures on the basis of the design schemata of a ship power system. In order to describe a design schema in the form of a list of component elements and a list of dened connections, one can generate simulation program les, describing simulation models of power systems. 5. Algorithms of formal and substantial correctness evaluation of collected data In order to evaluate the correctness of the collected data, the knowledge in the form of production rules is used. The rules evaluate the correctness of the employed elements and the assumed structure of the designed power subsystems. Some of the principles are presented: it is unacceptable to connect directly two Diesel engines (10/10); between a Diesel engine and a propeller screw an intermediate elementa shaft should be used (5/10), if Diesel engines work parallely through an summing gear, then between the Diesel engine and the gear a coupling should be used (6/10);

Fig. 3. The structure of a model of the propeller unit, shown in Fig. 2.

Fig. 5. The structure of a model of the tubular rudder, shown in Fig. 4.

R. Arendt / Expert Systems with Applications 27 (2004) 493499

497

an output torque of a Diesel engine should not be transferred to another one by a shaft (10/10); many shafts should not be used in one line transferring an output torque (5/10); on the shaft output a coupling should not be used (2/10); on the shaft output an adding gear should not be used (2/10); on the shaft output a distributing torque gear should not be used (2/10); on the shaft output a shaft generator should not be used (2/10); an output torque of an adding gear should not be transferred to a diesel engine (10/10); between an summing gear and a propeller screw a shaft should be used (3/10); an output torque of coupling should not be transferred to Diesel (10/10); two couplings should not be connected in series (10/10); an output moment of a distributing torque gear should not be transferred to a diesel (10/10); between a distributing gear and a propeller screw an intermediate elementa shaft should be used (5/10); between a coupling and a propeller screw an intermediate elementa shaft should be used (5/10); The rule concerning the last principle has the following form: Rule number: 94 IF: [CONNCOUPLING2] CONTPROP1 OR [CONNCOUPLING2] CONTPROP2 OR [CONNCOUPLING2] CONTPROP3 OR [CONNCOUPLING2] CONTPROP4 THEN: INCORRECT STRUCTURE OF THE SUBSYSTEMTHE COUPLING OUTPUT CONNECTED TO THE SCREW WITHOUT THE INTERMEDIATE ELEMENTA SHAFT CONFIDENCE 5/10 During the formulation of production rules evaluating the formal correctness of the collected data the following connections were taken into account: The rules evaluating the correctness of the collected data indicate the achievement of the adopted goal within the accepted condence (0/10 10/10). Depending on the goals shown by the rules and its condence, the rules determine the further work of the program. The following cases are possible: 1. the collected data are not complete, wrongly correlated or the designed subsystem shows an unacceptable structure; 2. the collected data show an untypical structure of the designed power subsystem; 3. the collected data are correct and complete.

In the case 1 a designer can nish the program or restart the data collection describing the designed power subsystem. In the case 2 a designer can restart the data collection describing the designed subsystem or go on to a simulation session. In the case 3 the program starts a simulation session without the help of a designer.

6. Creation of simulation models of power subsystems The data describing elements and the structure of the designed subsystem collected in a frame (in the le structur.frm) are the basis of simulation model description. The process of creating M-les of a simulation program SIMULINK can be divided into two basic parts: the subsequent adds of M-les containing the description of component element models to the M.-le of the simulation model of the designed propeller subsystem; the creation of a graphical image of the simulation model of the designed subsystem. The created graphical image of the power subsystem model is divided into discrete elds, to which row and column indices m; n are assigned. For an initial element DIESEL1 a eld (1,1) is assigned, a following elements connected with DIESEL1 have elds 1; 21; n; until nite element is reached. A presence in the row of a distributing torque gear (2 outputs) gives effect of assigning elds with indices 2; n to elements connected with second output. Subsequent Diesels are placed in elds m 1; 1; and the process of assigning eld indices to subsequent elements is repeated. The source of information concerning the following connections of elements is a frame structur.frm. The evaluated data are collected in a frame placement.frm, which is used at the creation of M-le of simulation program. To dene algorithms of creation of simulation program M-le describing subsystem model, the formal analysis of M-le description is done as well as processing of semantic and linguistic data. One can say, that created M-le describing the simulation model of the designed power subsystem should have the following structure: the le name, declarations and formal initial descriptions; the following base parts of le describing models of used component elements, each nished with a line declaring parameters of a macro-block placement on a graphical image; the set of lines declaring macro-block connections, results from an accepted structure of the designed subsystem; the declarations and formal close of the le.

498

R. Arendt / Expert Systems with Applications 27 (2004) 493499

According to what is stated the following les are edited: a le containing the name subsyst.m, declarations and initial descriptions; les containing models of engines; les containing models of shafts; les containing models of a exible coupling and friction clutches; les containing models of constant pith propeller screw and controllable pith propeller screws; les containing models of an adding moments gears; les containing models of a distributing moments gears; les containing models of shafts generators; a le with declarations and formal M-le close. The procedure of the M-le creation with description of the model of the designed power subsystem open the le subsyst.m (clears an existing one). Next it adds to le subsyst.m the name, declarations and formal initial descriptions. The subsequent rows and columns of the pleacement.frm are reviewed and for non-zero indexes of rows and columns the subsequent les with component element models are added. After each le is added the procedure of calculating a placement of a macro-block in the graphical image is started. The added le with a line declaring placement parameters is complemented. After review of the all pleacement.frm the procedure calculating a placement of subsequent lines describing the macro-block connections in the graphical image is started. The procedure of the le subsyst.m creation after adding the le with declarations formal le close is nished. In the procedure of M-le creation the knowledge in form of production rules took part. The algorithm of the creation of simulation models of power subsystems in the Fig. 6 is shown.

Fig. 6. The algorithm of the creation of the simulation models of the designed power subsystems.

7. An exemplary session of ship power subsystem design A power system in conguration: middle speed Diesel engine working through a coupling on constant pitch propeller screw is considered. A designer introduces to the expert system data describing component elements and the structure of the subsystem. The production rules check a formal and substantial correctness of collected data of the designed subsystem. Because there are no mistakes, the expert system creates a frame structur.frm with data of the designed subsystem and creates its simulation model in environment of MATLAB-SIMULINK program (Fig. 7). An image of middle speed Diesel engine with governor (macro-block DIESEL1), two-stage coupling (macro-block COUPL1) and a constant pitch propeller (macro-block CONSTPROP1) is displayed. Accepting the following binary input signals: 0 s step to nominal value of angle velocity of Diesel engine, 12 s the coupling is switching on (I stage),

15 s the coupling begins to work on II stage, 30 s the coupling is switching off, the simulation investigations are carried out. For a simulation the following parameters are accepted: Runge-Kutta 5 method of numerical integration, time of simulation 40 s, maximum step size 0.01,

Fig. 7. A structure of a simulation model of the ship power subsystem.

R. Arendt / Expert Systems with Applications 27 (2004) 493499

499

Fig. 8. Simulated responses of ship power subsystem; angle velocity of Diesel engine (1), propeller (2), load torque of Diesel (3), a state of coupling (4). X-axis in seconds, Y-axis in relative values 1 100%.

minimum step size 0.0001, tolerance 0.001. The achieved results of simulation investigations are presented in Fig. 8. During 4 s an angle velocity of Diesel engine achieves a value close to nominal onethe difference depends on a steady-state error of governor. During the coupling switching on a disturbance of angle velocity in a shape of dumping oscillations are observed. At 15 s due to transferred by coupling load torque (plot 3) an angle velocity of the propeller is growing from null to 0.6 nominal value (plot 2). In Fig. 8 a nominal value of load torque is multiplied by 0.3 to achieve more clear plot. At 15 s the II work step of coupling is switching on; the transferred load torque is growing and at 16 s the shaft is fully coupled. Plot 4 shows the moment of equal angle velocities of both halves of coupling (step from 0 to 1). After uncoupling a great oscillations of angle velocity and load torque (plots 1 and 3) due to step change of load torque and shaft exibility are observed. Angle velocity of the propeller inertial goes to null. The presented plots quite good map real phenomena which are observed during the switching on and switching off a coupling. It is necessary to correct at the I step of coupling work a value of transferred load torque and eventually a time of its work.

the output of many projects made on ships. This knowledge should be subject to structuralisationit should be systematised and documented, which is, in part, done in the expert system (in the knowledge database and in the form of reports). To a large extent, this takes place also in the database during the generation of technical specications using similar ships, and during making result documents such as: materials and suppliers lists, and at further stagesthe statement of checking and measuring apparatus for the engine room automation. The developed application of simulation investigation makes possible to run very effective simulations of the designed systems. The process of acquiring data describing the designed power subsystem and automatic transition to an investigations session takes only a few minutes. The investigations allow for an assessment of basic technical requirements regarding transient processes taking place during: start-up, shut-down, change of parameters of the subsystem, as well as the appearance of factors disturbing the systems operation.

Acknowledgements This research was supported by the Polish Scientic Research Committee under grant No. 4T11A 009 25.

References
Arendt, R. (1998). The choice of component elements for model structures which assure the topological compatibility of a simulation model with the conguration of a power unit. Proceedings of XVI National Conference on Multi-criteria optimisation and CAD MIELNO98, pp. 2330. Arendt, R. (2001). Problems of identication and validation of component elements of ship power systems. Proceedings of the 7th IEEE International Conference on Methods and Models in Automation and Robotics MMAR2001, pp. 613 618. Arendt, R. (2002). An expert system for aided design of automation of ship power subsystems. Proceedings of the 8th IEEE International Conference on Methods and Models in Automation and Robotics MMAR2002, pp. 851 856. Arendt, R. (2003). An application of simulation investigations in design procedures of ship power system automation. Proceedings of the 9th IEEE International Conference on Methods and Models in Automation and Robotics MMAR2003, pp. 163168. Bertram, V. (1999). Knowledge-based systems for ship design and ship operation. First International EuroConference on Computer Applications and Information Technology in the Marine Industries COMPIT2000, Potsdam, pp. 63 71. Kowalski, Z., Arendt, R., Meler-Kapcia, M., & Zielinski, S. (2001). An expert system for aided design of ship systems automation. Expert system with Applications, 20, 261 266. Lee, D., & Lee, K. H. (1999). An approach to case-based system for conceptual ship design assistant. Expert system with Applications, 16, 97104. Wu, T. N., Guo, J., Chen, Y. N., & Chen, W. C. (1999). Control system design and performance evaluation of anti-pitching ns. Journal of Marine Science and Technology, 4, 117 122.

8. Concluding remarks The process of designing ship automation is well qualied for application of an expert system, as it is a complex process in which various kinds of information are processed. This processing is based, to a large extent, on experience and intuition of experts (designers) and should be carried out in as short time as possible. On the other hand, the result of the designing process should have quality, as it directly inuences the operating advantages of the ship, including its safety. There is not fully systematised knowledge about the designing process in designing ship systems automation, because, often, the only source are the people involved in solving problems in this area. Apart from that, there is yet