Maria DEMS Sławomir WIAK Artur JEŻYK DESIGN OF THE SMALL INDUCTION MOTORS BY MEANS OF DEDICATED EXPERT SYSTEM ABSTRACT The paper describes the architecture of dedicated expert system building with using declarative and also algorithmic language programming. In first case was utilized pc-shell environment and in second Delphi 2005.NET. This expert system is dedicated to designing of the small induction motors. In the proposed expert system, the database of knowledge was created. For the codification of knowledge simple rules, complex rules, frames and neural networks were used. Keywords: induction motor, experts system, electrical motor design Assoc. Prof. Maria DEMS, Ph.D., D.Sc. e-mail: mdems@p.lodz.pl, Prof. Sławomir WIAK, Ph.D., D.Sc. e-mail: wiakslaw@p.lodz.pl Artur JEŻYK, Ph.D. Student, M.Sc e-mail: ajezyk@p.lodz.pl Technical University of Lodz, Institute of Mechatronics and Information Systems, 18/22 Stefanowskiego Street, 90-924 Lodz, POLAND PROCEEDINGS OF ELECTROTECHNICAL INSTITUTE, Issue 240, 2009
262 M. Dems, S. Wiak, A. Jeżyk 1. INTRODUCTION The expert systems constitute part of artificial intelligence field. They may be divided into two groups: shell expert systems and dedicated expert systems. The first one is applied for various purposes using only diversified knowledge bases, and these are more flexible. The second group is used in one main domain. These systems are more advanced, complex and usually give the best results and help in sorting out very difficult problems or tasks. In the design of electrical machines the best results give the use of the dedicated expert systems. 2. ARCHITECTURE OF THE DEDICATED EXPERT SYSTEM Architecture of the dedicated expert systems includes at least three blocks [5, 6]. The first one is user interface, responsible for the communication with the user and also for printing figures, results or graphs. The next one is inference engine that is used during the reasoning process while the system runs. These processes often use knowledge base, which is the last but not least block of the whole expert system. Knowledge bases include rules, neural networks and fuzzy logic [2, 3]. Sometimes the architecture can be more advanced and include also databases or computing blocks if needed, particularly for engineering systems. The proposed expert system uses different programming technologies: pcshell and prolog [1] as programming logic language and Delphi 2005.NET as object language. During the process of programming of the application only in logic language, we don t have the possibility of implementation of complex computing, creating effective user interface etc. On the other hand, we have a very efficient compiler written in C language, possibilities of forward or backward inference and tools which are very useful in verification knowledge bases. Delphi 2005.NET was chosen to build a very efficient application with wide a range of abilities. In the final version, application knowledge from fields experts was implemented. The detailed architecture of the dedicated expert system is shown in Fig. 1. This expert system cooperates with the existing computer program of electromagnetic calculations of the induction motor - STAT [7]. This program
Design of the small induction motors by means of dedicated expert system 263 allows to calculate all electromagnetic parameters and static characteristics of work of the induction motor for the known input data, i.e. all dimensions of the motor. In the designing process of the induction motor it is necessary to make the initial computations of the initial dimensions of the motor, and after that repeat the calculations a few times so as to obtain good final results. This process is made by the proposed designed expert system [4]. File *.res STAT Entry value: Type of machine, rated power, rated voltage, stator circuits connection, rotor speed, power fa ctor, efficiency, starting current, starting torque, maximal torque Expert System Initial Computation: dimensions of the motor core, number and shape of the stator and rotor slots, parameters of the stator and rotor circuits. Criterion referring to: magnetic flux density, current density, technological requirements.res Analysis Knoledge Bases + Inference Engine SE KB1 KB2 File *.dat Fig. 1. Architecture of the dedicated expert system The input dates of the designing expert system are: rated power, rated voltage, connection of the stator circuits (star or triangle), rotor speed, power factor, efficiency, starting current, starting torque and maximal torque. In the first step, the expert system calculates initial dimensions of the motor core, i.e: * external and internal diameter of the stator core (Dse, Dsi), * width of the air gape, * external and internal diameter of the rotor core (Dre, Dri), * number and shape of the stator and rotor slots, * parameters of the stator and rotor circuits and generates the input file (*.dat) to the program STAT. The basic criterions in this step are values of magnetic flux density in the air gape, stator and rotor
264 M. Dems, S. Wiak, A. Jeżyk yoke and stator and rotor tooth, values of the stator and rotor current density and also technological requirements (i.e. minimal values of the teeth width). In the program STAT the other electromagnetic parameters are calculated and the output file (*.res) is generated. This file is analyzed in the engine of the expert system and is compared with the input data. In the case when we have good results, the static characteristics of work are calculated. If the differences between the input and calculated data are too big, the calculations are repeated using the part of the system including knowledge bases with reasoning procedures. 3. CODIFICATION OF KNOWLEDGE IN THE EXPERT SYSTEM In the both steps the different rules are used. In the first step simple rules applied in the system will be included because of low complication level and no necessity of frequent modification. Simple rules are mostly utilized in simple expert systems, because of the fact that their implementation is very easy and procedures of forward chaining are accessible practically in every programming environment. The example of a simple rule enabling the delimitation of indirect conclusions with the use of the system is in the form: if h<=56 then // the rule for depending of the external diameter of the stator Dse from // frame size h begin Dse:<=87; else Dse:<=95; end; Complex rules are recorded in the expert system using pc-shell environment which gives possibilities to use the backward reasoning procedures and the tools to construct or verify knowledge in knowledge bases. These rules enable the realization of the complex actions by the reasoning engine, which requires the activity of many rules. Successive proofed indirect conclusions create a chain of reasoning. In the expert system, the investigation of the components is often connected with lots of effects, for example analyzing other rules,
Design of the small induction motors by means of dedicated expert system 265 questioning users etc. The example of the complex rule codification in pc-shell which allows for determining proper formula to compute internal diameter of stator dsi basing on such data as the number of the pair pole p and external diameter dse has a form: Induction_motor(dsi)= 0.61*dse-4 if Induction_motor (p)=1 & Induction_motor (dse)>80 & Induction_motor (dse)<360 Knowledge in form of the proper formula is transmitted to the system part where this formula is converted from symbolic form to numeric expression and the next expert system finds the correct value (in this case internal diameter of stator dsi) which will be used for the next part of the computation or parameterizations of knowledge base and the further results or conclusions will be received. A frame is a structure which describes a particular object and is assembled with the substructure called cages, known also as slots. Each cage represents a special property of the object described by the frame. Frames are similar to the representation of knowledge through statements, and have the form: (<object> <attribute> <value>) Induction_motor (p)=1 We have to take under consideration that the value of codification in the frame is one of many possible elements of frame. The frame could be divided into minor parts known as facets. The example of the definition of the facet in knowledge base in pc-shell environment enabling to determine the number of the rotor slots depending on number of the stator slots and number of the pole pair of the motor is in the form: facets p: single yes val oneof (1,2,3,4,5,6,8); qs: single yes val range<10, 80>; qr: val range <0, 100>; end; rules 01: qr=22 if qs=24,
266 M. Dems, S. Wiak, A. Jeżyk p=1; 02: qr=28 if qs=36, p=1; 03: qr=40 if qs=48, p=1; 04: qr=26 if qs=36, p=2; 05: qr=42 if ( qs=48 qs=54), p=2; Neural networks implemented in the system allow to approximate complex functions of many variables and also give possibilities of determination of their flow outside the measurement range. In the designed expert system the multilayer neural network was used. Another very big advantage of using neural network is a possibility to utilize network for modeling continuous decision variables which are used for parameterizations of knowledge. 4. CONCLUSIONS The dedicated expert system can be used to design of the small induction motors. This system enables the cooperation of the different programming environments, i.e. pc-shell and prolog as programming logic language and Delphi 2005.NET as object language, which allows to make the complex computing and create effective user interface. In the designed expert system, the database of knowledge is created. For the codification of knowledge simple rules, complex rules, frames and neural networks were used.
Design of the small induction motors by means of dedicated expert system 267 LITERATURE 1. Jeżyk A., Dems M., Wiak S.: Programowanie systemów ekspertowych przegląd technologii, PD FCCS 2005, Wydawnictwo WSHE, Łódź, Poland 2. Jeżyk A., Dems M.: Verification Rule-Based Systems, PD FCCS 2006, Wydawnictwo WSHE, Łódź, Poland 3. Jeżyk A., Dems M., Wiak S.: Knowledge Codification In Expert System On Example Of System Supported Designing Electromechanical Converters, Proceeding of PD FCCS 2007, Bronisławów, Poland 4. Jeżyk A., Dems M.: Expert System Supporting Designing Electromechanical Converters, Proceeding of SMC 2007, October 17-19, 2007, Zakopane, Poland. 5. Jacobson P.: Introduction to Expert Systems, University of Edinburg, 1985. 6. Muławka J.: Systemy ekspertowe, WNT, Warszawa 1996 7. Dems M., Rutkowski Z. Computer program STAT v.4 Electromagnetic Calculations of Small Induction Motors, Computer programs library of Institute of Mechatronics and Information Systems, Technical University of Lodz, Poland, 2004. Manuscript submitted 09.02.2009 Reviewed by Jan Zawilak PROJEKTOWANIE MAŁYCH SILNIKÓW INDUKCYJNYCH Z WYKORZYSTANIEM DEDYKOWANEGO SYSTEMU EKSPERTOWEGO M. DEMS, S. WIAK, A. JEŻYK STRESZCZENIE W pracy przedstawiono strukturę systemu ekspertowego wspomagającego projektowanie małych silników indukcyjnych. Proponowany system ekspertowy został opracowany w środowisku programowania w logice pc-shell, natomiast interfejs użytkownika wykonano w środowisku Delphi 2005.NET. Proces projektowania silnika składa się z dwóch kroków. W pierwszym kroku na podstawie katalogowych danych znamionowych silnika, takich jak: moc znamionowa, napięcie znamionowe, układ połączeń uzwojenia stojana, sprawność, współczynnik mocy oraz wznios osi wału, wykonywane są obliczenia wstępne, w wyniku których wyznaczany jest zbiór danych *.dat zawierający główne wymiary obwodu magnetycznego silnika, liczbę oraz kształty i wymiary żłobków stojana i wirnika oraz parametry uzwojenia stojana (rodzaj uzwojenia,
268 M. Dems, S. Wiak, A. Jeżyk liczba zwojów szeregowych, gałęzi i drutów równoległych, rodzaj i średnica drutu nawojowego). Głównymi ograniczeniami na tym etapie projektowania są maksymalne wartości indukcji magnetycznej w szczelinie powietrznej oraz w poszczególnych elementach obwodu magnetycznego, a także maksymalne wartości gęstości prądu w uzwojeniach stojana i wirnika. Ponadto, w opracowywanych regułach brane są również pod uwagę ograniczenia technologiczne, np. minimalna, możliwa do wykonania szerokość zęba stojana lub wirnika, maksymalny współczynnik zapełnienia żłobka stojana, itp. Wygenerowany przez system ekspertowy zbiór *.dat staje się zbiorem danych wejściowych do programu obliczeń elektromagnetycznych małych silników indukcyjnych STAT v. 4, w wyniku których otrzymujemy zbiór wyników *.res. Zbiór ten jest następnie analizowany przez system ekspertowy pod kątem zgodności danych wynikowych z danymi wejściowymi i w przypadku nie spełnienia narzuconych tolerancji system ekspertowy dokonuje niezbędnych korekt, po czym obliczenia powtarzane są samoczynnie, aż do uzyskania zadawalających wyników. Wykorzystywana w systemie ekspertowym baza reguł i baza wiedzy korzysta z różnych metod kodyfikacji wiedzy, w postaci reguł prostych, reguł złożonych, ramek oraz wielowarstwowej sieci neuronowej. IEl, Warszawa 2009. Nakład 130 egz. Ark. wyd. 22,67. Ark. druk. 16,75. Pap. off. Kl.III. 80 g. Oddano do druku w październiku 2009 r. Druk ukończono w październiku 2009 r. Redakcja Wydawnictwo Instytutu Indeks nr 37656