Application of Fuzzy Control Technique to Thyristor DC Drive Krzysztof Zawirski Institute of Industrial Electrical Engineering Poznañ University of Technology Poznañ, Poland Konrad Urbanski Institute of Industrial Electrical Engineering Poznan University of Technology Poznan, Poland Abstract. In the paper a study of fuzzy logic controller (FLC) application for solution of some control problems of thyristor DC drive is presented. The control problems, which are taken into account during FLC synthesis, consist in variable structure of the current control plant and variation of moment of inertia in the speed control loop. Comparison between fuzzy control system and traditional digital cascade control, carried out by simulation method, proved that FLC as a robust control gives better performance in the range where non-linearity and parameter variation is observed. continuous and discontinuous converter current, and variation of moment of inertia in speed control loop. I. INTRODUCTION Fuzzy logic controllers (FLC) based on fuzzy logic principle and expert knowledge, converted into automatic control systems, still increase their industrial application. In recent years this new technique has been successfully implemented in control system of power electronic converters and converter-fed drives [1,2,3]. Most of these systems are strongly non-linear with variable parameters and structure, what makes them a difficult plant for traditional control technique. Application of fuzzy logic controllers in these conditions provides an efficient tool to design robust control systems [4,5]. In the paper a study of FLC application for solution of some control problems of thyristor DC drive is presented. Improvement of DC drives performance is still an important task because of their application in many industrial branches. Most of the control problems of these drives were solved on the base of traditional control technique. Interesting research tasks is to investigate how the new fuzzy technique can deal with some difficult control problems of DC drives, consisting in non-linear characteristics and parameter variation. Implementation of FLC system to a solution of problems solved yet in traditional way, should provide to better control performance or similar performance but achieved on the more simple way of realisation [4,5,6]. In the presented work a comparison of two control systems for DC drive was provided. One investigated system is a digital control system with connected in cascade controllers of current and speed. Second one is a FLC system designed with similar general structure, decomposed into speed and current controllers connected in cascade, but realised in fuzzy technique. This decomposition gives a chance to apply in synthesis of FLC some expert knowledge, collected during analysis and synthesis of traditional control system. General structure of this FLC system is presented in fig.1. The control problems, which should be taken into account during FLC synthesis, consist in variable structure of the plant in current control loop, being a result of Fig.1 Structure of FLC system for DC drive II. CURRENT CONTROL LOOP A. Control problems in current loop It is well known that as a result of two possible mode of converter operation, with continuous and discontinuous current, the current control plant has variable structure. This is a reason that in many application adaptive current controller - with variable structure PI/I is applied [7]. Proposed in some papers more sophisticated control algorithms require time consuming complex calculations and are more sensitive on error of system parameters identification [7,8]. In the paper such a structure of current FLC was assumed, which can be comparative with traditional adaptive PI/I current controller. Among many known solutions of digital current controller that one was chosen, which concept bases on calculation of current average value only once per converter period [6,7]. This simple solution has structural limit of its dynamic behaviour but it is not important for comparative character of provided analysis. Each structure of adaptive current controller is optimised individually at its range of operation: PI in the range of continuous current and I in the range of discontinuous current mode [7]. Exemplary simulation results for investigated controllers during separate operation of each structure (PI/I) are shown in fig. 2. Investigations were provided with exemplary system consisting of DC motor 15kW/220V/8A/1450rpm supplied from proper 3-phase thyristor rectifier. The problem of adaptive control consists in proper changing over PI and I structures during the changes of converter mode. The simple changing over, at the point of boundary between both ranges of converter operation, provides sometimes to unexpected oscillations or an overshoot in the step response. adaptive current controller concept of PI/I structure changes. First of proposed structures (FLC-I) is the most complex and represents three-input fuzzy controller. In normal realisation these controllers are two (PI) and one (I) dimensional fuzzy processes but in proposed control system the third input signal is applied for proper realisation of adaptation mechanism [9]. As third signal the armature current is introduced. Its value well determines a boundary between rectifier operation modes. Obtained the three-inputs FLC structure is presented in fig.4. Fig.2. Step responses of digital current controller: A) I controller in discontinuous current range, B) PI controller in continuous current range. Fig.4. Structure of 3-inputs fuzzy current controller (FLC-I) This is a result of ambiguous control characteristic of converter, which requires different input signal (control voltage) in discontinuous and continuous mode of operation for the same value of output voltage. Fast controller integration, during discontinuous current mode, provides to much higher value of the control voltage than it is necessary in the continuous range. In result some overshoot takes place, like it is shown in fig.3. This undesired effect can be avoided if correction of the controller settings is provided, but unfortunately it must reduce dynamics of control operation. The provided adaptation mechanism acts in this way that in area near boundary between two modes of converter operation an mixed control effect of both PI and I controller, as a result of proposed fuzzy rules, is achieved. This proposed commutation mechanism between both controller structures is a new farther development of fuzzy switch concept described in [7] and could be named "soft commutation". The linguistic fuzzy control rules for both controllers are presented in table I and II [9]. TABLE I FUZZY CONTROL RULES FOR PI CURRENT CONTROLLER Change of error ( DE ) NB NM NS NZ PZ PS E NB NM NM NM NM NM NS r NM NM NS NS NS NS NZ r NS NM NS NS NZ NZ PZ o NZ NM NS NZ NZ NZ PZ r PZ NS NS NZ PZ PZ PZ PS NS NZ NZ PZ PZ PS (E) PM NZ PZ PZ PS PS PS PB PZ PS PS PM PM PM Fig.3.Step response of digital current controller with simple switching over adaptive mechanism; transient process in both ranges of converter operation B. Structures of current FLC Introduction of the FLC gives a chance to obtain robust adaptive current controller due to implementation of fuzzy adaptive mechanism. For this purpose three different versions of FLC structure were analysed. General idea of all these versions is similar and bases on well known PM NS NZ PZ PS PS PS PS PM PB NZ PZ PS PS PM PM PM PM TABLE II FUZZY CONTROL RULES FOR I CURRENT CONTROLLER Change of error ( DE ) NB NM NS NZ PZ PS PM E NB NB NB NB NB NB NB NB r NM NB NB NB NB NB NB NB r NS NB NB NB NB NB NB NB o NZ NM NM NM NM NM NM NM r PZ PM PM PM PM PM PM PM PS PB PB PB PB PB PB PB (E) PM PB PB PB PB PB PB PB PB PB PB PB PB PB PB PB PB NB NB NB NM PM PB PB PB The adaptation mechanism operates according the membership function of third (current) signal, presented in fig.5. There are only three linguistic values of current signal distinguished. Values PB an NB mean that the current value is enough high to be in continuous current mode and linguistic value ZE (zero) means that the operation point is in the range of discontinuous current mode. digital PI controller. The structure FLC-III is shown in fig.7. TABLE III FUZZY CONTROL RULES FOR ADAPTATION MECHANISM kd ke Current ( Id ) E r r o r (e) NB ZE PB NB PB PB PB ZE ZE ZE ZE Current ( Id ) PB PB PB PB E r r o r (e) NB ZE PB NB ZE PS ZE ZE PS PB PS PB ZE PS ZE The controller is equipped with fuzzy adaptation mechanism operating in the same way like in FLC-II. The same fuzzy rules like presented in table III are used but Fig.5. Diagram of membership functions for current signal The structure FLC-I, the most general and interesting from theoretical point of view, creates some difficulties in practical realisation. This was a reason that for analysis two other versions of structure with more simple realisation, were proposed. Structure FLC-II contains a two-inputs FLC equipped with fuzzy adaptation mechanism. This structure is shown in fig.6. Fig.7. Structure of current controller with adaptive mechanism (FLC-III) instead of normalisation factor coefficients of proportional kp and integral ki controller operation are determined. C. Analysis of adaptive current controller operation Fig.6. Structure of current FLC with adaptive mechanism (FLC-II) The controller has one fixed table of fuzzy control rules but variable normalisation factors due to the adaptation mechanism. Changes of the factors provides to realisation of two different control characteristics PI/I. The adaptation mechanism bases on current signal and on the "soft commutation" process of switching over characteristics, similar like in structure FLC-I. Additional adaptation mechanism is used for variation of the error normalised factor. This mechanism requires as an input signal a control error. Tables III show fuzzy rules of adaptation mechanism for normalisation factors of control error ke and change of error kd. On the base of the rules using simplified reasoning method values of factors ke and kd are calculated in each step of controller operation. Table of fuzzy rules for FLC-II is identical with table I. Membership functions of current signal are similar to presented in fig. 5. The structure FLC-III is similar to the structure FLC-II but contains instead of FLC a traditional Investigations carried out by simulation method showed that structure FLC-I is very difficult for proper adjusting by trial error method. Lack of separated adaptation mechanism makes the adjusting process more complex so in result the performance of the FLC-I is not as good as for other two described structures. The properties of the structures FLCII and FLC-III, which are based on the same adaptation mechanism are comparable. Exemplary simulation result for FLC-III, presented in fig.8, was obtained in the same conditions like transients in fig.3. The presented wave forms proved that effect of robust control was achieved. In the range where each controller PI or I operates individually FLC gave the same control properties like traditional control system. Operation of FLC-II provides to similar wave forms like FLC-III. Fuzzy evaluation of boundary between continuous and discontinuous operation mode (membership functions in fig.5) gives a chance to avoid problem with incorrect estimation of converter parameters. In case of traditional controller with simple (binary) switch over of the structures, accuracy of boundary evaluation plays important role. Fig.8. Step response of the current controller with fuzzy adaptation mechanism (FLC-III). Transient process in both ranges of converter operation Fig.9.A illustrates transient process in the case when controller switching over with wrong evaluation of boundary takes place. Introduction of fuzzy switching mechanism to this controller (FLC-III) provides to better (without oscillations) transient process (fig.9.B). exist in the case of other types of servo drives, DC or AC [2,5,10]. In the paper two fuzzy speed controllers were taken under consideration: • a PI fuzzy controller (FPI), • a sliding-mode fuzzy controller (FSM). These two versions were compared with traditional digital PI controller. In realisation of all investigated speed controllers the same sampling time, equal one period of converter operation, and the same digital current controller, operating with identical sampling time were assumed . Fuzzy control rules for FPI are presented in table IV. Membership functions of input and output signals were modified during tests to achieve controller robustness. Final version of membership functions is shown in fig. 10. Operation of traditional PI and FPI controllers was compared on the base of simulated transient process with different value of moment of inertia. Both controllers were tuned in such a way that for the same big value of moment of inertia, equal triple value of motor moment of inertia, similar transients of control process were obtained. TABLE IV FUZZY CONTROL RULES FOR PI SPEED CONTROLLER (FPI) NB E NB NB r NM NB r NS NB o NZ NB r PZ NB PS NM (E) PM NS PB PZ Change of error ( DE ) NM NS NZ PZ PS PM NB NB NB NB NM NS NB NM NM NM NS NZ NM NS NS NS NZ PS NM NS NZ PZ PS PM NM NS NZ PZ PS PM NS PZ PS PS PS PM PZ PS PM PM PM PB PS PM PB PB PB PB PB NZ PS PM PB PB PB PB PB Fig.9. Step response of adaptive current controller with incorrect evaluation of boundary between continuous and discontinuous mode. Controller with : A) - traditional adaptation, B) - fuzzy adaptation mechanism. Simulation analysis proved that general properties of fuzzy controllers FLC-II and FLC-III are equivalent. The FLC-III is a little more sensitive on non linearity of converter control characteristics but this problem can be avoided by characteristic linearisation using an cos-1 functional block. Investigation results provide to general conclusion that the most simple structure FLC-III should be recommended for future application. III. SPEED CONTROLLER The FLC is implemented in speed control loop with the purpose to obtain a robust speed controller in presence of moment of inertia variation. The same control problems Fig.10 Diagram of membership functions for FPI speed controller The exemplary results are presented in fig. 11.A and 11.C. Calculated in these conditions an integral index of speed control error ( Integral of Absolute Error - IAE) has similar value for both controllers , equal 35.5. Fig.11.B and 11.D present transient process for reduced three times moment of inertia (equal only motor moment of inertia). In this case FPI controller proves to be more robust than traditional PI. These conclusion is confirmed by values of IAE calculated during control process. For FPI its value is equal 26 while for PI it is equal 32. FUZZY CONTROL RULES FOR SLIDING-MODE SPEED CONTROLLER (FSM) Change of error ( DE ) NB NM NS NZ PZ PS PM E NB NB NB NB NB NB NM NS r NM NB NB NB NB NM NS ZE r NS NB NB NB NM NS ZE PS o NZ NB NB NM NS ZE PS PM r PZ NB NM NS ZE PS PM PB PS NM NS ZE PS PM PB PB (E) PM NS ZE PS PM PB PB PB PB ZE PS PM PB PB PB PB PB ZE PS PM PB PB PB PB PB Proposed membership functions for FSM controller are shown in fig. 12. The effect of FSM controller operation can be observed in fig. 13, which presents wave forms of the control process in the same conditions like in fig.11. Comparision of both figures shows that FSM controller provides to much better performance. Calculated values of IAE in this case are equal 35 for big and 21 for small value of inertia moment. Fig.11. Step response of speed controllers : A,B) digital PI controllers, C,D) FPI controller; moment of inertia : A,C) J=3JM, B,D) J=JM More interesting result was obtained by introducing fuzzy sliding mode controller - FSM. The simple FSM structure, easy for realisation, was proposed. The controller has only partial compensation effect and can be described in the form: . Iref = C ⋅ Id + U fuzzy(e,e) (1) where Iref is the output signal of FSM speed controller (reference signal for current controller), C Id is a component of partial compensation and Ufuzzy(e,e) is the output signal of fuzzy term, which realises sliding mode control with boundary layer [1]. Value of compensation coefficient C was assumed equal 0.85. Final version of fuzzy rules is given in table V. TABLE V Fig.12 Diagram of membership functions for sliding mode fuzzy speed controller (FSM) IV. CONCLUSIONS Some example of fuzzy control technique application to the control of adjustable speed thyristor DC drives was elaborated. Fuzzy technique provided a new control tool to deal with control problems of such non-linear plant. [5] [6] [7] [8] [9] [10] Conference PEMC’96, Budapest 1996, vol.3, pp. 3.462-3.466. M.A. Denai and A. Hazzab, "Fuzzy control of a separately excited DC motor", Proceedings of the 7th International Power Electronic and Motion Control Conference PEMC’96, Budapest 1996, vol.3, pp. 3.477-3.480. A. Debowski, "Armature-current fuzzy controller for DC-motor drives", Proceedings of 6th International Power Electronic and Motion Control Conference PEMC’94, Warsaw 1994, pp. 136-141. A. Monti, A. Roda, A. Scaglia and P. Vas "A new fuzzy approach to the control of the synchronous reluctance machine", Proceedings of the 7th International Power Electronic and Motion Control Conference PEMC’96, Budapest 1996, vol.3, pp.3.106-110. C. Karaali C and D. Naunin, "Implementation of fuzzy control structures to an intelligent synchronous motor", Proceedings of 6th International Power Electronic and Motion Control Conference PEMC’94, Warsaw, 1994, pp. 115-120. W. Leonhard, Control of electrical drives, SpringerVerlag Berlin Heilderbeg , 1990 H.T. Nguyen and D. Sands "Real-time self organising fuzzy logic controller for DC servo", Proceedings of the European Conference on Power Electronics EPE´93, Brighton 1993, pp.174-179. 1 Fig.13 Step response of sliding mode fuzzy speed controller (FSM); moment of inertia : A) J=3JM, B) J=JM Comparison between proposed fuzzy control system and traditional digital cascade control, carried out by simulation method, proved that FLC as a robust control gives better performance in the range where non-linearity and parameter variation are observed. Presented in the paper FLC structure for adaptive current control and fuzzy sliding-mode speed controller can be recommended for practical implementation. V. REFERENCES [1] D. Driankov, H. Hellendoorn and M. Reinfrank, An introduction to Fuzzy Control, Springer-Verlag, Berlin Heidelberg, 1993. [2] A. Farjon and M. Tholomier, "Why and how using the fuzzy logic in the power electronic field", Proceedings of the European Conference on Power Electronics EPE’95, Seville 1995, pp.3.127-3.132. [3] W. Pedrycz, Fuzzy Control and Fuzzy System, Research Studies Press LTD., Taunton Somerset England, 1993. [4] P. Brandstetter and P. Sedlak ,"Fuzzy control of electrical drive using DSP", Proceedings of the 7th International Power Electronic and Motion Control 1The work was supported in part by a grant from State Committee for Science Research (Poland), contract 42-926/T10/96.