011 nd Power Electronic, Drive Sytem and echnologie Conference An Experimental Examination of a Fuzzy Logic- Baed DC Scheme H. Valizadeh Haghi, B. Ekandari, E. Pahajavid, M. avakoli Bina Dept. of Electrical Engineering, K.N. ooi Univerity of echnology, ehran, Iran E-mail: valizadeh@ieee.org, ekandari@ieee.org, pahajavid@ieee.org, tavakoli@eetd.kntu.ac.ir Abtract Direct torque control (DC) of an induction motor i a relatively imple cheme making rapid computation while requiring no peed enor. It indicate low enitivity to parameter variation a well. Accordingly, it ha become a promiing alternative to the claic vector control method. hi paper report an experimental implementation of a DC cheme. A one of the main characteritic of the conventional DC i the torque ripple, the mot of our effort have been placed to reduce thi ripple in practice. A fuzzy logic-baed approach ha been employed to achieve a better drive performance by varying the duty ratio of the elected voltage vector during each witching period according to the magnitude of the torque error and poition of the tator flux. he main objective throughout the reported reearch project i to achieve an operational etup but not to propoe a new theory. he effectivene of the control method wa verified by imulation uing Simulink, and experimentally uing MS30F81 DSP-controlled etup. Keyword- Digital ignal proceor; direct torque control; pule-width modulation. I. INRODUCION Following the outtanding advance in emiconductor technology and power electronic during the lat 0 year, the movement toward developing an ideal induction motor drive ha been coniderable. However, the improvement either a the efficiently performing witche or a the poibility of implementing complex algorithm in the new microproceor need to be complemented by ome efficient control method. Indeed, conidering non-linear and multivariable characteritic of a real motor-drive ytem make the development of an ideal control method a profound reearch work. A coniderable reearch effort ha been devoted to thi field with the aim of finding impler method of peed/torque control for induction machine. A a reult, field-oriented control (FOC) and direct torque control (DC) are becoming the indutrial tandard for induction motor torque control [1]. hee two method indicate comparable performance by and large; however, the DC i impler to be implemented. he le computational time of implementing a DC a a enorle control technique with a good performance in teady-tate and tranient operating condition, make thi technique a viable choice for indutrial ue. On the other hand, there have been propoed variou type of technique to olve ome of the problem, inherently aociated with the DC [1]; among them i uing fuzzy φ * * φ Switching able Flux Poition Calculation Flux and orque Etimator Figure 1. Baic DC cheme. Gate Signal Motor Signal technique to moderate the torque ripple a one of the main diadvantage (e.g. [-7]). he torque and flux ripple in the claical DC induction motor drive are due to the fact that none of the inverter witching vector i able to produce the deired change in both torque and tator flux. Variou technique ha been propoed to reduce thee ripple; for example, etting up a higher witching frequency or a change in the inverter topology [8]. However, the duty ratio control by fuzzy rule [4-5, 7] eem to be a good alternative a it would not deteriorate the computational efficiency very much while it keep the witching loe at an acceptable level. A hown [5, 7], fuzzy logic approach control the duty ratio of witching tate uch that the torque ripple can be ignificantly reduced. It hould be noted that uing a higher witching frequency require a fat proceor and increae the cot a well. On the bai of the experience of the author, all improvement to the baic DC contributed to a better performance, neverthele, they could deteriorate the computational efficiency a a main advantage. herefore, according to the pecific feature of the related reearch project, uing a fuzzy control plan ha been preferred in thi paper. An improved DC cheme baed on the fuzzy duty ratio approach ha been examined experimentally. he baic theoretical idea i taken from [4] with ome improvement and practical conideration. Furthermore, an optimized flux reference i employed. Much emphai i being placed on etting up an experimental DC drive having a minimum torque ripple at high peed while preerving an acceptable computational efficiency. So, the method and trategie have been elected and optimized to achieve thi goal in practice. U i 978-1-6184-41-3/11/$6.00 011 IEEE 51
U 4 (011) U 3 (010) q-axi U (110) U 5 (001) U 6 (101) II. BASIC IDEA OF DC he baic DC idea i hown in Fig. 1 chematically. A three level hyterei comparator i ued for making a deciion ignal baed on the error between the etimated value and the reference value of the electromagnetic torque and a two level comparator i ued for the flux error in the ame way. Reference quantitie are depicted by a tar upercript. he baic idea of the DC, a the witching table block in Fig. 1 i in control of it, i to chooe the bet voltage vector, which relocate the flux vector to produce the deired torque. he tip of the flux vector hould be maintained inide a predefined band in the relocating proce. he reulting gate ignal correpond to a et of ix nonzero voltage vector and two zero voltage vector producible from a three-phae voltage ource inverter (VSI), a hown in Fig.. he real-time torque in Fig. 1 can be etimated by 3 = pi jϕ, (1) wherea the tator flux i calculated by a flux oberver or in relation to the meaured voltage and current vector uitable for high peed range [9], ( u R i ) dt. d-axi U 1 (100) Figure. he voltage vector of a two-level three-phae VSI of Fig. 1. ABLE I. DIFFEREN CHOICES FOR BASIC SWICHING ABLE OF DC e φ e φ e φ e φ Scheme 1 U K+1 U K+ U 0, U 7 U 0, U 7 Scheme U K+1 U K+ U K U 0, U 7 Scheme 3 U K+1 U K+ U K U K+3 Scheme 4 U K+1 U K+ U K-1 U K- ϕ = () he rotation of the tator flux vector could be conidered in ix ymmetrical ector correponding to the voltage vector hown in Fig.. A uitable voltage vector i applied to the motor to keep up the flux and torque with their reference value. he baic witching method that have been propoed for thi purpoe are hown in able I [10]. he parameter K repreent the number of the tator flux ector. A can be een, applying the vector U K+1 and U K+ increae the torque while applying the vector U K-1 and U K- decreae the torque within the current tep. he firt cheme i preferred in thi reearch. K III. IMPROVED-DC O MINIMIZE ORQUE RIPPLES A it i mentioned in the introduction, our purpoe i to et up an experimental DC with an emphai on achieving a mall torque ripple to be trade off againt computational efficiency. In thi regard, two well-etablihed method, optimal flux election and fuzzy duty ratio control, are tudied in the following ection and employed in parallel with each other for the experimental examination etup. A. Optimum Flux Reference Controlling motor away from their nominal operating point depreciate their efficiency. herefore, the flux hould be adjuted according to the load variation below the nominal value for optimal performance [11]. hree control method have been propoed for doing thi kind of adjutment: flux control a a function of torque [10], flux control baed on lo model [1], and flux control by a minimum lo earch controller [11, 13]. However, it i not intended here to employ a flux controller to minimize loe but to elect an optimal flux reference baed on the torque reference value. Indeed, the decreae in torque during the zero witching tate depend on both the tator flux magnitude and the frequency. In detail, becaue of a nonlinear dependence between the motor dynamic and the rotor peed, there have been oberved different behavior at low and high peed operation [5]. At low peed, the torque increaing vector are more effective than the torque reducing vector for changing the torque. Converely, the torque increaing vector are far le effective than the torque reducing vector at high peed. On the other hand, a very high torque ripple occur due to both thee behavior if the ampling frequency i too low. However, increaing the ampling frequency in order to adequately compenate the ripple deteriorate the computational efficiency to a great extent. herefore, to minimize the torque ripple (a a main objective of our project beide aving high computational efficiency,) the torque decrement during zero tate mut be kept to it minimum; for the reaon that, the pecific application require a DC drive at high peed range. hi can be done by applying a ufficiently large reference flux to generate the required reference torque. hi will decreae the reactive power conumption a well. he following equation enable calculating uch an optimal flux reference baed on the torque reference value [14]. ϕ ( L 3 3 3 + L ) m + ω( Lr ( L + Lm) + LmL ) 3 R 3 r p ( ω L ) Rr * 1 1 1 1 * = ω m = C * B. Fuzzy Logic-Baed Ripple Minimization A voltage vector i applied for the entire witching period in the baic DC cheme. he reulting voltage duty ratio caue the torque exceed it reference value almot near the beginning the cycle, provoking a high torque ripple. he idea of fuzzy logic-baed torque ripple minimization i to determine the duration for which each torque increaing vector i applied (3) 5
DSP30F81 Interface Control Signal DC Supply Gate Power Stage/VSI Protection Signal DC Control Meaured Current Conditioning Figure 3. Schematic diagram of the experimental etup. ABLE II. SPECIFICAIONS OF INDUCION MOOR Rated power Rated peed Rated voltage Number of pair pole Stator reitance Stator inductance Rotor reitance Rotor inductance Magnetization inductance otal inertia Friction coefficient 1.8 kw 80 rpm (@50Hz) 400 V RMS p = 1 r = 6.0 Ω L l = 3.0 mh r r = 4.9 Ω L lr = 3.0 mh L m = 340.0 mh J = 0.007 Kgm b = 0.0054 Kgm / (a) in anticipation of the torque error would be minimized. A with the baic DC, a look up table (jut like able I,) determine the mot appropriate torque increaing vector to apply. hi vector i then applied for a proportion of the ampling period determined by the fuzzy controller. he ret of the witching period a null tate i elected which won't almot change both the torque and the flux [5]. In thi way, the average inverter output voltage can take a omehow continuou range of value rather than the even dicrete voltage found in conventional DC. he relationhip between the optimal duty ratio per ampling period and the working point (i.e. the motor torque and peed) i non-linear. herefore, it i more practicable to perform a duty-ratio control every witching period by uing a fuzzy controller [6]. In [4], a fuzzy logic-baed DC ha been propoed in which the hyterei torque controller provide only two dicrete value and the zero tate i launched by the fuzzy rule. Alo, in addition to thi conventional fuzzy logic controller, a o-called adaptive block ha been introduced that become active when the two equentially elected tate are ame; and it objective i to put an increment or decrement in the previou duty cycle in order to minimize the torque ripple. he theoretical background and fuzzy rule i employed according to [4] in thi paper. Alo, it ha been ued the fuzzy method in parallel with an optimal flux reference for implementing a high peed induction motor DC drive ytem. he ue of an optimal flux reference complement the torque ripple reduction by fuzzy duty ratio controller; given that a maller decreae in torque ripple can be achieved at high peed by excluively uing the fuzzy controller [5]. Some practical experience are a follow in the next ection. (b) Figure 4. (a) he worktation, and (b) he main proceor, DSP, and a part of the ignal adaptation and the driver board.. IV. SIMULAION AND EXPERIMENAL RESULS Simulation and experimental tet have been carried out in order to verify the practical uability of the implemented etup. he experimental etup (a hown in Fig. 3 and 4) i baed on an induction motor according to able II, IGB-VSI, and a PC/digital ignal proceor (the platform ezdp baed on the fixed point 3 bit DSP MS30F81 from I) a a CPU, where the PC i the hot and the DSP the main proceor. he complete drive control ytem i dicrete. It i developed uing tool from Sim oolbox of Simulink, and i converted in C via Code Compoer Studio. he Simulink imulation ampling time of 5 μ for the DC control routine ha been elected for imulation and ampling time of 180 μ ha been elected for experiment. he reult obtained from the reported project are illutrated in Fig. 5-10. A magnified view of a torque tep change from to 15 N.m i hown in Fig. 5. he correponding experimental examination of a torque repone identical to Fig. 5 i illutrated in Fig. 6. here i a good match between the imulated and experimented change in the reference torque. he atifactory torque ripple i verifiable from thi figure a well. o further examine the operational characteritic, it i hown the magnitude of the tator current during the torque change of Fig. 6 in Fig. 7 and 8, from the imulation and experimental etup, repectively. hee two reult alo illutrate an identical behavior. 53
orque (N.m) 16 14 1 10 8 6 4 0-1.46 1.48 1.5 1.5 1.54 1.56 ime (ec) Figure 5. A magnified view of a torque tep repone. Figure 6. Experimental behavior of a torque tep repone correponding to the change in Fig. 5 (4 N.m/div.) 35 30 5 Stator Current (A) 0 15 10 5 0.8.9 3 3.1 3. 3.3 0.05 m per ample x 10 4 Figure 7. he magnitude of the tator current during the torque change of Fig. 5 and 6 (50m). Figure 8. Experimental behavior of the magnitude of the tator current during the torque change of Fig. 6 (5 A/div.) V. CONCLUSIONS hi paper i an implementation report of a part of a reearch project. he main plan wa to et up a fat-proceing DC for an induction motor with an emphai on reducing the torque ripple at high peed range. After preenting a general outline on electing the uitable method from previouly fullyetablihed method, a fuzzy logic-baed cheme in parallel with an optimal flux reference wa preferred and implemented. he reult illutrated atifactory performance. A the propoed approach i baed on fuzzy logic, it i eay to implement and doe not require much computational time. It hould be mentioned that the whole procedure of thi paper i baed on expert knowledge. REFERENCES [1] D. Caadei, F. Profumo, G. Serra, and A. ani, FOC and DC: two viable cheme for induction motor torque control, IEEE ran. Power Electron., vol. 17, no. 5, pp. 779-788, Sep. 00. [] S. A. Mir, S. Zinger, and M. E. Elbuluk, Fuzzy controller for inverter fed induction machine, IEEE ran. Ind. Applicat., vol. 30, pp. 78 84, Jan./Feb. 1994. [3] P. Z. Grabowki and F. Blaabjerg, Direct torque neuro-fuzzy control of induction motor drive. DSP implementation, in Conf. Rec. IECON 98, Aachen, Germany, Aug. Sept. 31 4, 1998, pp. 657 661. [4] L. Romeral, A. Aria, E. Aldaba, and M. G. Jayne, Novel direct torque control (DC) cheme with fuzzy adaptive torque-ripple reduction, IEEE ran. Ind. Electron., vol. 50, no. 3, pp. 487-49, Jun. 003. [5] I. G. Bird and H. Zelaya De La Parra, Fuzzy logic torque ripple reduction for DC baed AC drive, Electron. Letter, vol. 33, no. 17, pp. 1501-150, Aug. 1997. [6] P. Va, Artificial-intelligence-baed electrical machine and drive. Application of fuzzy, neural, fuzzy-neural, and genetic-algorithm-baed technique, Oxford Univerity Pre 1999. [7] A. Aria, J. L. Romeral, E. Aldaba, and M. G. Jayne, Fuzzy logic direct torque control, Proc. IEEE Int. Symp. Ind. Electron., pp. 53 58, 000. [8] C. A. Martin,. A. Meynard, X. Roboam, and A. S. Carvalho, A predictive ampling cale model for direct torque control of the induction machine fed by multilevel voltage-ource inverter, he Eur. Phy. J. Appl. Phyic. EDP Science 1999. [9] D. Caadei, G. Serra, A. ani, and L. Zarri, Aement of direct torque control for induction motor drive, Bull. Pol. Ac.: ech., vol. 54, no. 3, pp. 37-54, 006. [10] I. akahahi and. Noguchi, A new quick-repone and high efficiency control trategy of an induction motor, IEEE ran. Ind. Appl., vol. IA-, no. 5, pp. 80 87, Sep./Oct. 1986. 54
[11] S. Kaboli, M. R. Zolghadri, and E. Vahdati-Khajeh, "A Fat Flux Search Controller for DC-Baed Induction Motor Drive," IEEE ran. Ind. Electron., vol. 54, no. 5, pp. 407-416, Oct. 007. [1] I. Kiokeridi and N. Margari, Lo minimization in induction motor adjutable peed drive, IEEE ran. Ind. Electron., vol. 43, no. 1, pp. 6 31, Feb. 1996. [13] I. Kiokeridi and N. Margari, Lo minimization in calar-controlled induction motor drive with earch controller, IEEE ran. Power Electron., vol. 11, no., pp. 13 0, Mar. 1996. [14] A. Aria, Improvement in direct torque control of induction motor, Ph.D. diertation, Electron. Eng. Dept., echnical Univ. Catalonia, erraa, Spain, 000. 55