NguyenHoSiHung TV pdf Master’s Thesis Graduate Institute of Digital Mechatronic Technology College of Engineering Chinese Culture University Intelligent Control for Magnetic Levitation System Advisor[.]
ύ୯ЎϯεᏢπᏢଣᐒఓπำᏢسኧՏᐒႝᅺγ ᅺγፕЎ Master’s Thesis Graduate Institute of Digital Mechatronic Technology College of Engineering Chinese Culture University ! ᅶੌسϐඵችࠠ!ڋ Intelligent Control for Magnetic Levitation System ࡰᏤ௲:҅Ծ௲ Advisor: Professor Jeng-Tze Huang ࣴ زғǺٜचࢫ Graduate Student: Nguyen Ho Si Hung ύ҇୯ 103 ԃ Д June 2014 ACKNOWLEDGMENTS First of all, I would like to express my deepest gratitude to my research, Professor Zeng-Tze Huang, for giving me valuable advice, guidance, and encouragement on this research and influencing my way of thinking He helped me focus on this topic, gave me the corrections by all his patience He taught me many things not only on my research but also on how an engineer and a researcher should be I have benefited from his vast wealth of knowledge and his constructive comments during this research I would also like to thank to Shiu,Wei-Jen, assistant of the Graduate Institute of Digital Mechatronic Technology for her kindness and enthusiasm of providing me the useful and quick information, helping me to carry out all procedures of the graduation process My thanks are extended research friends who helped me complete this thesis I’m very happy to get in Huang’s laboratory and meet the members Especially, Tran Van Hung and Nguyen Xuan Bao are the members who get in this laboratory at the same time I am very grateful their great support and kindness Finally, I am thankful my parents, my brother, my sisters, girlfriend, all of my good friends and other relatives for their encouragement and support Nguyen Ho Si Hung June 27th 2014 i ABSTRACT This thesis presents an approach to control the magnetic levitation system (Maglev) with uncertainty To achieve precise control system, a neural network based intelligent control system was proposed to compensate the uncertainties First, the dynamic model of a magnetic levitation system was built Second, a sliding mode control (SMC) was applied to compensate the uncertainties that occurred in the magnetic levitation system The applied controller guarantees the stability of the system Moreover, to increase the robustness and to release the requirement of the uncertainty bound, a sliding mode control using a radial basic function network (SMCRBFN) is proposed The effectiveness was verified through the simulation and experimental results Keywords: Magnetic levitation system (Maglev), sliding mode control (SMC), sliding mode control using radial basic function network (SMCRBFN) ii CONTENTS CHAPTER : INTRODUCTION 1.1 Research background 1.2 ECP model 730 maglev machine 1.3 Motivation, methodologies and resources 1.3.1 Motivation 1.3.2 Task execution methodology 1.3.2.1 Research methodology 1.3.2.2 Approach analysis 1.3.2.3 Execution 1.3.2.4 Review 1.3.3 Resource analysis 10 1.3.3.1 Maglev apparatus 10 1.3.3.2 ECP control system software 11 1.3.3.3 MATLAB program and computer system 12 1.3.3.4 Data storage and protection 12 CHAPTER : LITERATURE REVIEW 14 2.1 Sliding model control 14 2.2 Neural network 16 CHAPTER : ECP MODEL 730 MAGLEV 21 3.1 ECP model 730 Maglev 21 3.1.1 Introduction 21 3.1.2 System overview 22 3.1.3 Electromechanical plant 22 3.1.4 ECP executive software 23 3.2 Dynamic system of the Maglev 24 3.2.1 ECP model 730 Maglev machine 24 3.2.2 Nonlinear actuator (coil) characteristic 26 3.2.3 Linearization using Taylor’ Series expansion 29 3.2.4 Nonlinear model of Maglev 31 3.2.5 Test Maglev model with simple algorithm 31 3.2.5.1 Test model at equilibrium positions 31 iii 3.2.5.2 Test model by PID 33 CHAPTER : CONTROL DESIGN 35 4.1 Design sliding mode controller 35 4.1.1 Theory 35 4.1.2 Design sliding mode controller for Maglev 37 4.1.2.1 Design control law of SMC 37 4.1.2.2 Stability of the robust sliding mode controller 39 4.2 Sliding mode control using RBF neural network 41 4.2.1 Theory 41 4.2.1.1 Problem statement 41 4.2.1.2 Controller design based on a radial basic function neural network 42 4.2.2 Design sliding mode control based on RBF neural network for Maglev 44 4.2.2.1 Design control law of SMCRBFN 44 4.2.2.2 Stability of SMCRBFN 47 CHAPTER : SIMULATION AND EXPERIMENTAL RESULTS 49 5.1 Simulation and plant comparison 49 5.2 Simulation results 52 5.2.1 Simulation of sliding mode controller 52 5.2.2 Simulation of SMCRBFN 54 5.2 Experimental results 56 5.2.1 PID control 56 5.2.2 SMC control 60 5.2.3 SMCRBFN controller 64 5.3 Analysis and evaluation of results 68 CHAPTER : CONCLUSION AND FUTURE WORKS 70 6.1 Conclusion 70 6.2 Further works 72 REFERENCE 73 iv LIST OF FIGURES Figure 1.1 : Transrapid Maglev Figure 1.2 : Magnetic Levitation System Figure 1.3: ECP Model 730 Magnetic Levitation Plant 10 Figure 1.4: ECP Control System Software 11 Figure 2.1: An approaching phase (a) and sliding phase (b) 15 Figure 2.2: Structure of a neuron 17 Figure 2.3: Structure of a multilayer feed-forward neural network 18 Figure 3.1: The model 730 Experimental Control System 21 Figure 3.2: MagLev Apparatus, side view (a) and front view (b) (Source: Educational Control Products 1999) 23 Figure 3.3: Free-body diagram 24 Figure 3.4 : Free-body diagram with a magnet 25 Figure 3.5: Physical measurements obtained from the Maglev plant 27 Figure 3.6: Representation of the estimated curve against the experimental data 28 Figure 3.7: Taylor’s linearization theorem representation, about point A (2 cm) 30 Figure 3.8: Nonlinear model of Maglev 31 Figure 3.9: Actuator Characteristics, measured data 32 Figure 3.10: Nonlinear Actuator Characteristic 33 Figure 3.11: Characteristic of simulation model and real system 33 Figure 3.12: Performance of PID through simulation program 34 Figure 3.13: Performance of PID through experimental program 34 Figure 4.1 : Free Body Diagram and Dynamic Configuration 37 Figure 4.2: Sliding mode controller 41 Figure 4.3 : Network structure of an RBFN 45 Figure 4.4 : Control block diagram of an SMCRBFN 46 Figure 5.1: Performance of simulation model with sampling time tdesign=0.0071s (a) and tdesign=0.001768s (b) 50 Figure 5.2: Error of simulation model with sampling time tdesign=0.0071s (a) and tdesign=0.001768s (b) 51 Figure 5.3 : Simulation results of an SMC system due to periodic step command 52 Figure 5.4 : Simulation results of an SMC system due to periodic sinusoidal command 53 v Figure 5.5 : Simulation results of an SMC system due to periodic trapezoidal command 53 Figure 5.6 : Simulation results of an SMCRBFN system due to periodic step command 54 Figure 5.7 : Simulation results of an SMCRBFN system due to periodic sinusoidal command 55 Figure 5.8 : Simulation results of an SMCRBFN system due to periodic trapezoidal command 55 Figure 5.9 : Performance (a) and error (b) of PID due to step command in case 56 Figure 5.10: Performance (a) and error (b) of PID due to sinusoidal command in case 57 Figure 5.11 : Performance (a) and error (b) of PID due to trapezoidal command in case 57 Figure 5.12 : Performance (a) and error (b) of PID due to step command in case 58 Figure 5.13 : Performance (a) and error (b) of PID due to sinusoidal command in case 59 Figure 5.14 : Performance (a) and error (b) of PID due to trapezoidal command in case 59 Figure 5.15: Performance (a) and error (b) of SMC due to step command in case 60 Figure 5.16: Performance (a) and error (b) of SMC due to sinusoidal command in case 61 Figure 5.17 : Performance (a) and error (b) of SMC due to trapezoidal command in case 61 Figure 5.18 : Performance (a) and error (b) of SMC due to step command in case 62 Figure 5.19: Performance (a) and error (b) of SMC due to sinusoidal command in case 63 Figure 5.20: Performance (a) and error (b) of SMC due to trapezoidal command in case 63 Figure 5.21 : Performance (a) and error (b) of SMCRBFN due to step command in case 64 Figure 5.22 : Performance (a) and error (b) of SMCRBFN due to sinusoidal command in case 65 Figure 5.23: Performance (a) and error (b) of SMCRBFN due to trapezoidal command in case 65 vi Figure 5.24 : Performance (a) and error (b) of SMCRBFN due to step command in case 66 Figure 5.25 : Performance (a) and error (b) of SMCRBFN due to sinusoidal command in case 67 Figure 5.26: Performance (a) and error (b) of SMCRBFN due to trapezoidal command in case 67 vii LIST OF TABLES Table 3.1: Definition of Maglev plant-specific variables 25 Table 3.2: Relation between position and control effort 28 Table 5.1: Performance measures of PID in case 58 Table 5.2: Performance measures of PID in case 60 Table 5.3: Performance measures of SMC in case 62 Table 5.4: Performance measures of SMC in case 64 Table 5.5: Performance measures of SMCRBFN in case 66 Table 5.6: Performance measures of SMCRBFN in case 68 viii