Research On Modeling And Control Methods For Magnetic Controlled Shape Memory Alloy Actuators

In the end of the20th century, a researcher named Vasilevf discovered a new andinteresting material which is called magnetic controlled shape memory alloy. This materialwill produce micro displacement when it is driven by magnetic flux density, and the materialwill recover to its original shape if the magnetic flux density is removed. By way of research,scholars found that magnetic controlled shape memory alloy is a functional material withhigh displacement and good frequency response and it will be the key material which isirreplaceable in micro-nano drives field. But due to the late discovery of this material, theresearch about its control method is far insufficient at home and abroad. In this paper, thehysteresis nonlinearity in magnetic controlled shape memory alloy actuator is analyzeddeeply, and some modeling method and control method aimed at eliminating this hysteresisnonlinearity is proposed. This established a foundation for the further control study ofmagnetic controlled shape memory alloy actuator in the further.In this paper, the advantages of magnetic controlled shape memory alloy material aredescribed first, and the cause and characteristics of hysteresis nonlinearity are introduced.The most major characteristic of hysteresis nonlinearity are that its present output, has thememory of the previous output and may has many corresponding inputs. Therefore, theartificial neural network which can only approach one to one mapping or many to onemapping must be modified to modeling the hysteresis nonlinearity. In order to describe thetwo main characteristics of this nonlinearity, the method adopted in this paper is that theneurons of input layer are set to2. The corresponding inputs are present input which containsthe information of present state, and the previous output which contain all historicalinformation. In this way, the multi-valued mapping of this nonlinearity is changed tosingle-valued mapping, and makes the neural network has memory function. Based on thismethod, PID neural network, BP neural network and PI model are adopted to establishhigh-precision models for hysteresis nonlinearity, and with these models we can lay thefoundation for the control strategy. The weight identification algorithms of these threemodeling methods are momentum BP algorithm, scaled conjugate gradient method andparticle swarm optimization with cross and heredity function respectively. Simulation results show that these three algorithms are better than their corresponding classical algorithm.The inverse PI model has the same structure with PI model, and the elementaryhysteresis operator of inverse model-stop operator can be parsed from the hysteresis operatorof PI model-play operator. Based on this knowledge, the inverse PI model was establishedand the recursion least square method with high computation speed was adopted as theweight identification algorithm. Then a feed-forward controller base on this inverse PI modelwas designed and used to compensate the hysteresis nonlinearity. Simulation results showthat the maximum error rate of this feed-forward control method is0.67%. It is illustratedthat the control signal can make the actual output of actuator approximate the desired outputeffectively.Finally, aims at the problem that the feed-forward control method has weakantijamming capability, the PID hybrid control method was adopted to improve the controlaccuracy based on the feed-forward control. The more parameters particle swarmoptimization searching, the longer the searching time is. Because the PID control has onlythree parameters, particle swarm optimization with cross and heredity function was adoptedto search the optimal parameters in this paper. The PID parameters which were set manuallyare adopted as the initial position value of particle swarm optimization, the optimalproportional, integration and differential parameters are searched. Simulation results showthat the control accuracy of PID hybrid control based on manual adjustment is improvedfrom0.67%(feed-forward control) to0.57%. The control accuracy of PID hybrid controlbased on particle swarm optimization with cross and heredity function improves by0.19%.The system can get higher accuracy and higher anti-jamming performance than thefeed-forward control, the high precision displacement control of magnetic controlled shapememory alloy actuator was realized.