Research Of Hysteresis Nonlinearity Compensation Control Method For Magnetic Shape Memory Alloy Actuators

Micro-nano actuators are the key of micro-nano positioning technology which is used tocontrol the plant. Actuators are generally required to meet high displacement precision andhigh displacement resolution. As a new type of smart actuator, magnetic shape memory alloyactuator possesses the advantages of fast response frequency and outstanding straincapability. Therefore, it has a wide application and promotion in the field of aerospae,robotics and micro-control projects.Firstly, the magnetic shape memory alloy actuators’ background and significance wereintroduced. Then the strain principle of actuators was analyzed, and the research actuality ofthe hysteresis compensation control method with different hysteresis models wassummarized.To achieve hysteresis compensation for magnetic shape memory alloy actuator, radialbasis function neural network model and KP model were used for modeling. The radial basisfunction neural network model established in this paper is a two-input single-output model,and its modeling speed is fast. The accuracy of radial basis function neural network model is0.79%. In order to establish a single-input single-output KP hysteresis model, this paper usesBP algorithm and adaptive linear neural network identification method to identify the densityfunction of KP model. The accuracy based on BP algorithm is0.40%, while the accuracybased on adaptive linear neural network identification method is0.19%.In order to improve the controllability of nonlinear system, a feed-forward controllerwas designed to eliminate nonlinear hysteresis influence which was established by aGaussian radial basis function neural network inverse hysteresis model. By this way, theinput-output hysteresis relationship was transformed into approximately one-to-one mappinglinear relationship. The experimental simulation shows that the inverse hysteresis modelbased on neural network has higher accuracy and faster speed of modeling. The maximummodeling error of neural network is1.85%less than the use of traditional modeling approach. The open-loop experimental simulation shows that the maximum tracking error is0.38%which meets the controller design requirements.In order to improve the robustness and adaptability of the hysteresis nonlinear system,this paper designs a robust adaptive controller based on back stepping method. On the basisof high-precision KP hysteresis model, the control law and adaptive law are designed toensure the global system stability. The simulation results show that the robust adaptivecontroller enhances robustness and adaptive performance of closed-loop system, and theactual output can track the desired output. The maximum tracking error is0.76%whichachieves the control accuracy for magnetic shape memory alloy actuators.