Study On Micro-Displacement Actuator Of Magectically Controlled Shape Memory Alloy

Magnetically Controlled Shape Memory Alloy (short for MSMA) is a new functional material which is found in 1993. The material has the general shape memory alloy characteristics of large strain and high driving force, and it also can induce large strain under the magnetic field, which is a very suitable intelligent material for making drive. The research of MSMA is mainly concentrated on the influence on characters of microstructure, temperature and components, while little research is done on its application. The primary research of this paper concentrates on modeling and simulation for MSMA micro-displacement actuator external character to lay a foundation of further study on establishing the actuator control system.When the MSMA materials deformation mechanism is analyzed, we found that the relation between temperature, pressure, magnetic field and the rate of deformation is a complex nonlinear multi-variable system. So it is difficult to establish precise modeling through the mechanism analysis method. Therefore data-based machine learning method is used to build model of the MSMA actuator at different temperatures, magnetic field and the pre-pressure based on experimental data of external character.Based on the experimental data obtained from static properties test of MSMA micro-displacement actuator, model is established by using the function approximation of BP neural networks. The neural network modeling avoids the trouble to calculate the modeling parameters based on the material physical properties and solves the material inherent coupling problem between magnetic field, the force field and temperature field. Simulation results show that the modeling has very good accuracy.Based on the experimental data obtained from dynamic properties test of MSMA micro-displacement actuator, dynamic model is set up by using least squares support vector machine regression. The magnetic shape memory alloy dynamic modeling problem is changed into a nonlinear regression function of small sample. The simulation results show that the least squares support vector machine both in the accuracy and the generalization of functions do the best compromise, which is a very effective method for regression analysis and dynamic model building of magnetic controlled shape memory alloy micro-displacement actuator.