Shape Changes Identification Of The Shape Memory Alloy Based On The Neural Network

Shape Memory Alloys (SMA) are interesting alternative for micro-actuators. Among them, copper alloys are quite promising since the potential strain of monocrystal can reach 12%. Besides, once miniaturized, time constants are reduced to values compatible with mechatronic applications. The unique properties make them the perfect choice for many applications. In the last decade, some of the successful application of SMA actuators has emerged. However these applications only require on-off actuation. In many potential applications, such as muscle wires, position actuator, more precise actuation of SMA is required. And these materials exhibit hysteresis which can seriously affect the performances of the actuator, and make the control quite tricky. To make things worse, the aging of the material can affect in a non-trivial fashion behavior. This research will concentrate on precision control of SMA actuators.This project is the continuation of previous work which focused on the use of Preisach Model to cancel the hysteretic behavior of the material, and to relize open loop operations. This solution shows to be successful but still suffered from a long identification process, and it is not robust to the change of parameters which result from aging.Introduce the Preisach model in order to eliminate the non-linearity hysteresis. This algorithm can avoid the double integration and will be executed more accurately, which is highly desirable feature for real-time control. Furthermore, the identification of the inversion hysteresis model only requires first-order descending curves of hysteresis.A novel memory matrix are adopt for the investigation of the Preisach model and its inversion.With the help of Gauss Neural Network the simulation results demonstrate the performance of the algorithm work. Nonetheless, this approach shows that the non-linearity is separable, and the Preisach Model provides a sound basis for a neural network based inline identification of the hysteresis .It could be extended to the control of the actuator.