Prediction Of Shape Memory Alloy Recovery Stress Based On BP Neural Networks

As a material of novel function, shape memory alloy(SMA) used as an integrated sensor and actuator has drawn much attention in engineering in recent years.The utilizaion of SMA to improve the intelligent control of structures has become an important research topic. Many scholars carry out the relevant research work and have achieved some results.However, some areas of the study are still in an exploratory or developing stage,such as the mechanical properties and constitutive modeling of SMAs and the BP neural network model of SMA recovery stress.Therefore,it is necessary to continue and further the knowledge base in this area.In this paper, the recovery stress of NiTi SMA wires'has been tested on the BZ2216 Digital Force Measurement Apparatus after refit, and its'change law with the increase and decrease of temperature has been studied, after being prestrained on the Hydraulic Universal Testing Machine, using the self-made fixture on the principle of interference coordinating. Meanwhile, the effect of temperature, incentive modes and incentive times, as well as prestrain level on NiTi SMA wires'have been discussed. It is shown that, the recovery stress of NiTi SMA increases with the increase of temperature, when over the softening point, the alloy becomes soft and the stress decreases; The higher the prestrain level, the greater the recovery stress, but when over 8%, it decreases instead; Compared with being incentived with high current intensity, under low current intensity, the amount of austenite became from martensite is much more, the stress is greater; After being incentived many times, there is still recovery stress.Meanwhile,utilizing the nonlinear mapping capability of artificial neural networks, the BP neural network model of SMA recovery stress is proposed according to the experimental results of SMA under different strains.The results show that the numerical results agree well with experimental observations.This model can describe the nonlinear relationship among stress,strain and temperature of SMA and has good forcasting capability.To a certain extent,this model can predict the correct stress of SMA under a given strain,and the results proved the practicality of BP neural network model.