Optimal control of SMA cables system based on genetic optimized BP neural network constitutive model

This paper firstly measures the super-elasticity of SMA through the mechanical property test of austenite SMA wire. On the basis of the SMA material property test, it establishes a genetic optimized BP network constitutive model for SMA by considering the effect of loading/unloading rate on the mechanical properties of SMA, and using the experimental data as the training data of neural network. Then, the author processes the constitutive model in MATLAB, uses the improved genetic algorithm to optimize the location and number of SMA in a spatial model structure, and makes seismic response analysis of the optimal configuration. The results show that: the prediction curve of the genetic optimized BP network constitutive model is better agreement with the experimental curve and more stable than that of non-optimized BP network; the BP network constitutive model is easy to invoke, high in precision and beneficial to the MATLAB simulation analysis of SMA control system. Moreover, optimized by the genetic algorithm, the SMA control system can more effectively reduce the seismic response of the structure. For example, the seismic response of the controlled structure is lower than that of the uncontrolled structure by more than 15%, and the control effect of the interlayer displacement response of the structure is more obvious than that of the acceleration response.