Constitutive Modelling for Restrained Recovery of Shape Memory Alloys Based on Artificial Neural Network

This paper attempts to optimize the constitutive modelling of restrained recovery for the shape memory alloy (SMA). For this purpose, a backpropagation neural network (BPNN) model was developed to predict the restrained recovery of the SMA. The modelling data were collected from restrained recovery experiments on the SMA. Thanks to nonlinear function mapping and adaptation, the proposed model can learn the complete restrained recovery stress and temperature hysteresis of the SMA and predict the complete restrained recovery stress at different initial strains. The result analysis shows that the predicted data agree well with the experimental data. Compared to mathematical constitutive models, the proposed model is simple, cheap and convenient, and especially suitable for real-time applications.