Electroshock plastic constitutive modelling of high strength aluminum alloy based on GA-BP neural network

Abstract. Electrically-assisted forming is an effective way to improve the difficult-to-form materials. On basis of electro- and magneto- plastic effects, our team put forward a novel electroshock treatment (EST) process, which can not only improve the formability of the material but also improve the service performance of the formed parts. However, electrically-assisted forming involves nonlinear deformation with multi-physics coupling, the microstructural evolution and plastic mechanical response of materials under the coupling of electric current field and stress field are highly intricate, posing significant challenges in predicting macroscopic deformation behavior. The study of the characterization and precise control of plastic deformation's constitutive relation is therefore highly significant. In this work, the stress-strain curves of AA7075 under, different current densities, periods and power duration are obtained by means of the pulse electric current assisted tension test, and the intelligent prediction of the electroshock plastic constitutive relationship of AA7075 is realized based on the Genetic Algorithm (GA) optimized Back Propagation (BP) (GA-BP) neural network. The results show that the proposed algorithm can reduce the input parameters of the constitutive model to current, temperature and strain, and improve the computational efficiency. GA is used to optimize the initial weight and threshold of BP neural network, and the best hidden node number was selected as 12 by error verification to train the network. Compared with the traditional BP neural network, the neural network optimized by GA has higher accuracy, and in the plastic deformation stage, the coefficient of determination R2 between the predicted results and the experimental results is basically 99%, which is about 20% higher than that of the BP neural network. At the same time, it solves the data fluctuation in the long-term prediction due to the standard of satisfying the error in the prediction.