Research on hot deformation behavior of Zr-4 alloy based on PSO-BP artificial neural network

Abstract The high-temperature deformation behavior of Zr-4 alloy was investigated by using isothermal compression tests performed on the Gleeble-3500 thermomechanical simulator in the range of temperatures 750–1000 °C with strain rates of 0.001–10s−1. Based on the above hot deformation conditions, the flow stresses are predicted efficiently and accurately by the PSO-BP artificial neural network model. The architecture of the network includes two hidden layers; the first hidden layer includes 15 neurons and the second 10 neurons. And the established model was further evaluated in terms of the correlation coefficient, absolute error, and average absolute relative error. The results show the measured and predicted flow stresses have a good agreement, which indicates that the PSO-BP neural network model can be widely used to analyze and predict the hot deformation behaviors at different experimental conditions. As revealed by micrographs, the phenomenon of spheroidization appears at a strain rate of 0.001s−1 below the transformation temperature of the β phase, while the lath-shaped microstructure is transformed into a microstructure that the grain is nearly equiaxed at the same condition mentioned above. However, the microstructure becomes uneven at a high strain rate above the transformation temperature of the β phase.