Estimation of flow stress and grain size uniformity of nickel alloy steel for the heavy plate rolling process

Abstract

Artificial Neural Networks (ANN) is considered one of the most practical technologies in the fields of intelligent engineering and manufacturing. In the hot rolling process of visco-plastic characteristics, the ANN can be applied not only for improving the machining accuracy but also for relaxing the experimental constraints to analyze the critical data such as flow stress for a precise control. With this points, the ANN allows materials with non-linear properties and high strength such as nickel alloy steel to be machined in a wide range of temperature and dimension because the force and torque prediction of rolling process can be stable. In this study, the accuracy estimation between constitutive calculation based on Arrhenius type equation and multi-layer ANN for the nickel alloy steel rolling process is carried out. The flow stress prediction error by constitutive calculation could not represent the nonlinear characteristics of nickel alloy materials because the calculation method includes the average concept of rate of change for influence factors such as alpha, n, InA, activation energy Q. But the ANN of backpropagation could be applied to improve the prediction accuracy over the nonlinear tendency of flow stress. The reliability of flow stress prediction by the ANN of multilayer type is verified by checking the nonlinear characteristics of nickel alloy steel rolling process with the Karman and Orowan's theory. It is found that the standard deviation of flow stress is within 2.7%. It is also found that the ANN method could be applied to plate rolling process with a high accuracy of flow stress prediction of 3.5%. Finally, a higher uniformity of grain size could be obtained through the multi-pass rolling size than that by the forging process. (C) 2019 Elsevier B.V. All rights reserved.