Machine learning in the prediction of formability in aluminum hot stamping process with multiple variable blank holder force

Abstract

ABSTRACT Multiple blank holders with variable blank holder forces is a promising method to effectively improve the formability of aluminum alloy for the hot stamping process. However, it introduces much more variables of holder forces in the process. The conventional simulation- or experiment-based methods cannot effectively analyze and optimize the hot stamping process. This paper provides a new way to solve this problem by utilizing the advantages of machine learning techniques. Finite element models (FEM) of hot stamping of a box-shaped part considering eight separate blank holders with varying forces were established first. Based on the model, 1000 sets of process parameters and corresponding hot stamping results were generated randomly to provide enough data. A machine learning model was then established to predict the maximum and minimum thickness of the stamped parts. Fourth, a convolutional neural network was established to predict the thickness variation distribution. The results showed that a series of optimal multiple variable blank holder forces were obtained to improve formability, and the machine learning models could accurately predict the thickness distribution. This technique provides an efficient tool in applying multiple blank holders with variable blank holder forces in the hot stamping process.