Neural Networks with Input Dimensionality Reduction for ‎Efficient Temperature Distribution Prediction in a Warm ‎Stamping Process

Hot stamping involves deforming a heated blank to form components with increased mechanical strength. More recently, warm stamping procedures have been researched. The forming occurs at lower temperatures to improve process efficiency. The process is non-linear and inefficient to solve using finite element simulations and surrogate models. This paper presents the use of dimension-reduced neural networks (DR-NNs) for predicting temperature distribution in FEM warm stamping simulations. Dimensionality reduction methods transformed the input space, consisting of assembly, material, and thermal features, to a set of principal components used as input to the neural networks. The DR-NNs are compared against a standalone neural network and show improvements in terms of lower computational time, error, and prediction uncertainty.