A deep neural network model for parameter identification in deep drawing metal forming process

With the growing demand for complex-shaped, lightweight, and high-strength metal components in manufacturing, using thinner sheets has become a trend. However, this can lead to decreased formability and excessive thinning. Thus, accurately and efficiently predicting thickness distribution is essential in engineering applications. During the deep drawing process, material deformation is complex and involves multiple nonlinear problems, including geometry, physics, and boundary conditions, which often require extensive computation time for analysis. This study presents a data-driven model based on Deep Neural Networks (DNN) that captures the intrinsic relationship between thickness distribution and process parameters (such as punch radius, blank holder force, and temperature) in deep drawing. The cylindrical cup is divided into regions based on theoretical analysis and numerical simulation to determine thickness at key points. By predicting the thickness distribution and maximum thinning of SPCC steel sheets under various process parameters, the results demonstrate that the model has strong applicability and efficiency, providing reliable thickness prediction for engineering design.