Machine learning approach integrated with finite element modelling, and experimentation for optimizing the forming accuracy of U-shaped thin-walled tubes manufactured by free bending forming technology

This study proposed a robust framework that integrates machine learning (ML), finite element (FE) modelling and experimental validation to optimize the free bending forming (FBF) process for manufacturing U-shaped thin-walled metallic tubes. Firstly, a theoretical analysis of the relationship between guide mechanism displacement and bending radius was established. Subsequently, an ML approach-based back-propagation artificial neural network (BP-ANN) model was developed to predict the optimal bending radius (R) and angle (<i>θ</i>) values. The predicted results were verified through FE modelling and experimental trials. The results revealed a high forming accuracy for bending radii up to 99.01% and angles up to 99.59% across varying materials and thicknesses. The obtained results are attributed to the comprehensive selection of training data and the optimization of neural network hyperparameters. The proposed framework demonstrates a reliable and data-driven approach for enhancing precision and efficiency in the FBF process, offering valuable insights for advanced tube-forming applications.