Research on the Curvature Prediction Method of Profile Roll Bending Based on Machine Learning

Roll-bending technology has a high degree of flexibility and does not require special molds. However, based on the existing plastic mechanics theory finite element simulation, it is difficult to accurately analyze complex spatial relationship profile roll forming. Therefore, fixed-curvature prediction model constructed XGBoost (extreme gradient boosting), coupling effect process parameters material performance roll-forming explored. Combined with Bayesian optimization algorithm, hyperparameters are optimized. In addition, result fixed curvature, variable-curvature established using conditional random field (CRF). To further improve accuracy, an error compensation network added after CRF in order map discrete sequence continuous sequence. The experimental results show that mean square (MSE), absolute (MAE), percentage (MAPE) predicted by models above much smaller than other methods, which verifies superiority models.