[成形过程的数据挖掘与深度学习方法]基于机器学习的管材弯曲回弹有效预测与补偿

采用基于优化的误差反向传播(BP)神经网络的机器学习算法建模，提出了考虑材料参数、几何参数等多因素的弯管回弹精确预测和高效控制方法。该方法通过引入非线性惯性权重及遗传算法的杂交算子，改进了粒子群优化(PSO)算法，进而通过改进的PSO算法对BP神经网络进行优化，构建了基于改进的PSO-BP神经网络机器学习回弹预测和补偿模型。以多种规格的铝合金数控弯管构件为对象，将实际生产中不同规格、批次、成形参数下回弹数据作为训练样本，实现了所建机器学习预测模型的应用验证。所建模型获得的预测结果平均相对误差为6.3%，与未优化的BP神经网络等传统模型相比，预测精度最大提高了18.5%，计算时间可从1.5 h缩短至300 s，同时实现了回弹预测与补偿精度以及计算效率的显著提高。

Effective Prediction and Compensation of Springbacks for Tube Bending Using Machine Learning Approach

The machine learning algorithm modeling was adopted based on the optimized back propagation(BP) neural network and the precise prediction and efficient control method of bend springbacks was proposed. In this method, the particle swarm optimization(PSO) algorithm was improved by introducing the nonlinear inertia weight and hybrid operator of genetic algorithm, and then the BP neural network was optimized by the improved PSO algorithm, and the machine learning springback prediction and compensation model was constructed based on the improved PSO-BP neural network. Based on the springback data of different specifications, batches，and forming parameters in the actual productions, the applications of the machine learning prediction model were verified. The average relative error of the prediction results obtained by the model is as 6.3%. Compared with the traditional models, the prediction accuracy is increased by 18.5% at most, and the calculation time may be reduced from 1.5 h to 300 s. The prediction and compensation accuracy of springback and the calculation efficiency are improved significantly.