基于遗传算法优化反向传播神经网络的激光铣削层质量预测

Quality Prediction of Laser Milling Based on Optimized Back Propagation Networks by Genetic Algorithms

为了有效地控制激光铣削层质量,建立了激光铣削层质量(铣削层深度、铣削层宽度)与铣削层参数(激光功率、扫描速度和离焦量)之间的反向传播(BP)神经网络预测模型。利用遗传算法(GA)优化了BP神经网络的权值和阈值,构建了基于遗传算法神经网络的质量预测模型。用GA-BP算法对激光铣削层质量进行了仿真预测,并将仿真结果与BP神经网络模型仿真结果进行了对比。仿真结果表明,两种网络模型的平均误差较小,网络训练后检验精度较高,说明两种网络模型用于激光铣削层质量预测是可行的,并且遗传算法优化BP神经网络能够有效地提高网络的收敛性和预测精度。

In order to control the quality of laser milling layer, back propagation （BP） neural network model of the milling laser quality including milling depth and width, and milling layer parameters including laser power, laser velocity and defocus amount is set up. The weight and threshold of the BP neural network is optimized by genetic algorithm （GA）, and a quality prediciton model is constructed based on BP neural network. The quality of the laser milling layer is forecasted by the model of GA-BP neural network. The results from BP neural network are compared with that of GA-BP neural network. The results of simulation show that the errors of the two network models are smaller, and the test accuracy are higher. Therefore, the two network models can he used to predict the quality of the laser milling. It is also shown that both the astringent and prediction accuracies of the GA optimized BP neural network are improved.