基于BP和RBF神经网络的表面质量预测研究

目的 研究RBF和BP神经网络在铣削加工中的作用，实现对铣削加工质量的预测，改善铣削性能。方法 对环形铣刀与常用的球形铣刀进行对比，然后基于MATLAB平台，建立以铣削速度、进给量和铣削深度为输入参数，表面粗糙度为输出参数的RBF神经网络模型。通过大量的试验数据对RBF神经网络模型进行训练，然后再用训练好的RBF神经网络模型预测表面粗糙度，将预测值与实测值进行比较，验证RBF神经网络的预测性能。将训练好的BP神经网络模型与RBF神经网络所建模型的预测结果进行比较。结果 发现用RBF方法预测的表面粗糙度相对误差的绝对值不超过6%，最大误差为0.056 098，平均误差为0.022 277，而BP方法的最大误差为0.074 947，平均误差为0.036 578。结论 环形铣刀加工质量更好。RBF神经网络的预测精度较高，具有比BP神经网络更优的预测能力，且拥有建模时间短、收敛速度高、训练过程稳定以及学习速度快等优点，能有效进行铣削质量预测。

Surface Quality Prediction Based on BP and RBF Neural Networks

The work aims to study the role of RBF and BP neural networks in milling, so as to realize the prediction of milling quality and improve the milling performance. Firstly, the circular milling cutter was compared with the commonly used spherical milling cutter, and then based on the MATLAB platform, an RBF neural network model was established with the milling speed, feed and milling depth as input parameters and surface roughness as output parameters. The RBF neural network model was trained through a large amount of experimental data, and then the trained RBF neural network model was used to predict the surface roughness, and the predicted value was compared with the measured value to verify the prediction performance of the RBF neural network. After the BP neural network model was trained in the same way, the prediction results of the model established with the RBF neural network were compared. The absolute value of the surface roughness relative error by the RBF method did not exceed 6%, the maximum error was 0.056 098, the average error was 0.022 277, while the maximum error of the BP method was 0.074 947, and the average error was 0.036 578. The processing quality of circular milling cutter is better. RBF neural network model has better prediction ability than BP neural network, and has shorter modeling time, higher convergence speed, stable training process and faster learning speed, and can effectively predict the milling quality.