基于BP神经网络的滚刀工艺参数预测

Prediction of Hob Process Parameters Based on BP Neural Network

为了保证滚刀加工质量的一致性,缩短滚刀工艺文件的制定周期,在对滚刀的粗加工工艺进行研究后,采用机器学习方法,将滚刀的几何特征参数作为反向传播(BP)神经网络的输入变量,滚刀粗加工中每个工序的工艺参数作为输出结果,对滚刀粗加工过程中每个工序的工艺参数进行预测。针对传统BP神经网络最速下降法收敛速度慢的问题,在研究了“锯齿现象”产生的原因后,提出了一种“修正下降方向”的反向传播神经网络算法。仿真结果说明,与传统BP神经网络相比,同等条件下,改进的BP神经网络收敛速度加快,预测结果可靠。

In order to ensure the quality consistency of the hob process,shorten the cycle of the hob process file formulation.Based on the hob roughing process studying,the machine learning method is used in this paper,which takes the geometric characteristic parameters of the hob as the input variable of the backpropagation(BP) neural network,and the process parameters of each operation in the roughing of the hob as the output results.The process parameters of each operation in the roughing of the hob are predicted.Aiming at the problem of slow convergence speed of the fastest descent method of traditional BP neural networks,after studying the causes of "saw-tooth phenomenon",a backpropagation neural network algorithm of "correcting the descent direction" is proposed in this paper.The simulation results show:compared with the traditional BP neural network,the improved BP neural network converges faster under the same conditions,and its prediction results are reliable.