基于神经网络的机械加工工序能耗预测

摘要

针对工序级能耗难以用数学方法精确估算的问题,提出了一个基于神经网络的机械加工工序能耗预测方法.给出了输入变量及输出变量的选取及其归一化处理方法,进行了隐含层节点数和传递函数的选取.以各切削用量组合及其对应能源消耗的历史数据作为神经网络训练的样本集,建立切削用量组合方案输入和能源消耗输出间的非线性关系,从而对新的切削用量参数组合进行能耗值的预测.以某企业导叶片的粗铣加工为例,验证了该能耗预测方法的有效性.%To solve the problem that the energy consumption result of machining procedure can not be calculated accurately by mathematical method, the estimation method of energy consumption in machining procedure layer based on neural network is presented. The choice and unitary of input variables and output variables are illustrated. The node number of latent layer and transfer function are selected. The combination of cutting parameter and the corresponding history data of energy consumption are served as training data set, and the nonlinear mapping relation of the combination of cutting parameter with corresponding energy consumption is set up. Thus, energy consumption of the new combination of cutting parameter is estimated. The effectiveness of the prediction method is verified by the example of rough machining procedure of a guide blade in an enterprise.