基于改进果蝇优化算法的铣削力信号特征选择方法

A feature selection method for the milling force signal based on the improved Fruit Fly Optimization Algorithm

特征选择是刀具模式识别的关键问题之一。采用果蝇优化算法(FOA)将铣削力特征选择转换成果蝇寻优过程,得到了一种可用于铣刀磨损状态识别的适应度强的特征选择方法。该方法用力传感器提取铣削力信号,把特征选择过程模拟成果蝇觅食行为,采用Fisher辨别率作为特征寻优标准,将优选后的特征集输入BP神经网络,刀具磨损量为输出。实验证明,该方法易调节,寻优效果好,适应度强,BP神经网络表现好,可以快速有效地对铣削加工过程中的力信号特征进行选择。

Feature selection is one of the key processes in pattern recognition. To solve the problem of identification of tool wear condition,a feature selection method based on the improved Fruit Fly Optimization Algorithm was proposed.Feature selection of cutting force was converted to food finding process of the fruit fly. The experiment was conducted on a Makino CNC milling machine equipped with： milling cutter,EGD440R; and insert material was A30 N. Cutting force was extracted using Kistler 9257 B three-phase dynamometer,analyzed by wavelet packet theory to reduce noise and extract the energy feature of the signal as a basis for feature selection. Then,an improved Fruit Fly Optimization Algorithm was established,in which Fisher discrimination was chosen as optimization criteria. The optimal feature subset was put into a BP neural network,which output the flank wear. The result of experiment indicates that the parameter of the model is easy to adjust,has good optimization result. As shown in Table1,the BP network performance has ample potential for cutting feature selection.