基于集成神经网络的刀具磨损量监测

提出了一种基于集成神经网络识别铣刀磨损量的监测方法．利用小波包变换将切削力和振动信号分解为不同频带的时间序列，从每个信号中选择与刀具磨损状态最相关的3组频段的均方根作为监测特征；通过信号的组合和不同子网络输出决策间的融合，集成神经网络输出刀具磨损的识别结果．试验和仿真分析表明，此方法能够满足刀具磨损量实时监测的要求．

Tool Wear Monitoring Based on Integrated Neural Networks

A tool wear condition monitoring approach based on integrated neural networks was proposed to recognize and predict tool wear conditions in milling operations. In this approach, vibration and cutting force signals are decomposed into time sequences in different frequency bands by wavelet packet transform, and the root mean square values of each signal in three frequency bands, extracted from decomposed signals, with a close relation to wear conditions are selected as monitoring features. The final recognition results of tool wear are given by the integrated neural networks through the combination of signals and the decision fusion of different subnets. Experiments and simulations show that the proposed approach can meet the requirements of on-line monitoring of tool wear conditions.