基于模糊神经网络的刀具磨损识别

为了实现刀具磨损状态的自动识别，采用机床功率法进行了刀具自然磨损和不同切削参数（切削速度、进给量和切削深度）对功率信号影响的实验。在此基础上，建立了功率信号的时序AR模型。在提取作为刀具磨损特征量的AR模型参数时，考虑了切削用量对模型参数的影响，提出了特征量选取的准则，使所提取的特征量更加实用化，通过具体自学习和良好函数逼近能力的神经网络获得了特征量对刀具状态的隶属函数，并利用模糊神经网Fuzzy ART实现了刀具磨损状态的自动识别，识别正确率为95%，说明所提出的方法是有效可行的。

Tool Wear Monitoring Using Fuzzy Neural Network

Tool condition monitoring is important in achieving advanced manufacturing systems. The fuzzy neural network fuzzy ART was applied to realize the tool wear monitoring. Experiments in turning were implemented to obtain the motor power signals under normal tool wear and different cutting parameter (cutting velocity, feed velocity and cutting depth) conditions. The time series AR model of the signals was established. The influence of cutting parameters was considered when the features sensitive to tool flank wear are extracted from the time series AR model. The principle of feature extraction was proposed. Feed forward neural network was employed to obtain the fuzzy member function of the features. Tool wear condition was recognized by Fuzzy ART. The results show that the proposed method is feasible.