刀具状态的ANN—模糊识别

提出了利用人工神经网络（ＡＮＮ）及模糊识别理论融合多监控参数进行刀具状态识别的方法．该方法首先对各监控参数按刀具不同状态的敏感性进行分组，并利用多个ＡＮＮ子网络建立各组参数与刀具状态的模糊隶属度关系，然后利用模糊决策法对各ＡＮＮ子网络确定的刀具状态模糊隶属度进行综合评判并按最大隶属度判定刀具状态．该方法不仅具有ＡＮＮ的并行运算特点，而且具有模糊综合评判的容错性，从而提高状态识别的实时性和正确率．结合功率信号的多个特征对大量实验数据的测试表明，该方法可将ＡＮＮ的识别正确率从平均８８％提高到９５％．

ANN Fuzzy Recognition of Tool Conditions

The methodology in sensor fusion based on artificial neural networks (ANN) and fuzzy recognition theory for tool condition recognition is introduced. First, the monitoring indices are grouped according to their sensitiveness to different tool conditions. Then, several ANNs are used to model the membership function between tool conditions and each group of monitoring indices. After that, a fuzzy decision strategy is used to synthesize the fuzzy outputs of the ANNs and tool conditions are identified according to the maximum membership degree. The method has not only the parallel computation characteristics of ANN, but also the fault tolerance of fuzzy synthesis strategy. Therefore, the proposed ANN fuzzy recognition strategy can improve tool condition systems with a shorter time response and a higher rate of correct recognition. Finally, the multiple monitoring indices from cutting power signal are fused with the ANN fuzzy recognition method. The experiments show that the correct recognition rate is increased from 88%(the average ANN correct recognition rate) to 95%. The proposed method provides a basis for the implementation of ANN and fuzzy theory in practical tool condition monitoring system.