摘要
A principal component analysis-cerebellar model articulation controller （PCA-CMAC） model is proposed for machine performance degradation assessment.PCA is used to feature selection,which eliminates the redundant information among the features from the sensor signals and reduces the dimension of the input to CMAC.CMAC is used to assess degradation states quantitatively based on its local generalization ability.The implementation of the model is presented and the model is applied in a drilling machine to assess the states of the cutting tool. The results show that the model can assess the wear states quantitatively based on the normal state of the cutting tool.The influence of the quantization parameter g and the generalization parameter r in the CMAC model on the assessment results is analyzed.If g is larger,the generalization ability is better,but the difference of degradation states is not obvious.If r is smaller,the different states are distinct,but memory requirements for storing the weights are larger.The principle for selecting two parameters is that the memory storing the weights should be small while the degradation states should be easily distinguished.
提出了一种用于设备性能退化评估的PCA-CMAC(主成分分析小脑模型节点控制器)模型.该模型利用PCA进行特征提取,去除多个传感器信号特征的冗余信息,并且减少CMAC的输入维数;利用CMAC的局部泛化能力定量地评估设备的性能退化.给出了模型的实现过程,并将模型应用于钻削过程刀具状态的评估,试验结果证明该模型能基于刀具的正常状态,对刀具的磨损状态进行定量的评估.分析了CMAC中泛化参数g和量化参数r对评估结果的影响,g越大,CMAC的泛化能力越好,但各退化状态之间的区别越不明显;r越小,各退化状态之间越容易区分,但所需的权存储空间越大.2个参数的基本选择原则是CMAC的权存储空间应尽量小,与此同时,各退化状态之间应容易区分.
基金
The National Natural Science Foundation of China(No.60443007,50390063).
