Abstract: Considering the limitations of the available methods and the features of parts coupling and fault concurrent, a method for spindle lubrication fault prediction based on fault symptom decision model and dynamic confidence matching is proposed, to improve the ability of lubrication fault prediction and predictive maintenance on spindle transmission system of a computer numerical control (CNC) machine tool. Fault symptom parameters of spindle are reduced first according to correlation degree and fault symptom condition sequences of parameters are defined on basis of historical fault data set. Fault symptom decision models are built based on wavelet analysis and probabilistic neural network technologies, which are used to identify real-time condition of fault symptom parameters. Dynamic confidence matching algorithm is designed and multiparameter matching results with fault symptom condition sequences are fused from reliability and accuracy point of view. On this basis, fault probability and occurrence time can be predicted online. Experimental results show that the method can accurately predict spindle lubrication fault.

Key words: CNC machine tool, Dynamic confidence matching, Fault prediction, Fault symptom decision model, Spindle lubrication