A Fault Detection Approach For Motorized Spindle Based On Non-linear PCA

With the development and application of science and technology,excellent performance of NC machine tools is a prerequisite for high-speed processing technology.The high-speed spindle is a core component of NC machine tool,motorized spindle is the main form of high-speed spindle.The performance of the motorized spindle directly affects machining precision and product quality.The consequence of the fault are range from economic losses to casualties.The domestic and foreign research for motorized spindle fault diagnosis technology has made some achievements,but mainly focused on analysis methods of the vibration signal.Taking the motorized spindle system,as the research objective an improved nonlinear PCA method was used for fault diagnosis of motorized spindle system,the main contents are as follows:1.The structure and the fault mechanism of motorized spindle were studied,test system was established vibration signal was collected and treated by AIC9916F system.2.The basic theory and method for PCA and kernel function are studied.considering the nonlinear characteristics of the motorized spindle system,the concept of nonlinear PCA methods was introduced.Through the characteristic analysis of polynomial kernel and radial basis function,a mixed kernel function are abtained by combining polynomial kernel function and RBF kernel function,then combined with the PCA to improve kernel principal component analysis3.Faults diagnosis rate is considered as the optimization objective.1)KPCA based on the dichotomy optimizing method(dichotomy-KPCA)is provided.and is used to diagnosis the unbalance problem of motorized spindle.The faults diagnosis rate of dichotomy-KPCA is significantly higher than the other two methods,thus dichotomy-KPCA used for the fault diagnosis motorized spindle system is feasible.Though dichotomy is simple and easy to apply,it can't simultaneously optimize multiple parameters,In addition dichotomy give more errors compared with intelligent optimization algorithm.2)KPCA based on the PSO optimizing method(PSO-KPCA)is proposed,and used to diagnosis the fault of unbalance and Rub-Impact of motorized spindle.The diagnostic effect of PSO-KPCA is compared with dichotomy-KPCA method,the comparison showed that the fault diagnosis rate of PSO-KPCA was significantly higher than the other methods.PSO has some advantages compared with the dichotomy.Among is simple,and easy to apply,but fall into local optimum.3)Then KPCA based on the ABC optimizing method(ABC-KPCA)is proposed.All methods are used to diagnosis the fault of unbalance and Rub-Impact of motorized spindle.The results show:ABC could overcome local optimum and has the strongest global search ability.ABC-KPCA has a highest faults diagnosis rate,and is a ideal faults diagnosis method.