Research on diagnosis method of heavypressure and high-speed dynamic sealing fault based on layer support vector machine

A new method of the monitoring and diagnosis of the heavy-pressure and high-speed dynamic seals was presented. The transient changes of sealing operating state are reflected by vibration displacement signals of main shaft which could be utilized for monitoring and diagnosis , and at the same time the general method of the Ant Colony Algorithm (ACA) and Layer Support Vector Machine(LSVM) was proposed to research seals fault features extraction and on-line diagnosis. Based on the analysis of fault mechanism and fault pattern of dynamic seals, the kernel principal component analysis (KPCA) method was firstly applied to extract the features, which were used as the input of Support Vector Machine (SVM) classifier, from normal and fault samples. ACA was applied to optimize the compromise parameters of kernel function. Furthermore, thinking of the integrated character of the kernel principal component and kernel function, a hybrid algorithm, an ACA-LSVM model, was proposed in order to dignose the seals fault effectively. Finally, an example of the sealing monitoring and diagnosis was introduced in the test system of the heavy-pressure and high-speed dynamic sealing, and the results were proved the validity of this proposed algorithm, which gives a new way for designing and manufacturing the dynamic seals in the high-power driving devices.