Research Of Extrusion Granulating Machine Fault Diagnosis Based On Principal Component Analysis

It can be proved that application of fault diagnosis bring more economic benefit. At present, there are many technology of diagnosis, such as time domain analysis, frequency domain analysis and time-frequency analysis. However, the signal measured in factory is more complex than the one researched and tested. This brings inconvenience when the signal is studied.The research object of the paper is an extrusion granulating machine in No.l factory, Beijing Yanshan Co. It is enormous, and has many vibration sources. Its technique is complicated. Because the sensor is absorbed by magnetic force on the shell when the signal is measured, the signal can be taken for nonstationary signal. Therefore, the vibration signal is interfered and the redundancy information is much.According to the characteristic of research object, four parts content are integrated in this paper, which is about signal measure and process, feature extraction, state recognition and diagnosis decision. Then an integrated diagnosis and analysis method is advanced. And it is applied to the diagnosis project of machine. The integrated method contains Local-Wave Decomposition theory, Principal Component Analysis theory and Time Series Analysis theory.The course is as the following:First, decompose signal based on local wave theory. Second, Analyze based on principal component analysis theory to obtain some principal component. Finally, modeling and draw AR spectrum and find the fault information. The local wave theory has more applicability than other theory. Principal component analysis can decrease interference, emphasize the attributive information, reduce the spatial dimension and keep the important information. The AR spectrum is smoothness and can give exact characteristic frequency.It can be proved by theory and practical diagnosis. The method is applicable for nonsteady state signal and signal interfered. The method is simple, effective and exact. It can be applied to practice diagnosis project widely...