Study On Information Entropy Methods And Its Application On Fault Diagnosis

With the rapid development of automation in industrial field, the structure of mechanical system become more and more complex, and the requirement to fault diagnosis is getting much higher. Thus, feature extracting becomes one of the key factors of system fault diagnosis. In this article, theories on information entropy and feature extracting methods are researched into aiming at the difficult problem of feature extracting from complex mechanical signal.Firstly, statuses of domestic and foreign research on information entropy theory in fault diagnosis field are analyzed based on opinions form lots of reference materials. Muti-resolution singular spectrum entropy module and algorithm are proposed based on wavelet analysis theory and singular spectrum analysis theory. Methods on optimized parameters in Muti-resolution singular spectrum entropy are discussed.Secondly, a new information probability and its algorithm are proposed to describe the energy distribution. Based on this, complexity information entropy module is proposed to measure the complexity feature of signal, due to the disadvantages of traditional information entropy theory in abrupt information detecting. Further more, to extract feature from coupling signal in complex system, information transfer index analysis methods are adopted, and a new information transfer index and its algorithm based on new information probability are proposed. The new index is used in describing information coupling feature between signals. Methods in this article are proved feasible to be used in signal feature extracting by theoretical analysis and comparing experimentation.At last, an experiment is adopted on rotating device. Feature of vibrating signals under crack shaft, rotor locally collide and bearing loosen conditions, are researched and described using complexity information entropy and information transfer index. These states are well recognized. It is proved that methods in thisarticle are usable in fault diagnosis system.