Research Of Multi-fault Diagnosis Metheds Of Rotating Machinery Based On Constrained ICA

It’s very necessary and important to monitor the condition of rotating machinery and conduct real-time fault diagnosis, along with the requirement of reliability of some rotating machinery on vital position increasingly high. Rotating machinery fault diagnosis is mostly based on vibration signals of equipment, after some signal processing, obtaining interesting signal. During the operation process of the mechanical system, vibration signals from different sources are mixed together generating the sensor observed signals. It’s practical significant to seek for some methods to separate or extract the signals representing fault information from the observed signals.This paper starts from the point of rotating machinery multi-fault, researches the mixed mode of the rotating machinery vibration signals, and establishes the mixed fault simulation signals based on gear meshing model and bearing fault response model. We use independent component analysis(ICA) to diagnose the simulation signals, separate every fault source, and analyze the deficiency and limitation of ICA through the separated result.Aiming at the problem of the ICA method, we consider the spatial information between fault components with sensors in mechanic system and the prior information about frequency characteristics of fault signals, integrate different constrained condition into classic independent component analysis method, generate the extracted methods of the temporal constrained ICA and the spatial constrained ICA. We import the spatial constraint into ICA considering, then propose the spatially constrained independent component analysis(SCICA) method. We emphasize the acquisition and processing process of spatial constraints, and find general spatial constrained vectors. Then we use frequency characteristics of fault signals as temporal constraints, introduce the method of independent component analysis with reference(ICA-RT), converting the temporal constraints into temporal reference signals. At the same time, we also use spatial constrained vectors to generate the reference signals, propose the independent component analysis with spatial reference(ICA-RS) method.We test different kinds of multi-fault signal on the platform of gearbox dynamic system, using the separated method of ICA and the extracted methods of SCICA, ICA-RS and ICA-RT to diagnose the fault signals, conduct comparative analysis of effect between these methods, finding that the constrained ICA methods can extract the fault sources effectively and the stability and accuracy of the spatial constrained ICA methods’ diagnosis results is more outstanding.