Weak Signal Extraction And Detection Of Mechanical Equipment In Heavy Noise

In actual fault diagnosis of mechanical equipment, the measured vibration signals which for bad environment, not only contain useful features but also have much noise. Especially for incipient fault of mechanical equipment, its the feature is weak and usually submerged in heavy noise, so it is hard to be extracted. Aiming at mechanical equipment, this dissertation focuses on weak signal extraction and detection in heavy noise.The dynamic signals of mechanical equipment often possess nonstationarities due to variance of operation and inherent nonlinearity of equipment. Empirical mode decomposition (EMD) method is an efficient tool for the nonlinear and nonstationary signals analysis. But EMD method have inherent drawback called end effect, and this drawback will be more serious when weak signal in heavy noise. For this prolem, this dissertation puts forward a new method whivh uses cascaded bistable stochastic resonance system (CBSRS) for noisy EMD. The successful application of fault diagnosis rolling bearing shows the method's efficiency.Singularity value decomposition (SVD) as a nonlinear filter method is widely used for signal de-noising and detection. But the traditional SVD method is usually in time domain and the singularity values are sensitive to noise, so SVD only can process signals mixed weak noise. For this reason, an improved method called frequency domain SVD is proposed. In addition, as the de-noising ability of single SVD system is limited, another novel method of cascaded SVD system (CSVD) is presented in this dissertation. Simulation experiments prove these two methods'feasibility.Weak signal detection when submerged by large amplitude ones is analysed, and two method based on different point of view are proposed. One is based on signal and noise separation of independent component analysis (SNICA) method to eliminate large amplitude signal and extract weak feature ones. The other is based on multi-taper technique in order to reduce frequency leakage and enhance weak signals, Satisfactory results have been achieved when using them respectively to an axis eccentricity fault diagnosis and the oil whirl in early stage monitoring.In order to meet the requirement of different applications and as the carrier of the key technology, an innovative portable fault diagnosis system with its advantages of high performance and low coast is presented. This system adapts dual CPU hardware structure and uses Linux as software development platform. In addition, it contains many communication interfaces to realize information transmission between portable instrument and upper condition monitoring system.