Investigation On Bearing Fault Diagnosis And Life Prediction Based On JADE

Rotating equipment plays an important role in the field of modern industry, and rolling bearing is a main component of rotating equipment, so the running state of rolling bearings is important for the stable movement of the rotating machinery. For the rolling bearings are always running in high speed, heave load or in bad condition, they get failure easily. The faults happened on bearings are often complicated, for a long time, the fault diagnosis of the rolling bearing used in rotating machinery is the priority and difficulty in the industry research. Understand the running condition and health status of bearings accurately before the equipment is in failure, so that can take effective measures to prevents the industrial accident and reduce the loss in the accident. Once the fault has occurred, we should also identify the fault and solve it. Analyze bearing signals that collected in the experiment and extract the effective feature of bearing signals is basis of bearing diagnosis. Bearing status signal will change while the fault happened. We can get the fault information after analysis the signal we have collected. Researchers have put forward many signal processing methods to extract the effective characteristics of the bearing. This paper combined use the Joint Approximate Diagonalization of Eigen-matrices method, EMD decomposition and spectral correlation methods to discuss the bearing state feature extraction methods, and further investigation on bearing fault diagnosis and life prediction combining support vector machines and extreme learning machine.In the first chapter, this article discusses the basic aspects and fault diagnosis methods, and presents the development of bearing life prediction theory and the fundamental knowledge of bearing signal. Chapter Ⅱ introduced the calculation method of failure frequency calculations and the knowledge of signal acquisition, summarizes the calculation methods of different fault features. Chapter III combined use the Joint Approximate Diagonalization of Eigen-matrices method, EMD decomposition and spectral correlation methods in effective feature extraction, further combining with the SVM realized the recognition of different fault types, and get good results. In chapter IV, improved the JADE algorithm and proposed two-layer-JADE algorithm, extracted the parameter vector that can reflect the degradation state of bearings, and set up residual life prediction model based on ELM. Different bearing fault types in the whole life data are analyzed, and completing the bearing life accurate prediction. Chapter V introduced bearing fatigue test in detail. The last chapter made summary to full text.