Research On Ensemble Fault Diagnosis Methods Oriented To Motor Bearing

Motor bearing is one of the most widely used in motor parts.When the rolling bearing has a failure,it can damage motor lightly,cause huge economic loss in worse and even endanger personal safety.Therefore,how to detect and positioning bearing failures in time becomes an increasingly important issue.The motor bearing fault diagnosis method of research mainly in two aspects at present: one is how to extract the motor bearing fault characteristic effectively;The second is how to troubleshoot effectively.Motor bearing working environment is relatively poor,the vibration signal contains many noise,often include the non-gaussian and nonlinear noise,the key is extracting the fault feature in multifarious signal effectively.With this situation,the collection of ensemble empirical mode decomposition(EEMD)was combined with energy entropy to extract the fault information effectively in this paper,which can be used in extract the characteristics of non-gaussian and nonlinear fault vibration signal.Then the energy entropy was used in extracted the energy of fault characteristics.The bearing fault feature can be effectively extracted through this combination method.Neural network is widely applied in many fields such as pattern recognition and classification.BP network is one of the most classic method,but conventional single BP neural network has its own limitation.An new model of the motor bearing fault diagnosis,which combined Improved Gravitational Search Algorithm(IGSA)combined BP neural network,is proposed in this paper.Firstly,because standard GSA can be involved into the local optimum in solving optimization problem and optimize the accuracy is not high,an IGSA algorithm was proposed based on time-varying weight and boundary mutation.And t has been verified by the simulation that IGSA has better optimization effect.Then,the optimized BP neural network by IGSA was adopted in motor bearing fault diagnosis.The experimental results show that the EEMD-IGSA-BP neural network model has higher diagnosis optimization effect than previous method,which proved the feasibility and effectiveness of the proposed method.