Rolling Bearing Fault Diagnosis Based On Blind Source Separation And Multi-scale Entropy

As one of the most important parts of most mechanical equipment,the operating condition of rolling bearings is related to whether the mechanical equipment is normal or not.The traditional fault diagnosis methods of rolling bearings often ignore the fact that the vibration signal collected by the sensor is mixed by multiple sources,and the non-stationary vibration signal is directly processed by fast Fourier transform(FFT)which is suitable for stationary signal analysis,and it is difficult to analyze the fault types of source signal comprehensively and accurately by the traditional methods.In this thesis a new fault diagnosis method of rolling bearing based on blind source separa tion and multi-scale entropy is proposed to address the issues that exist in the traditional rolling bearing fault diagnosis.As a kind of precision component,when one component of a bearing is in abnormal state,the other parts of the bearing often produce a chain reaction,and the vibration data collected by the sensors are often the superposition of abnormal vibration data of the components.In order to identify the abnormal situat ion more accurately,a separation method of single channel vibration signal based on blind source separation is proposed.In this method,the underdetermined blind source separation problem is converted to a positive definite blind source separation problem by the Extreme-point Symmetric Mode Decomposition Method(ESMD),and then a blind source separation method based on time-frequency analysis is used to separate the mixed signal.The simulation results show that the correlation coefficients between the separated source signals and the actual source signals can reach 0.9771,0.9784 and 0.9660 respectively,and the method can separate the multi-source mixed signals with a high separation precision.The empirical mode decomposition(EMD)and multi-scale entropy method are used to extract the characteristic values of the separated signals.In practice,the empirical mode decomposition method is often affected by the end effect.End effect reduction based on waveform averaging is proposed to address the above issue.This method which extends the signal according to its characteristics has good adaptability,and can restrain the end effect of empirical mode decomposition to an acceptable degree.Finally,BP neural network is used to identify the fault types.The experimental results show that the identification rate of the inner ring fault,outer ring fault and normal state can reach 97%,86% and 90% respectively by the method proposed in this thesis,and this method can effectively identify the fault type of r olling bearing to a certain extent.Hybrid programming with C# and MATLAB technique is used to develop fault diagnosis and analysis software for rolling bearings.Through the analysis of the vibration signal of actual rolling bearing,the validity of the method is verified in practical application.