Study On Compound Fault Diagnosis Method Of Rolling Bearing Based On Blind Source Separation

As an important mechanical parts, rolling bearing has been widely used in mechanical equipment,which is used to bear and deliver load. The operation state of rolling bearing will directly affect the equipment overall performance, once the bearing fails, the fault may result in the damage of equipment, even cause major catastrophic incidents. Therefore, it is very important to diagnose the fault of rolling bearing.In the actual production environment, a fault of rolling bearing parts often comes with the fault of other parts, which namely compound fault of rolling bearing. In this case, multiple fault source and noise signals mixed with each other, so, the vibration signals are complex, and the diagnosis of fault is particularly difficult. Blind source separation is an effective method to solve the problem of compound fault signal separation, therefore, in this thesis, the research object is the vibration signals of rolling bearing compound fault,and combining with the blind source separation theory as well as the method of time-frequency analysis and pattern recognition, research is mainly focused on the problem of source signals separation under the normal and extreme conditions, the problem of feature extraction and compound fault diagnosis. The main research contents of this thesis are as follows:The basic theory of blind source separation is studied in this thesis, then this thesis analyses the effectiveness of several classic blind source separation algorithm. On this basis, JADE algorithm is applied to separate rolling bearing compound fault signals.The fault signal of rolling bearing is non-stationary and nonlinear. However, traditional time domain and frequency domain analysis method is difficult to give consideration to the time-varying characteristics of non-stationary signals, so fault characteristics of rolling bearing can not be accurately characterized. Aiming at this problem, a novel method to extract the sample entropy and energy ratio based on time-frequency method of empirical mode decomposition(EMD) is presented, which can reveal the feature information of faults signals more comprehensively and more accurately. In addition, a diagnosis method combining blind source separation,feature extraction and support vector machine is presented in this thesis.Traditional blind source separation method based on the assumption that the number of observation signals is not less than the number of source signals, which cannot adapt to the separation of fault signals under the condition of single channel. Thus variational mode decomposition method is introduced to the field of blind source separation, and extreme underdetermined problem is converted into a well-posed or overdetermined problem by the method of variational mode decomposition, which provides an effective solution for the separation of fault signals under the condition of single channel. Experimental results show that the single channel blind source separation method based on the VMD has more advantages compared with traditional methods.