Research On Adaptive Method For Early Fault Diagnosis Of Rolling Bearing

Rolling bearing is widely used in rotating machinery.It is one of the most easily damaged parts.Once the fault occurs,it will seriously affect the performance of the whole equipment,so the fault diagnosis of bearing has important economic value.The working condition of the rolling bearing is complex,and its early weak fault signal is easily covered by the strong noise,which makes it difficult to extract the fault feature.At present,the commonly used methods of fault diagnosis mainly focus on noise reduction and fault feature extraction,which have achieved remarkable results.Based on this,EEMD and morphological filtering are used to deal with the signal.The key technology of EEMD is to select the useful IMF components for signal reconstruction.In this paper,according to correlated kurtosis choose the IMF components.Morphological filtering can effectively filter the noise in the signal,but its effect mostly depends on the length of the structure elements.The choice of the length of the traditional morphological filter structure elements is random and empirical,and sometimes it is difficult to obtain satisfactory results.Therefore,based on the average filter,using the global search ability of genetic algorithm to select the best length of structure element,which the correlated kurtosis of filtered signal is maximum.An adaptive morphological mean filter is established.Through simulation and experimental study,the effectiveness of the method is demonstrated.For those weak fault signal,which is seriously covered by the strong noise,just using one kind of signal processing method,the diagnosis effect is not ideal.Therefore,this paper combines EEMD and adaptive morphological filtering to give full play to their respective characteristics in order to improve the accuracy of fault diagnosis of rolling bearing.Because the morphological difference filter owns the good ability of feature extraction,an adaptive morphological difference filter based on genetic algorithm is established.At the same time,using kurtosis of auto-correlation coefficient as the quantitative evaluation criteria to reflect the contentof noise in the signal.Combining kurtosis of auto-correlation coefficient and kurtosis,together to illustrate this method not only can highlight the fault information,but also can effectively suppress the noise.The actual fault data of rolling bearing shows that this method has the ability of accurate fault diagnosis.