Fault Identification Of Rolling Bearing Based On Manifold Learning Research

The rolling bearing is one of the key parts in the rotating machinery and is also the most widely used part.The running state of rolling bearing directly affects the normal operation of the whole equipment.Because of its own working characteristics and working environment,rolling bearing is prone to failure.Therefore,the timely and accurate diagnosis and recognition of rolling bearing can not only make the device run stably and reliably,but also has important practical significance.Manifold learning is a new method of dimensionality reduction,which is suitable for dealing with nonlinear and non-stationary bearing vibration signals.In this paper,the phase space reconstruction technique and the LTSA algorithm in the manifold learning method are used to denoise the vibration signal of the rolling bearing;The signal after noise reduction is obtained through Fourier transform to get the characteristics of time domain and frequency domain,and the multi domain feature set is formed,and the sensitive feature is selected through the FCBF algorithm for the multi domain feature set.Finally,the fault identification model based on local linear discriminant embedding(LLDE)and support vector machine(SVM)rolling bearing is proposed,first to use the LLDE to extract the signature from the denoised signal,then after extraction of the characteristics of input into the trained SVM is used to identify the classification of rolling bearing.As the rolling bearing failure is a slight to serious development process,so the damage degree of the rolling fault is identified on the premise of determining the type of fault..Through the actual data of bearing failure,it is verified that the fault recognition model can identify the fault type and fault degree of rolling bearing well.By comparing the recognition effect of the k-nearest neighbor classifier,the recognition rate of the SVM classifier is the highest,and the validity of the model is verified.