Study On Fault Feature Extraction And Diagnosis Of The Rolling Bearing

Rotary machine is widely applied into the area of machinery manufacturing、electric power、motive power、and even the aerospace and the military department ect.Once some sections of the operation system are failure,which are being closely linked,it will directly influence the operation of the whole system,so it is particularly important for monitoring and diagnosis of rotating machinery.Rolling bearing,which is the main part of he rotating machinery,plays important role in our daily life and production.The monitoring and diagnosis of rolling bearing is of practical significance.The paper is written for the fault diagnosis through analyzing the vibration signal of rolling bearings. Feature extraction as the key link for fault diagnosis is the main content of the study The methods of feature extraction,including high order statistics, the fractional lower order statistics, the fractional order permutation entropy algorithm based on modern signal processing,are also studied in this paper.Facing with the difficulties in getting sample and the lack of fault sample in the mechanical fault diagnosis, support vector machine is used as a fault identification method,highlighting its advantages in terms of classification of small samples. This paper puts forward some new method of fault feature extraction and diagnosis of rolling bearing, the main research contents are as follows:(1) Introduced the concept of alpha stable distribution and probability density function, through the bearing fault signal probability density fitting, prove that the bearing fault signal more in line with stable distribution, and the relevant feature parameters reflecting the information of characteristics can be extracted by basing on it, decomposition theory and characteristic parameters of form feature vectors by using wave let packet, and then use the support vector machines for fault classification.(2) Detailed the high order statistics theory and its properties, taking into account the three order cumulant higher-order statistics theory and four order cumulant,characteristic index theory of the fractional lower order statistics and the dispersion coefficient, feature extraction method is presented in different order,forming four-dimensional features to a volume group, and finally getting the feature vector group income as the basis, using support vector machine vibration state of a mechanical system to make a judgment.(3) Explained the principle of permutation entropy algorithm, simulation results verify the effectiveness of the permutation entropy algorithm; based on permutation entropy was proposed on a new feature extraction method, namely fractional order permutation entropy, the simulation proved that the fractional order permutation entropy is more sensitive to mutation, significantly enlarged the tiny change of time sequence; select the permutation entropy of different orders as feature vector, and provides a new way for the fault diagnosis feature extraction, using support vector machine to complete the classification of different work conditions.