Research On Aero-Engine Bearing Fault Diagnosis Based On Texture Feature

With the rapid development of modern industry and the continuous progress of intelligent technology,in the context of supply side structural reform,industry is not only the leading force of national economy,but also the main battlefield of supply side structure reform,thus people require more on the safety and Intelligent of industrial equipment.With the "public entrepreneurship,multitude innovation" and "Internet +"swept China,which set off a wave of "Internet + manufacturing" to lead China's industrial industry into the 4.0 era,but also accelerated the transformation of China's industry from traditional industrial to intelligent industry.Therefore,it is imperative to study the status monitoring and intelligent fault diagnosis of industrial equipment.Industrial machinery system,which has numerous components and works in complex and harsh environment,is highly intelligent.In order to monitor completely the status of mechanical equipment and its components,we usually use multi-sensor,which install on multi-point,to collect information.The vibration signal collected by each sensor is usually a comprehensive response of the vibration trajectory of each component during operation,and the complexity of the actual operating conditions and the uncertain propagation path of each component also increase the complexity of the collected vibration signal.It is convenient to monitor status and diagnose fault if we can extract interested feature from vibration source signal.In this paper,we will study the vibration mechanism and signal characteristics of rolling bearing,which is one of the most common and most critical components of aero-engine,and study its one-dimensional frequency characteristics of vibration signals.Some fault diagnosis methods of aero-engine bearing based on texture feature are proposed.In the actual case,the rolling bearing is running in the strong noise background,and the collected vibration signal is a comprehensive reflection of the rolling bearing and the surrounding components and noise.Therefore,based on the failure mechanism of rolling bearings,by calculating the failure frequency of different faults and analyzing the vibration signal's frequency domain characters we can extract the rolling bearing fault information.However,the application of one-dimensional frequency feature extraction and fault diagnosis based on traditional rolling bearing is very limited in the case of strong noise background and complex fault-prone actual conditions.The feature extraction and fault diagnosis method based on the texture feature of the vibration signal have an important role in improving the recognition accuracy of the composite fault and enhancing the fault feature extraction ability of the rolling bearing in the complex working condition.The main contents of this paper are as follows:1)From the perspective of engineering application and theoretical analysis,elaborated the background of the thesis.This paper analyzes the development of fault diagnosis based on frequency feature and texture feature at home and abroad,and extracts the problems that need to be solved in the field of state monitoring and fault diagnosis by reading a lot of literature,so as to carry out research work and explain the idea of this paper and outline.2)On the basis of deep understanding of the vibration mechanism and signal characteristics of rolling bearing,the main forms and reasons of rolling bearing failure and the basic knowledge of characteristic frequency calculation are studied,which provide the theoretical basis for the follow-up study.3)Based on the feature extraction based on the traditional one-dimensional signal characteristic frequency,the new comparative Empirical Wavelet Transform and Variational Mode Decomposition are studied and introduced into the research of aero-engine rolling bearing fault diagnosis,we also optimized the VMD parameters4)Aiming at the problem that the feature extraction method based on the traditional one-dimensional signal frequency feature is limited under complex conditions and complex faults,the theory of texture feature extraction is studied,and fault diagnosis methods based on Complete Local Binary Pattern and Bag of Visual Words are proposed.