| In modern industry,mechanical equipment becomes more and more efficient,pre-cise and intelligent,and is closely connected with other equipment in the factory pro-duction.In order to ensure the normal operation of the mechanical equipment and prevent the equipment from failure effectively,it is necessary to conduct condition monitoring and fault diagnosis research on mechanical equipment.Make reasonable arrangement of repair or maintenance plan according to the result of health condition to reduce maintenance cost and maximize production.As one of the core components in modern machinery and equipment,rolling bearing is also one of the parts most prone to failure.Whether the rolling bearing is in a healthy working state will directly affect the normal operation of mechanical equipment,and even affect the safety of the whole production line.In this study,rolling bearing is taken as the research object.And a se-ries of research work is carried out in combination with the vibration analysis method for feature extraction,which is the crucial task in health condition monitoring and fault diagnosis.This study first expounds the purpose and significance of the study on rolling bear-ing health condition monitoring and fault diagnosis,and summarizes its development process.Nextly,the basic knowledge of rolling bearing is introduced,including the ba-sic structure and the main failure forms.In addition,the key technologies and existing scientific problems in this subject are summarized.The related works of three aspects including information acquisition,feature extraction and pattern recognition method at home and abroad are introduced.On this basis,the main ideas and contents of this study are established.For condition monitoring,this paper proposes a feature extraction method based on internal correlation information,which uses the temporal image description of the vi-.bration signal to realize machine condition monitoring.Firstly,the minimum distance optimization matching method based on dynamic time warping(DTW)is used to ob-tain the period of vibration signal.Then the raw vibration signal is divided into several sub-sequences by the length of one or several cycles.Then,the signal reconstruction method is used to extract the interrelated information among the data in the segment,and the time sequence image that can describe the dynamic health condition of the rolling bearing is constructed.Then the grayscale image is analyzed and processed.For the obtained image feature,the statistical distance is employed to measure the similarity of images,and the hypothesis test based on Gaussian distribution is used to make decisions on potential change points.Finally,a general application of the typical application is explored.For the fault diagnosis problem,a novel fault feature extraction method based on local temporal self-similarity is proposed,which is another expression of internal cor-relation information of vibration signal.It is a hybrid approach that combines hand-crafted feature design with automatic feature learning,which takes into account the local characteristics and variation trend of the signal.Firstly,some points in the neigh-borhood of the signal are collected to construct a local temporal self-similarity ma-trix,and the block-based descriptors are used to count the gradient information of the matrix.The standard bag-of-word framework is employed to assemble these local descriptors.Then the unsupervised machine learning algorithm is adopted to learn the codebook composed of codewords.The vibration signals are represented by his-tograms of learned codewords.Finally,two common machine learning methods,i.e.,k-nearest neighbor and support vector machine,are employed to classify faults.The effectiveness of the proposed fault diagnosis algorithm based on local temporal self-similarity is verified by the case study of bearing fault diagnosis.Finally,we summarize the main content and innovation points,and discuss the shortcomings of the paper and the future work. |
PDF Full Text Request