Research And Hardware Implementation Of Rolling Bearing Fault Diagnosis Based On HMM

Rolling bearing is a key component in rotating machinery.In order to ensure that the machinery and equipment can work reliably and efficiently under long and heavy loads,the research on the condition monitoring and fault diagnosis of rolling bearings has great significance.Therefore,this thesis studies the fault diagnosis method of rolling bearing,focusing on the analysis of fault feature extraction,fault feature reduction and fault diagnosis,and puts forward an effective and complete fault diagnosis method for rolling bearings.On this basis,the fault diagnosis method of rolling bearing is implemented on the heterogeneous So C hardware platform,and the design of computing architecture is studied.The feasibility and superiority of the method in the implementation of high performance computing on the heterogeneous So C platform are verified.The specific research contents are as follows:After comparing and analyzing several fault features extraction methods of rolling bearings,the advanced variable mode decomposition(VMD)is used to extract the complex fault characteristics.On this basis,the singular value decomposition(SVD)is applied to the reduction of primary fault features.In the case of ensuring that the fault feature information is valid,the calculation amount of subsequent model input training is greatly simplified.For the fault diagnosis method,the Hidden Markov Model(HMM),which is widely used in the field of natural language processing,is applied to the field of rolling bearing fault diagnosis.After discussing and analyzing the relevant concepts and basic algorithms of HMM,combined with the characteristics of actual signals,the Gaussian Mixed Model(GMM)is integrated into the Hidden Markov Model.In this way,a rolling bearing fault model based on GMM-HMM is trained for the extracted fault characteristics.The fault diagnosis method proposed in this thesis is tested and verified by actual data of different rolling bearings.The experimental results show that the feature extraction method combined with VMD and SVD can provide effective training feature samples for GMM-HMM.With the GMM-HMM model,the training of the rolling bearing failure model can be completed quickly.The accurate identification of the running state of the rolling bearing can be realized,and the rationality and high efficiency of the fault diagnosis can be ensured.The fault diagnosis method of rolling bearing is implemented on the Zynq-7000 based heterogeneous So C hardware platform,and the design of rolling bearing fault diagnosis system is carried out in combination with the development of Lab VIEW interface.Based on this,combining the architecture of heterogeneous So C platform,the rolling bearing fault diagnosis method is divided into reasonable hardware and software partitions,and a suitable data path for hardware acceleration is designed.The reliability and rationality of the hardware acceleration of the rolling bearing fault diagnosis method on the heterogeneous So C platform are verified.At the same time,it also provides a design reference for the complex algorithm to implement the high performance calculation on the embedded platform.