Research On Health Condition Assessment Method Of A Rolling Bearing Under Varying Loads Based On Unsupervised Domain Adaptation

Rolling bearing is one of the key component of rotating machinery,and its health condition is closely related to the safety and reliability of rotating machinery.Rolling bearing often works under varying loads condition,and the change of load will directly affect the vibration feature.Therefore,under varying loads condition,fault diagnosis and health condition assessment of rolling bearing is of great significance for ensuring the running operation of machinery and preventing major accidents.The dissertation takes the rolling bearing condition assessment method under varying load condition as the center line.and the research is expanded from several aspects of feature extraction,domain adaptation method,the establishment of fault diagnosis model and health state assessment model.For feature extraction,aiming at the problem of single-domain features cannot comprehensively reflect the degree of failure of rolling bearings,the time-domain and frequency-domain features are extracted for rolling bearing vibration signals,and the time-frequency features are extracted by combining variational mode decomposition and singular value decomposition to construct a multi-domain feature set.For domain adaptation method and the establishment of fault diagnosis model,the application of domain adaptation method in rolling bearing fault diagnosis is studied deeply.Aiming at the subspace alignment(SA)method is not suitable for dealing with non-linear feature,the dissertation improves the SA method.It is proposed to combine the kernel mapping method with SA method to align the features of the source domain to the features of the target domain under different loads.Aiming at the kernel mapping SA method can only complete the domain adaptation from the perspective of maintaining the same feature,it is proposed to combine the kernel mapping SA and balanced distribution adaptation,while maintaining the same feature,the probability distribution difference of data in different domains is further reduced to obtain better domain adaptation effects.Combining the data processed by the two domain adaptation methods with the support vector machine,the diagnosis of different fault states of rolling bearings under varying loads is achieved.The comparative experiments prove the effectiveness of the improved algorithm in solving varying loads problems.For health condition assessment,under the lack of prior knowledge,it is difficult to fully describe the spatial distribution of feature samples with different degrees of failure based on a single state model,a multi-state model evaluation method is proposed.By measuring the membership of feature samples to the hyperspheres of sample with different degrees of failure,the change process of the feature samples of each failure degree in the degradation process is better expressed,and the health condition of rolling bearing is effectively assessed.