Research On Operational Modal Analysis Method Of Mechanical System Based On Single Channel Blind Source Separation

Modal analysis is of great significance to the analysis of structural dynamics,fault diagnosis,and online health monitoring.In recent years,operational modal analysis can realize modal parameter identification only by measuring the vibration response of the structure without affecting the normal work of the structure.Therefore,it has been widely used in largescale mechanical structures and civil engineering fields.Blind source separation technology has developed rapidly in the field of operating modal recognition due to its advantages of nonparameterization,simplicity and efficiency.Existing methods mostly address the problem of positive-definite blind source separation,but they often encounter under-determined problems,especially the extreme ill-conditioned situation with only one sensor-the problem of singlechannel blind source separation.This paper develops how the classic blind source separation algorithm-second-order blind identification(SOBI)applies in operational modal analysis,and combined with the virtual multi-channel method based on time-frequency decomposition,proposes an operational modal analysis method based on variational modal decomposition(VMD)and SOBI.The specific content is as follows:Firstly,the mathematical model of blind source separation,separability analysis and separation effect evaluation criteria are introduced,and the mathematical assumptions and problem solving of independent component analysis(ICA)and SOBI are explained from a mathematical point of view,and they are introduced into the technology of operational mode analysis.Aiming at the problem that the frequency ratio of each order in the modal analysis may be an integer,two simple harmonic sources are used to analyze the identifiability of the source signal of the classical blind source separation algorithm,and the physical system model is used to compare the pros and cons of ICA and SOBI applied to the operational mode analysis.The results show that the SOBI algorithm has higher recognition accuracy and better noise resistance,and is more suitable for operational modal analysis.Secondly,in view of the problem that the in-service mechanical system is often accompanied by strong background noise,the accuracy and anti-noise ability of the VMDbased virtual channel reconstruction method are studied,and a criterion for selecting the mode number is proposed.The single-channel blind source separation method based on virtual multichannel is discussed.Through simulation analysis,it is concluded that the VMD algorithm can overcome the modal aliasing problem of empirical modal decomposition methods when the signal is intermittent,and has good noise Robustness.Aiming at the problem that the number of modal components of VMD needs to be set artificially in advance,an optimized VMD method for effectively selecting the number of modal components based on the basic idea of lower energy ratio of low-order modal and stable graphs is proposed,and the calculation example verifies the effectiveness of the method.Finally,a single-channel blind source separation algorithm based on K-value optimized VMD-SOBI is proposed and applied to the modal parameter identification of rail vehicle suspension systems.First,use VMD decomposition to convert the single-channel vibration response into multi-channel signals,and then filter out the effective components as the input matrix of SOBI to obtain the modal coordinates and modal response,and finally use the singlemode recognition technology to realize the recognition of modal parameters.The simulation verifies the feasibility and noise robustness of the algorithm.This method is applied to a 1/5 scale roller drilling rig system that can simulate a rail vehicle suspension system,and the first three-order modal parameters of the 1/5 scale bogie are successfully identified,and the factors of single channel blind separation is studied by this experiment.In technology,the installation position of a single sensor and the influence of data collection under different working conditions on the recognition results show that the installation position of the sensor is not sensitive to the method in this paper,and the data collection under different working conditions has few effect on parameter identification of the first-order bouncing mode and the secondorder pitch mode of the bogie,but influence the third-order rolling mode much.