Study On Automatic Identification Of Modal Parameters Based On Stabilization Diagram And Machine Learning

As one of the research hotspots in the field of structural health monitoring(SHM),operational modal analysis(OMA)is not only the basis of structural dynamic analysis,but also has important applications in structural optimization design,model modification,damage identification and safety assessment,etc.Among the current modal parameter identification methods,the stochastic subspace identification(SSI)method is widely used due to its many advantages,but this method requires manual participation in the process of modal parameter extraction,which is time-consuming and labor-intensive,and is easily affected by subjective factors,which reduces the accuracy of the identification results.It is not conducive to online monitoring of structural modal parameters.To solve this problem,this paper selects the most widely used covariance-driven stochastic subspace identification(Cov-SSI)method in the time-domain modal analysis method under environmental excitation as the basic algorithm and combines popular deep learning techniques and stabilization diagram theory to realize modal parameter identification.The whole process is automated,and the calculation efficiency meets the needs of online structural monitoring.First,the calculation formula of covariance-driven stochastic subspace identification method based on the reference point and the uncertainty of modal parameters are derived.Through the four-degree-of-freedom spring mass model and the ASCE four-layer frame model,the typical stability diagram method and the modal parameter identification process are demonstrated.The accuracy of the written program is verified by comparing the manual recognition result with the theoretical value.In the calculation example,it is pointed out that there are many false modal points on the stability diagram.To solve this problem,it is proposed to use the uncertainty of the modal parameters to eliminate the false mode,and two numerical examples are used to verify the effectiveness of the algorithm.Furthermore,an automatic identification method of modal parameters based on the Fast Search Density Peaks Clustering(FSDPC)algorithm is proposed,and the basic principles of the algorithm and the selection of default parameters are explained.Then,in view of the low efficiency of the FSDPC algorithm in the online automatic identification of modal parameters,an automatic identification method of modal parameters based on the combination of Convolution Neural Network(CNN)and FSDPC algorithm is proposed.The accuracy and efficiency of the two algorithms are discussed through two numerical examples.The results show that two algorithms can not only better identify the modal parameters of the structure,but also have higher consistency in the identification results.However,the CNN+FSDPC based automatic identification of modal parameters has higher computational efficiency and is suitable for real-time online identification of modal parameters of structure.Finally,two modal parameter automatic identification algorithms are applied to actual engineering structures: Tianjin Yonghe Bridge,Queen’s Park Suspension Bridge and Hong Kong NF276 Pedestrian Bridge,further verifying the accuracy,robustness and practicability of the two algorithms.The recognition results show that the two methods have good recognition results,and there is no need to adjust the parameters of the algorithm during the identification process,and the modal parameter recognition can be completed under the default parameters,which proves the versatility of the algorithm and achieved fully automated of the algorithm.