Power Spectrum Density Transmissibility Based Automated Modal Identification Methods And Engineering Applications

Due to many influences such as environmental corrosion,material aging,load effects,etc.,structural damage will continue to accumulate during normal operation,and serious accidents will be caused under severe conditions.In recent years,more and more structural health monitoring systems have been installed on large-scale infrastructures(such as large-span bridges,high-rise buildings,and special-shaped structures)in order to realize real-time online health monitoring of structures,the structural states are dynamically grasped and structural safety are ensured.Among the many indexes that reflect the inherent characteristics of structures,structural modal parameters are the most widely studied and recognized.In recent decades,modal recognition has always attracted attention in the field of structural health monitoring.As one of the difficulties in real-time state assessment of structures,many researches have been carried out on the automated identification of modal parameters,but there are still some critical issues to be solved urgently.This thesis has carried out detailed and in-depth research on the subject of automated modal identification method and applications based on Power Spectrum Density Transmissibility(PSDT).The main research contents include the following four parts:(1)Spurious modal identification based on PSDT;(2)Closely spaced modal identification based on the Peak Slope of PSDT;(3)Automated identification of modal parameters combined with machine learning algorithm;(4)Automated identification of modal parameters combined with machine vision technology.In chapter 2,a spurious modal identification method based on PSDT is proposed.Firstly,the spurious modal identification under harmonic interference is achieved by comparing the Transmissibility of different measuring points at the same reference point under the harmonic response.The accuracy of the method in identifying spurious modes and its robustness against noise interference are verified by the ASCEBenchmark model.Finally,through the actual engineering application of offshore wind turbine structure,the harmonic interference and structural working modes can be effectively distinguished by this method and the ability to eliminate spurious modes under harmonic interference are further verified.In chapter 3,a closely spaced modal identification method based on the peak slope of PSDT is proposed.In order to identify the possible closely spaced modes in complex structures,the Peak Slope is defined and the closely spaced modes are identified through peak positioning and slope calculation,and then the accuracy of the closely spaced modes is verified through the Relative Difference Coefficients of modal parameters.Finally,the numerical model and the ASCE-Benchmark model are employed to verify the Peak Slope and Relative Difference Coefficients.The complex signal pre-processing is not required in this method,the closely spaced modal parameters can be effectively distinguished and further accurately identified.In chapter 4,an automated identification method of modal parameters combined with machine learning algorithm is proposed.Based on Peak Slope and Relative Difference Coefficients of modal parameters,the Support Vector Machine in the machine learning algorithms is introduced to determine the threshold of indexs above.After the feasibility of this method for automated modal identification fully verified through laboratory model,it is applied to the steel corridor structure of Xiamen International Trade Center.The research demonstrates that the automated identification of frequency and damping ratio can be completed at the same time,which provides greater possibilities for real-time structural health monitoring.In chapter 5,the optical flow method in the field of machine vision is employed to process image vibration signals,and the automated identification method of modal parameters combined with machine vision technology is proposed.Then the feasibility of the optical flow method for automated modal identification was verified through carbon fiber board and C-shaped aluminum beam in the laboratory,and this method was applied to membrane structure and bridge cable.The degree of automation of modal identification is further enhanced through this approach.