The Application Of Improved Independent Component Analysis In Structural Identification

With the development of economic construction and the improvement of science and technology,the country has carried out a large-scale construction in the field of infrastructure.These civil engineering structures during service will be subject to many factors,such as fatigue corrosion,natural disasters and so on,which may lead to a certain degree of structure damage and threat the overall security of the structure.Therefore,the structural health monitoring system is installed in the large structure,however,how to quickly and accurately analyze the structure of the modal parameters and identify the structural damage is still a difficult problem in the health monitoring system.In this paper,unsupervised parameter identification algorithm(Independent Component Analysis)is studied.The limitations of the algorithm are improved to improve the ability of algorithm parameter identification,and finally introduce the algorithm in structural damage identification.The main research contents and achievements are as follows:(1)The structural modal parameters identification methods were reviewed and systematically describes the importance and research status of parameter identification and damage identification technology based on the output,at last,a brief description of the development process of the unsupervised blind separation algorithm is presented.(2)The basic theoretical knowledge of the Independent Component Analysis algorithm,the classical model,the independence of the decision basis,the pretreatment process,as well as the algorithm analysis process are introduced.Independent component analysis is to find the multidimensional data statistically independent and non Gaussian components of a core of the blind source separation algorithm,can be directly with a mixed signal response and more convenient to solve practical problems.Mainly introduced the theory,identify the process and advantages of the fixed-point algorithm(Fast ICA).(3)The IDT-Fast ICA algorithm has been put forward.The connection between Independent Component Analysis algorithm and the structure parameter identification has been studied,and discussed the limitations of Independent Component Analysis in the field of modal parameter identification,to analyze the causes of limitations.Aiming at the limitations,this paper puts forward the improved Fast ICA algorithm based on Inverse Decaying Window,to improve the ability of Independent Component Analysis of parameter identification.Finally,through the results of the spring-mass model,simply supported beams model,and the laboratory experiment of three-story frame to prove the feasibility and effectiveness of the improved algorithm.(4)The DWT-Fast ICA algorithm has been put forward to identify structural damage.Through a combination of Discrete Wavelet Transform(DWT)and Fast ICA algorithm,to quickly identify the structure damage time and corresponding damage location.The effectiveness and feasibility of the proposed algorithm are verified by the multi free spring-mass and three story frame model under different operating conditions.