Study On Bridge Damage Detection In Time-Frequency Domain For Vehicles-Induced Vibration Responses

Under the vehicles-induced vibration, the bridge structural states are load-related, time-varying and nonlinear. But the most exsited damage detection methods do not match these characteristics. So it is necessary to establish the damage detection method complying with these characteristics. The pattern recognition method based on modern statistics can avoid the sick and ill-posed optimization solution, and the whole damage detection process is divided into a serious of independent calculation, such as finite element calculation, sample calculation, partern classification, etc. Thus, the damage detection becomes easier. The time-frequency analysis can reflect both time domain and frequency domain information, and has been used to analyse non-linear and non-stationary signals.Therefore, based on the idea of pattern recognition method and time-frequency analysis, the problems of bridge structural damage detection under the vehicles-induced vibration are studied. The following works are completed:Firstly, the research situation of structural damage detection under the vehicles-induced vibration and structural damage detection based on Hilbert-Huang transform are reviewed. The problems that bridge structural damage detection under the vehicles-induced vibration faced are presented.Secondly, the core contents of pattern recognition and the process of Hilbert-Huang Transform are introduced in detail.Thirdly, a kind of method for damage alarming of bridge structure is proposed. Regarding damage alarming as a two-class problem of pattern recognition, the support vector classification algorithm which applies to small-sample learning is adopted for alarm.Fourthly, a kind of method for damage localization is proposed. The support vector classification algorithm and layered detection method are adopted for location.Fifthly, a kind of method for damage extent detection of bridge structure is proposed. Regarding damage extent detection as a regression-estimation problem of pattern recognition, the support vector regression algorithm which applies to small-sample learning is adopted for detection.Lastly, the simple supported beam and continuous beam bridge are taken as examples for validation of the above proposed method. The results show the method is correct and effective for bridge structural damage detection.