Bridge Online Damage Alarming Method Based On Residual Analysis Of Time Series Models

The normal operation of bridges is the premise of the normal running of transportation system which is crucial to national economy. The damage phenomenon of bridges appears to increase with the rising traffic volume which can lead to serious loses without timely damage detection. Structural health monitoring has become of great significance in the research community in the last few years. A time-series algorithm is presented for damage alarming by using ambient vibration data. This thesis conducts the following research with emphasis on bridge structural damage alarming methods:1. A damage alarming approach is presented based on a two-stage prediction model Autoregressive model and Autoregressive model with eXogenous inputs(short for AR-ARX model) after the analysis and summary of the existing structural damage identification methods and time series theory. In this method, a novel sequence matching formula is suggested for data normalization. Besides, an optimization method of the selection of the orders is presented. Prediction errors of AR-ARX models at unknown and reference conditions were used to extract damage feature. A statistical index that is derived from a large number of data samples based on the extracted damage features is used for evaluating the method.2. The applicability and robustness of this approach is demonstrated by numerical simulations performed on a simple supported beam with single and double-zone damaged cases. The vertical ground accelerations emulated by Gaussian white noise are input to the bases to excite the beam.3. Experimental study on a two-span continuous I-beam model will be operated with single and double-zone damaged cases. Numerical simulation will be performed before the test. Acceleration sensor placement scheme will be determined according to actual demand and acceleration time histories will be collected at different damage conditions. The feasibility of the methodology is investigated by analyzing the collected signals using the proposed damage alarming procedure.