Nondestructive evaluation (NDE) methods are widely used to detect defects in bridge decks. These methods evaluate the bridge deck condition from different aspects. Data fusion is a viable approach to efficiently process and combine such heterogeneous NDE data for making more informed decisions. In order to fuse multi-resource NDE data, it is crucial to avoid the counter-intuitive results due to potential conflict between the measurements. In this study, discrete wavelet transforms (DWT) and improved Dempster-Shafer (D-S) evidence combination theory are proposed to develop a multi-resource NDE data fusion framework. A series of NDE data periodically collected through a full-scale bridge accelerated testing program are used to create the proposed framework. The deployed NDE methods include half-cell potential, ground penetrating radar, electrical resistance, and ultrasonic waves. The results from the data fusion analysis are compared with those derived using individual NDE, visual inspection, and advanced vision-based methods. By leveraging the access to the unique data sets collected from the Bridge Evaluation and Accelerated Structural Testing (BEAST) facility, the feasibility of the proposed method has been evaluated for the entire bridge lifetime under a controlled environment. The efficacy of the fused NDE data in detecting various bridge defects is further discussed via correlating the information captured during the accelerated bridge testing with a representative bridge in the state of Pennsylvania.
Read full abstract