Dof-based submatrix scaling factors for damage detection in reinforced concrete bridges

This study focuses on the system identification and the damage detection of reinforced concrete bridges using neural network algorithm, eigenvalue analysis and parallel computing. First, autoregressive coefficients (ARCs) of both temporal output and forced input of the real structure are computed. The ARCs are used for the eigen-system realization algorithm (ERA) to obtain the modal parameters of the structure. Second, the ARCs are utilized as the input variable of the neural network algorithm while the outputs are the submatrix scaling factors that contain information about the degeneration of each element and each mode within the element. However, the neural network algorithm requires training to output reliable results. The training is the most challenging task of this study and finite element analysis is used to compute the modal parameters of the model built around the neural network outputs. The model is compared with the ERA results to update the neural network coefficients. Due to the scale of the neural network used parallel computing is necessary to reduce the computational time to a reasonable amount. © COPYRIGHT SPIE. Downloading of the abstract is permitted for personal use only