A damage identification method based on time series analysis and multi-hierarchical improved BP network

A structural damage identification method for smart wireless sensor network is presented by integrating time series analysis and multi-hierarchical improved BP network. Using time series analysis and analyzing ASCE Benchmark' acceleration response data, five kinds of parameters based on the residual errors are extracted and compared. The standard deviation (std) of the residual errors is chosen as the damage-sensitive parameter. Employing multi-hierarchical damage identification approach based on self-adjusting BP network with momentum item and std as the inputs of neural networks, damage identification of the ASCE Benchmark structure is implemented in three steps, which include locating damaged story, identifying damaged member and identifying the damage severity. The results show the presented approach can quickly and effectively identify damages of ASCE Benchmark structure with high accuracy and precision.