Title :
BP algorithm based on LAC and its application in structural damage detection
Author :
Xianwei, Xiong ; Zhuobin, Wei ; Guanglei, Li
Author_Institution :
Dept. of Logistics Command & Eng., Naval Univ. of Eng., Tianjin, China
Abstract :
An improved BP algorithm is proposed adopting the LAC (least absolute criteria). Based on the research of BP algorithm and its limits, the improved BP algorithm adopts the LAC by using the iteratively reweighted least squares method. The algorithm is analyzed, employed and validated in a structural damage detection of a wing component. Comparing with the ordinary algorithm based on the ordinary least-squares criterion, the improved BP algorithm has a better performance when there are violations of the assumption of normality of residuals, and it is robust to the outliers. The results of the simulation demonstrate the effectiveness of the improved BP algorithm for damage detection.
Keywords :
backpropagation; construction components; fracture; least squares approximations; structural engineering computing; BP algorithm; iteratively reweighted least squares method; least absolute criteria; structural damage detection; wing component; Artificial neural networks; Atmospheric modeling; Monitoring; Presses; Robustness; Testing; BP algorithm; least-absolute criteria; neural network; robustness; structural damage detection;
Conference_Titel :
Electric Technology and Civil Engineering (ICETCE), 2011 International Conference on
Conference_Location :
Lushan
Print_ISBN :
978-1-4577-0289-1
DOI :
10.1109/ICETCE.2011.5775421
