دانلود مقاله Comparison of RBF,GR and BP with Wavenet Bach Propagation Neural Networks in Approximation Dynamic Analysis of Structure
سال انتشار: ۱۳۸۵
محل انتشار: هفتمین کنگره بین المللی مهندسی عمران
تعداد صفحات: ۸
S.Gholizadeh – ph.D. student
In recent , neural network are considered as robust tools for fast approximation with arbitrary accuracy in time consuming problems. Dynamic analysis of structures against the earthquake has the time consuming process, in this study we employ Radial Basis Function (RBF) , Generalized Regression (GR), Back propagation (BP) and Wavenet Back Propagation (WBP) neural network , for approximating dynamics time history responses of frame structures. RBF , GR and BP are traditional network while WBP is a wavelet neural network as sigmoid activation function of hidden layer neurons are substiluted with wavelets. To train WBP the optimum weights are determined using Scaled Conjugate Gradient algorithm (SCG). Comparison of WBP results with the other network indicates that generality of the properly trained WBP is excellent and all of them can be efficiently used for approximating dynamic analysis of the frame structures.