TY - GEN
T1 - Comparisons of three meta-models for structural reliability analysis
T2 - 11th International Conference on Structural Safety and Reliability, ICOSSAR 2013
AU - Li, B. Q.
AU - Lu, D. G.
PY - 2013
Y1 - 2013
N2 - State Key Laboratory of Disaster Reduction in Civil Engineering, Tongji University, Shanghai, P.R. China ABSTRACT: This paper aims at comparing the performance of three meta-models for structural reliability analysis, namely, Response Surface Model (RSM),Artificial Neural Network (ANN) and SupportVector Regression (SVR) machine. The polynomial RSM with the cross terms, the BP-ANN, and the SVR are respectively used to fit the original implicit limit state function in the global random variable space. In order to minimize the number of simulation and to fill the space more uniformly, the uniform design (UD) method is applied to choose experimental points in the space of basic random variables. In addition, the Genetic algorithm (GA) incorporating the gradient information in FORM is employed to search for the global design point to avoid falling into the local optimal solutions. The proposed approaches are programmed in MATLAB by calling and integrating the commercial finite element analysis program ANSYS. The performances of these models are compared in different aspects. The parameter analysis with respect to the BP-ANN and SVR model is performed to optimize the performance of the two meta-models under different conditions. Two numerical examples are provided to demonstrate the accuracy, efficiency and applicability of the three meta-models for structural reliability analysis.
AB - State Key Laboratory of Disaster Reduction in Civil Engineering, Tongji University, Shanghai, P.R. China ABSTRACT: This paper aims at comparing the performance of three meta-models for structural reliability analysis, namely, Response Surface Model (RSM),Artificial Neural Network (ANN) and SupportVector Regression (SVR) machine. The polynomial RSM with the cross terms, the BP-ANN, and the SVR are respectively used to fit the original implicit limit state function in the global random variable space. In order to minimize the number of simulation and to fill the space more uniformly, the uniform design (UD) method is applied to choose experimental points in the space of basic random variables. In addition, the Genetic algorithm (GA) incorporating the gradient information in FORM is employed to search for the global design point to avoid falling into the local optimal solutions. The proposed approaches are programmed in MATLAB by calling and integrating the commercial finite element analysis program ANSYS. The performances of these models are compared in different aspects. The parameter analysis with respect to the BP-ANN and SVR model is performed to optimize the performance of the two meta-models under different conditions. Two numerical examples are provided to demonstrate the accuracy, efficiency and applicability of the three meta-models for structural reliability analysis.
UR - https://www.scopus.com/pages/publications/84892429269
M3 - 会议稿件
AN - SCOPUS:84892429269
SN - 9781138000865
T3 - Safety, Reliability, Risk and Life-Cycle Performance of Structures and Infrastructures - Proceedings of the 11th International Conference on Structural Safety and Reliability, ICOSSAR 2013
SP - 3353
EP - 3360
BT - Safety, Reliability, Risk and Life-Cycle Performance of Structures and Infrastructures - Proceedings of the 11th International Conference on Structural Safety and Reliability, ICOSSAR 2013
Y2 - 16 June 2013 through 20 June 2013
ER -