인공신경망을 활용한 매입형 합성기둥의 비선형 변형능력 예측

초록

To perform seismic performance evaluation, nonlinear modeling parameters are essential. However, for concrete-encased steel columns, a relevant parameter set has not been established, probably due to the lack of sufficient test results. In this study, the database of test results was first established and artificial neural network (ANN) models for prediction of the nonlinear deformation capacity (which corresponds to the nonlinear modeling parameter a in ASCE 41) were sought. To improve the prediction accuracy, test data were compared with the results of section analysis, then divided into two groups, which are Group C (Consistent) and Group NC (Not-consistent). Two ANN models, one using all data for training and the other using the Group C only, were established. Comparisons of prediction results between two groups indicated that selection of consistent data for training was very effective to obtain the accurate predictions.

키워드