人工神经网络预测混凝土柱屈服性能

建立了一种基于人工神经网络的矩形混凝土柱屈服性能预测方法.该方法采用经验模型进行柱屈服性能影响因素的分析来确定该神经网络的输入参数,并通过敏感性分析验证了所选神经网络输入参数的合理性.为验证该方法的可行性与有效性,通过对PEER 210组矩形混凝土柱的屈服性能进行预测分析并与经验预测模型的预测结果进行比较.比较分析结果表明:神经网络预测结果与实验结果吻合程度远高于其他经验预测模型;同时也证实该方法在实验数据稀少的情况下为预测结构在地震作用下的性能提供一条新途径.

Prediction of the Yield Performance of RC Columns by Neural Network

This paper developed an artificial neural network (ANN) based method for the prediction of the yield performance of rectangular RC columns. In this method, the inputs of ANN were determined on the basis of the empirical studies of the impact factors of the yield performance of RC columns. The sensitivity analysis of the yield performance of RC columns was also investigated to validate the reasonability of the inputs selection of ANN. In order to demonstrate the feasibility and effectiveness of the proposed method, the ANN models were applied to predict the yield performance of rectangular RC columns by using 210 sets of experiment data provided by PEER, Furthermore, the predicted results were compared with empirical model results. Comparative analysis has shown that the prediction degree of agreement with the experiment results of ANN models is much better than that of other empirical prediction models. Also, the result reveals that the proposed method provides a new way for accurately estimating structural performance under earthquake with data scarcity.