神经网络在混凝土多轴强度预测中的应用

The Application of Artificial Neural Networks in Prediction of Multiaxial Strength of Concrete

目的采用人工神经网络技术来处理混凝土多轴强度间的非线性关系.方法运用BP网络模型对混凝土多轴强度试验数据进行分析,并与数学回归模型进行了比较.结果研究表明,只要选取合适的隐层节点个数和最优化的网络结构,建立的神经网络模型可以合理地模拟具有复杂非线性关系的混凝土多轴强度模型.结论该方法具有较高的预测能力,可以作为混凝土多轴强度准则研究的有益途径.

The strength of concrete under multiaxial stresses is a function of the stress state, it is quite difficult to create a precise mathematical expression of the strength surface since the surface is a complex threedimensional one. The artificial neural network is regarded as a good tool to model the highly nonlinear systems, so in this study, the back propagation neural network （BPNN）is used for training and testing the multiaxial experimental data of concrete from document[9]. When choosing the appropriate number of hidden nodes and the optimal architecture of the network, the artificial neural network is effective in predicting the multiaxial strength. Finally, the conclusion is drawn that compared with regression-based strength models, the neural network approach provides better results as well as a new way for the further study of the failure criterion of concrete.