混凝土三轴循环本构关系的神经网络模拟

利用BP神经网络的模拟能力代替传统的力学方法，对混凝土材料的循环本构关系进行了模拟研究．首先回顾了此类模型的研究进展和构造方法。然后直接从试验数据出发，建立了混凝土三轴受压循环比例加栽条件下的本构模型．试验结果和模拟结果的比较说明，该模型具有较高的精度和良好的泛化能力．本研究结果进一步说明神经网络模型适合于描述多影响因素的非线性复杂因果规律，为研究材料本构特性提供了一条新的途径．

Acquiring the Triaxial Cyclic Constitutive Relationship for Concrete Using an Artificial Neural Network

The modeling capabilities of neural networks are used as substitutes for the classic approaches to investigate the cyclical constitutive relationship of concrete. Based on reviewing of the advances in study on constitutive model and constructive methods by neural networks, a back-propagation (BP) neural network for triaxial constitutive model of concrete subjected to cyclic proportional loading is developed. A good agreement between the measured data and predicted results demonstrates that the BP neural network is able to capture significant variability inherent in the concrete samples, and has promising applications in modeling the constitutive behavior of concrete subjected cyclic proportional triaxial compression. It can be concluded that artificial neural networks have unique learning capabilities that can be used in learning complex nonlinear relationships, and offer a fundamentally different approach in modeling of constitutive behavior of materials.