基于BP神经网络的离线迭代混合试验方法研究

利用反向传播(误差逆传播算法back propagation algorithm,BP简称反向传播算法)神经网络对非线性结构位移-力系统进行拟合,将训练好神经网络作为试验子结构与数值子结构联合求解,省去了混合试验中试验子结构与数值子结构之间的实时数据交互,并通过迭代训练样本的方法不断逼近真实响应,克服了神经网络需要大量训练样本的问题。通过对两个自由度的非线性结构进行混合试验数值仿真,验证了该方法的可行性。以实际工程的一榀框架为混合试验对象,取一个隔震垫作为试验子结构进行数值仿真,进一步验证了该方法的有效性。

Research on the Offline Iterated Hybrid Test Method Based on BP Neural Network

This study utilizes back propagation (BP) neural networks to fit the displacement-force system of nonlinear structures. The trained neural network is then used as a physical substructure in conjunction with a numerical substructure to solve the problem, eliminating the need for real-time data exchange between physical and numerical substructures. By iteratively training sample data, this approach continually approaches the real response, overcoming the issue of requiring a large number of training samples for neural networks. The feasibility of the proposed method was verified through numerical simulation of hybrid experiments on two degrees of freedom nonlinear structures. An actual engineering framework was chosen as the object for hybrid experiments, and a seismic isolation pad was selected as the experimental substructure for numerical simulation, further confirming the effectiveness of the method.