基于在线神经网络算法的混合试验方法

发布时间：2018-04-03 01:03

本文选题：混合试验　切入点：神经网络　出处：《振动与冲击》2017年14期

【摘要】：混合试验是一种将数值模拟与物理试验相结合的新兴结构抗震试验方法,得到了相关研究者们的广泛关注。如何模拟具有强非线性的数值子结构仍是混合试验亟待解决的问题。在传统的离线神经网络基础上提出一种在线学习的神经网络算法,并应用于混合试验中来在线预测数值子结构恢复力。在线学习算法仅利用当前步的系统输入和观测样本,采用递推形式更新每一步的权值和阈值。针对两个自由度非线性结构,分别进行了基于在线学习和离线学习神经网络的混合试验数值仿真。研究表明:与离线学习神经网络算法相比,在线学习神经网络算法具有更好的自适应性,能够有效提高恢复力预测精度和计算效率;基于在线学习神经网络算法的结构混合试验方法可以提高混合试验结果精度。
[Abstract]:Hybrid test is a new structure seismic test method which combines numerical simulation with physical test.How to simulate numerical substructures with strong nonlinearity is still an urgent problem in hybrid tests.Based on the traditional off-line neural network, an on-line learning neural network algorithm is proposed, and it is applied to the on-line prediction of the numerical substructure resilience in hybrid experiments.The online learning algorithm only uses the system input and observation samples of the current step to update the weights and thresholds of each step in a recursive form.The hybrid experiment numerical simulation based on online learning neural network and offline learning neural network is carried out for the nonlinear structure with two degrees of freedom.The results show that: compared with the off-line learning neural network algorithm, the on-line learning neural network algorithm has better adaptability, and can effectively improve the prediction accuracy and computational efficiency of the recovery force;The hybrid structure test method based on online learning neural network algorithm can improve the accuracy of hybrid test results.
【作者单位】：中国地震局工程力学研究所中国地震局
【基金】：黑龙江省青年科学基金项目(QC2013C055) 国家自然科学基金项目(51408157;51308159;51308160)
【分类号】：TU317

【参考文献】