基于曲率模态和神经网络的结构损伤识别与仿真

探讨用曲率模态和神经网络对混凝土结构裂缝进行损伤识别和定位的方法.以一矩形截面悬臂梁为研究对象,通过完好结构和损伤结构的有限元分析,获得损伤标识量,输入Elman神经网络进行训练,以损伤位置和损伤程度作为网络的输出参数,进行单处损伤和多处损伤的定位研究.数值仿真结果表明,曲率模态振型对结构的损伤敏感,采用曲率模态和神经网络结合的方法可以同时确定结构损伤的存在、程度和位置,并且可以用于结构多处损伤的检测.该方法对于实际工程结构的损伤识别具有一定的指导意义.

The Damage Identification and Simulation by Curvature Mode and Artificial Neural Networks

The curvature mode with artificial neural networks to identify and locate the crack damage is presented in this paper.A cantilever is analysed by finite element method,and the damage indices of perfect structure and damaged structure are gained.Then Elman is used to analyse the single damage and multi-damages quantification and its location.Numerical simulation results show that the changes of curvature mode shapes are sensitive to the damage in the structure.Thus,the curvature mode shapes can indicate the presence of damage and its severity,and locate damage at the same time.Curvature mode shape is also suitable for detection of multiple damages.This method has certain guiding sense to damage identification in concrete projects.