基于神经网络技术的复杂框架结构节点损伤的两步诊断法

大量研究表明,对于发生损伤的大型复杂结构,采用常规的一步方法进行损伤诊断将是十分困难,甚至是不可能的。因此,本文对多层及高层复杂框架结构节点损伤,提出了基于神经网络技术的两步诊断方法,此方法先将结构划分为n个子区域,将损伤引起的结构前n阶模态频率变化比与损伤区域的关系输入概率神经网络,建立系统,进行损伤子区域判定;然后将结构损伤子区域内第二阶杆端应变模态变化量与节点损伤位置和损伤程度的关系输入径向基神经网络,建立系统,进行损伤位置和损伤程度具体诊断。数值仿真分析结果表明,此方法可对多层及高层框架结构的地震节点损伤做出成功诊断,且具有较好的抗干扰能力。

TWO-STEP APPROACH FOR JOINTS DAMAGE DIAGNOSIS OF FRAMED STRUCTURES BY ARTIFICIAL NEURAL NETWORKS

A two-step approach to diagnose the joint damage of framed structures by artificial neural networks is proposed in the paper. At first, the structure is divided into many sub-domains. The frequency ratio of damaged structure to completed structure of first n modes is inputted into the probability neural network. The damaged sub-domains are determined. The variation of the element-end strain modal is successfully used as a damage target. Then, the damaged location and degree are diagnosed. The example is shown that the proposed approach is effectively. It can be used to diagnose the joint damage with good anti-jamming abilities.