基于BP神经网络的空间索杆结构承重索损伤识别研究

Research on damage identification of load-bearing cables in a practical spatial cable-strut structure based on back-propagation neural networks

针对某实际空间索杆结构的承重索预应力松弛现象,采用BP神经网络与基于振动的损伤识别方法,分别对单榀承重索和双榀承重索的预应力松弛进行了识别研究。研究表明,虽然空间索杆结构的动力性能较为复杂,但归一化后的结构前10阶固有频率变化比和归一化后的对应结构第一阶模态的部分节点的损伤信号指标对承重索的损伤位置较为敏感,利用其进行损伤定位是可行的。在此基础上,再增加考虑结构的第一阶固有频率平方变化比即可进一步有效识别出承重索的损伤程度。

Using the back-propagation neural network and the damage identification method based on structural dynamic responses, prestress-looses of single load-beating cable and double load-bearing cables in a practical spatial cable-strut structure are diagnosed. Simulation results show that the first 10 natural frequencies change ratios and the damage signal indexes corresponding to the first mode are sensitive to the damage location of load-bearing cables, and the first natural frequency square change ratio is effective for identification of damage extent of load-bearing cables.