论文部分内容阅读
以单层柱面网壳试验模型为背景进行损伤定位研究,采用ANSYS软件对其螺栓松动损伤进行了有限元分析,利用神经网络的模式分类功能进行了螺栓松动的分步损伤定位,即首先以整体结构为研究对象,采用面向子结构的损伤定位,确定结构是否存在损伤及损伤的大致位置;然后采用面向节点的损伤定位法,确定子结构中螺栓松动损伤的节点位置。研究结果表明:损伤识别的成功率与杆件灵敏度密切相关,为了保证识别结果准确性,其灵敏度达到2%以上才能正确识别,当杆件的灵敏度过低时将无法通过动力测试识别;采用分步损伤定位方法可以显著减少损伤精确定位时训练样本的数量,适合于不完备的模态数据,利用低阶模态数据即可准确识别损伤的子结构与节点位置。
The single-layer cylindrical reticulated shell model was used to study the damage location. The ANSYS software was used to analyze the looseness of the bolt. The model was divided into two parts by using the neural network model classification function. Firstly, The whole structure is taken as the research object, and the damage localization of the structure is determined by using the damage localization oriented to the sub-structure. Then, the location of the node of the sub-structure loose bolt damage is determined by the node-oriented damage localization method. The results show that the success rate of damage recognition is closely related to the sensitivity of the rod. In order to ensure the accuracy of the recognition result, the sensitivity can reach 2% or more to be correctly identified. When the sensitivity of the rod is too low, it can not be identified by dynamic test. The step damage location method can significantly reduce the number of training samples when the damage is accurately located and is suitable for the incomplete modal data. The low-order modal data can accurately identify the damage substructure and node location.