针对网壳结构损伤识别中所面临的模态信息不完备、模态密集程度严重以及结构自由度巨大等困难,提出了一种基于时间序列自回归模型(AR模型)与BP神经网络的损伤识别方法.以凯威特型单层球面网壳为例,采用该方法识别损伤杆件位置.数值模拟结果表明,该方法具有较高的准确性和一定的抗噪性.进一步讨论了影响损伤识别结果精度的因素.研究表明,传感器的数目、布置位置对损伤识别结果有一定影响。
In order to overcome the difficulties of the damage identification for latticed shell structures, such as the modal information incompleteness, the mode intensity, and the low accuracy of modeling brought by the large freedom degrees, a methodology of structural damage identification based on autore- gressive (AR) model and BP neural network was proposed. Taking a Kiewit latticed shell as an example, methodology was used to identify the position of damaged members. The effectiveness and noise resist- e of the method were validated by the results of numerical simulation. Furthermore, the study showed t the accuracy of damage detection was influenced by the number and the location of the sensors.