论文部分内容阅读
利用结构损伤时损伤参数所具有的稀疏性,基于灵敏度分析的有限元模型修正方法,提出一种结合L1/2范数正则化过程的结构损伤识别方法。与以Tikhonov正则化为代表的二次型正则化过程相比,L1/2范数正则化可以有效改善识别结果过度光滑的缺陷;与以L1范数正则化为代表的一次型正则化过程相比较,L1/2范数正则化识别结果更准确。二维框架模型为例的损伤识别数值模拟表明,L1/2范数正则化方法与模型修正方法相结合可以有效抑制实测模态参数中噪声的影响,对于结构局部损伤有更好的识别效果。
Based on the sparsity of damage parameters in structural damage and the finite element model correction method based on sensitivity analysis, a structural damage identification method combined with L1 / 2 norm regularization process is proposed. Compared with the quadratic regularization process represented by Tikhonov regularization, L1 / 2 norm regularization can effectively improve the overly smooth recognition defect. Compared with the one-regularization process represented by L1 norm regularization In comparison, L1 / 2 norm regularization recognition result is more accurate. The damage identification numerical simulation of a two-dimensional frame model shows that the combination of the L1 / 2 norm regularization method and the model correction method can effectively suppress the influence of noise in the measured modal parameters and has a better recognition effect on structural local damage.