基于模型修正与图像处理的多尺度结构损伤识别

发布时间：2019-03-18 12:22

【摘要】：结构的损伤识别在既有结构的评价与维护中具有重要意义。论文结合结构动力指标与裂缝指标,利用基于频响函数灵敏度的模型修正方法进行结构整体损伤识别,利用基于图像处理与三维重建的裂缝识别技术进行结构局部损伤识别,形成由整体到局部、由宏观到微观的多尺度损伤识别方法。论文的主要研究成果有:(1)针对由整体到局部的多尺度损伤识别思路,论文利用模型修正的方法实现对整体结构的损伤定位及损伤程度初判;在此基础上,利用图像处理等方法实现局部裂缝的定量识别,从而实现了两种损伤识别方法在逻辑与操作上的衔接。(2)在基于频响函数灵敏度的模型修正中,针对参数敏感性分析、灵敏度方程组病态性、构造灵敏度方程的频响函数频率点选取、阻尼处理等问题展开系统讨论,给出了两类灵敏度指标,提出了综合考虑指标敏感性与灵敏度方程组病态性的频带选择指标,并结合数值算例对指标予以说明、验证。(3)在基于频响函数灵敏度的模型修正中,为处理数据不完备情况,提出了整体有限元模型缩聚、局部子结构频响函数扩展的两步自适应修正方法,实现了在测试数据严重不完备的情况下,对局部损伤的准确识别。通过四边固支有限元板的数值算例,验证了在测试数据严重不完备(约测得6%自由度)并存在噪声干扰的情况下,该损伤识别方法的有效性及准确性。(4)在基于图像处理的裂缝识别中,针对混凝土表面裂缝图像中可能存在的强噪声,提出了自适应的降噪算法。该方法结合背景减除损伤及Niblack’s二值化操作,实现在图像处理过程中维持裂缝形态不变的前提下,最优地消除噪声。同时,针对混凝土结构中常见的复杂裂缝形态,提出了复杂裂缝的分解编号、单独处理的方法,准确地完成裂缝宽度与长度的参数计算。利用多组试验对该算法的有效性及精度进行了验证。(5)在混凝土结构表面裂缝识别研究中,针对传统图像的识别方法中存在的问题,结合三维重建技术,提出了三维条件下的裂缝识别技术:通过裂缝体表面三维重建、二维图像裂缝参数识别、裂缝体三维并行投影等步骤,可实现裂缝图像的倾斜拍摄、多表面贯穿裂缝识别以及大面积裂缝的合并、集成识别。对该方法进行的试验验证表明其具有较强的应用价值。
[Abstract]:Damage identification of structures plays an important role in the evaluation and maintenance of existing structures. Combined with the structural dynamic index and crack index, this paper uses the model correction method based on the sensitivity of frequency response function to identify the whole damage of the structure, and uses the crack identification technology based on image processing and three-dimensional reconstruction to identify the local damage of the structure. To form a multi-scale damage identification method from whole to local, from macro to micro. The main achievements of this paper are as follows: (1) aiming at the idea of multi-scale damage identification from whole to local, this paper uses the method of model modification to realize the damage location and damage degree judgment of the whole structure; On this basis, the image processing and other methods are used to realize the quantitative identification of local cracks, so that the two damage identification methods are connected in logic and operation. (2) in the model modification based on the sensitivity of frequency response function, the two damage identification methods are connected with each other in logic and operation. Based on parameter sensitivity analysis, ill-conditioned sensitivity equations, selection of frequency response function frequency points for constructing sensitivity equations, damping treatment and so on, two kinds of sensitivity indexes are given. A frequency band selection index considering index sensitivity and ill-conditioned sensitivity equations is proposed and illustrated by numerical examples. (3) in the model modification based on the sensitivity of frequency response function, the index is verified by numerical examples. In order to deal with the incomplete data, a two-step adaptive correction method is proposed to reduce and aggregate the whole finite element model and expand the frequency response function of local substructure, which realizes the accurate identification of local damage when the test data is seriously incomplete. A numerical example of a four-sided fixed-support finite element plate is given to verify that the test data is seriously incomplete (about 6% freedom) and there is noise interference. The effectiveness and accuracy of the damage identification method. (4) in the crack recognition based on image processing, an adaptive noise reduction algorithm is proposed for the possible strong noise in the concrete surface crack image. Combined with background reduction and Niblack's binarization, the method optimally eliminates noise while maintaining the same crack shape during image processing. At the same time, according to the common complex crack form in concrete structure, this paper puts forward the decomposition and numbering of complex crack and the method of dealing with it separately, so that the parameter calculation of crack width and length can be completed accurately. The validity and accuracy of the algorithm are verified by multi-group experiments. (5) in the research of surface crack recognition of concrete structures, the problems existing in the traditional image recognition methods are analyzed, and the 3D reconstruction technology is used to solve the problem. The technology of crack recognition under three-dimensional condition is put forward: through the steps of three-dimensional reconstruction of fracture body surface, two-dimensional image recognition of crack parameters, and three-dimensional parallel projection of crack body, the slant shooting of fracture image can be realized. Multi-surface penetration crack identification and large-area fracture combination, integrated identification. The experimental results show that the method has strong application value.
【学位授予单位】：清华大学
【学位级别】：博士
【学位授予年份】：2015
【分类号】：TU317

【相似文献】