基于UKF的桥梁空间结构损伤识别研究

发布时间：2018-07-20 11:55

【摘要】：桥梁结构受自然环境和人为等因素的影响,在正常使用期间会出现损伤,并会随时间逐步积累,最终可能导致严重的工程事故。在事故发生前通过结构健康监测,识别结构损伤,掌握结构工作状态,对于保障结构安全、预防突发事故具有重大价值。在众多损伤识别方法中,基于UT变换的无迹卡尔曼滤波方法(UKF)是一种递推型的模型修正方法。与传统时域方法相比,UKF可以通过递推的方式得到状态的最优估计;与扩展卡尔曼滤波方法相比,对于非线性系统参数识别问题,不需要求解复杂的雅克比矩阵且具有更高的精度。针对UKF在结构损伤识别反问题中存在的不适定性,本文基于伪测量技术提出了结合pl范数正则化的UKF方法,可以有效利用不同类型的桥梁损伤先验信息来提高损伤识别效果;并以此为基础将结合正则化的UKF应用到考虑空间特性的桥梁损伤识别中。具体的研究工作如下:(1)利用模态坐标代替结构节点响应构建UKF的状态向量,采取模态截断技巧有效缩减状态向量维数;并将损伤参数添加到状态向量当中,构造结构状态与结构参数混合识别问题;进而利用结构的自由振动观测值,通过递推计算在得到状态最优估计的同时识别损伤。(2)为充分利用人工巡检等方式得到的桥梁损伤先验信息,本文基于伪测量技术,提出将UKF结合pl范数正则化方法识别结构损伤。根据结构的损伤特点可以选择不同的正则化方法,有效缓解了反问题求解的不适定性,提高了损伤识别的精度。论文首先以简支梁为代表的数值算例对两种pl正则化方法进行了比较,分析了不同正则化方法的识别效果和适用的损伤情况;进而考虑具有局部损伤特点结构,以平面桁架为代表重点验证了结合1l范数正则化的UKF方法有效性。(3)在桥梁结构损伤识别分析中,简化的梁式结构是目前最为常见的分析模型;但实际的桥梁都是三维结构,具有空间特性。考虑空间桥梁结构时,结构变形、损伤分区和测点布置跟一维梁式结构都有很大的区别。论文选取板梁桥和T型梁桥两种代表性桥梁作为研究对象,进行损伤识别的研究。首先利用ANSYS建立了具有空间特性的板梁桥结构,考虑结构损伤为局部损伤,研究UKF结合1l范数正则化方法对板梁桥的识别效果,重点研究板梁桥的空间分区和测点布置方案对损伤识别的影响;在针对T型梁桥的分析中,分别以湿接缝、横隔板、主梁为损伤对象进行损伤识别,同时还分析了不同的模态信息、不同的分区数目、不同的测点布置方案对湿接缝的识别影响。结果表明,对于具有空间特性的桥梁空间结构,UKF结合正则化方法可以有效识别损伤,同时具备很强的鲁棒性和一定的抗噪声能力。
[Abstract]:Under the influence of natural environment and human factors, the bridge structure will be damaged during normal use and will accumulate gradually with time, which may eventually lead to serious engineering accidents. Before the accident happens, it is of great value to monitor the structure health, identify the structural damage and master the working state of the structure to ensure the safety of the structure and to prevent the sudden accident. The unscented Kalman filter (UKF) based on UT transform is a recursive model correction method among many damage identification methods. Compared with the traditional time domain method, the UKF can obtain the optimal state estimation by recursive method, and compared with the extended Kalman filtering method, the parameter identification problem of nonlinear systems can be obtained. It is not necessary to solve complex Jacobian matrix and has higher accuracy. In view of the ill-posed nature of UKF in the inverse problem of structural damage identification, a new UKF method combining pl norm regularization is proposed based on pseudo-measurement technique, which can effectively use different prior information of bridge damage to improve the damage identification effect. Based on this, the UKF combined with regularization is applied to bridge damage identification considering spatial characteristics. The specific research works are as follows: (1) the state vector of UKF is constructed by using modal coordinate instead of structural node response, the dimension of state vector is reduced effectively by modal truncation technique, and the damage parameter is added to the state vector. The problem of mixed identification of structural state and structural parameters is discussed, and then the free vibration observation value of the structure is used. In order to make full use of the prior information of bridge damage obtained by manual inspection, this paper is based on pseudo-measurement technology. The UKF combined with pl norm regularization method is proposed to identify structural damage. According to the damage characteristics of structures, different regularization methods can be chosen, which can effectively alleviate the ill-posed problem solving and improve the accuracy of damage identification. In this paper, the numerical example of simply supported beam is given to compare the two pl regularization methods, and the identification effect and the applicable damage of the different regularization methods are analyzed, and then the local damage characteristic structure is considered Taking plane truss as the representative, the validity of UKF method combined with 1l norm regularization is verified. (3) in the damage identification analysis of bridge structures, the simplified beam structure is the most common analysis model, but the actual bridges are three-dimensional structures. It has spatial characteristics. When considering the spatial bridge structure, the structural deformation, damage zoning and measuring point arrangement are different from the one-dimensional beam structure. In this paper, two kinds of representative bridges, slab girder bridge and T beam bridge, are selected as research objects to study damage identification. Firstly, the structure of slab girder bridge with spatial characteristics is established by ANSYS. Considering the damage of the structure as local damage, the recognition effect of UKF combined with 1l norm regularization method for slab girder bridge is studied. In the analysis of T-beam bridge, the wet joint, transverse diaphragm and main girder are respectively used as the damage objects to identify the damage. At the same time, the effects of different modal information, different number of zones and different layout of measuring points on the identification of wet joints are analyzed. The results show that the UKF combined with regularization method can effectively identify the damage, and has strong robustness and anti-noise ability.
【学位授予单位】：南昌大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U441.4

【参考文献】