泄流结构损伤诊断的数据级与特征级融合方法研究

发布时间：2018-01-04 07:43

本文关键词：泄流结构损伤诊断的数据级与特征级融合方法研究　出处：《南昌大学》2015年硕士论文　论文类型：学位论文

【摘要】：近年来,高水头、大流量的水工泄流结构的安全问题在水利界受到广泛关注。在泄流结构服役过程中,其长期受到水流荷载作用,随之诱发的振动问题导致结构出现不同程度的病害或损伤,情况严重的甚至影响水利枢纽的正常运行。基于多源信息融合技术的泄流结构性态监测与损伤识别体系的建立,不仅能够全面系统地监测泄流结构的灾变演化,更有利于保障大坝的安全运行。在基于信息融合技术的水工泄流结构损伤诊断领域,数据级融合是将多个同类型传感器的振动响应信号,按一定准则,融合为一个更加全面、准确反映结构动力特性(频率)的信号,为泄流结构损伤敏感特征量的提取奠定基础;特征级融主要针对结构多个不同类型的损伤敏感特征量,在特征级层面进行融合,获得对同一损伤的一致性描述,实现对泄流结构实际健康状态的评估。目前有关泄流结构损伤诊断的信息融合方法研究不多,为此,本文主要开展以下两个方面的研究工作:(一)泄流结构多传感器振动信号的数据级信息融合方法研究。针对传统数据级融合方法对振动响应信号相似度要求高,且具有融合系数固定的缺陷,以及在高拱坝结构泄洪振动测试中,传感器测试精度受环境噪声及测点布置影响等问题,本文提出了基于改进方差贡献率的高拱坝多传感器振动响应数据级融合方法,该方法通过定义各测点不同时刻信号相对于该测点信号振动总能量的百分比进行修正,使各测点的融合系数在动态分配时保持在同一能量水平。并结合某拱坝数值模拟试验和二滩高拱坝原型工程实例,验证了该算法在实际工程中的可行性;与相关函数的融合算法进行对比分析,该方法在高拱坝振动测试的信息提取上有明显的优势,且具有更好的动态融合性和适用性。(二)泄流结构损伤诊断的特征级信息融合方法研究。以水工弧形闸门主框架结构为研究对象,采用基于柔度灵敏度动力损伤识别方法,分别对其单位置、非对称位置和对称位置单元三种损伤工况进行识别,发现该方法在识别对称结构非对称位置单元损伤情况下,存在单元误判和损伤程度识别偏差较大的缺陷;采用基于应变改变率的静力损伤识别方法,其损伤程度的识别效果依旧不好。本文分别利用D-S证据理论、Bayes定理,融合动静力损伤敏感特征量,通过识别效果对比验证:多源信息比单一信息更能够准确、全面反映结构的整体特性,特征级信息融合方法能够很好地应用于水工结构的损伤识别领域,且D-S证据理论和Bayes方法的融合效果都与实际损伤情况相符合,但Bayes方法具有更好的工程适用性。
[Abstract]:In recent years, the safety of high-head and high-flow hydraulic discharge structures has been widely concerned in the field of water conservancy. In the discharge structure service process, it has been subjected to the effect of water flow load for a long time. The resulting vibration problems lead to different degrees of disease or damage to the structure. Based on the multi-source information fusion technology, the monitoring of leakage structure and the establishment of damage identification system can not only monitor the catastrophic evolution of discharge structure in a comprehensive and systematic way. In the field of damage diagnosis of hydraulic discharge structure based on information fusion technology, data level fusion is the vibration response signal of multiple sensors of the same type, according to certain criteria. The fusion is a more comprehensive and accurate signal reflecting the dynamic characteristics (frequency) of the structure, which lays a foundation for the extraction of the damage sensitive characteristics of the leakage structure. The feature level fusion is mainly aimed at different types of damage sensitive feature quantity of the structure. It is fused at the feature level to obtain the consistent description of the same damage. To realize the evaluation of the actual health state of the discharge structure, there are few information fusion methods about the damage diagnosis of the discharge structure, so for this reason. This paper focuses on the following two aspects of research work: (1). Research on data level information fusion method for multi-sensor vibration signal of discharge structure. Traditional data level fusion method requires high similarity of vibration response signal. It also has the defect of fixed fusion coefficient, and in the high arch dam structure flood discharge vibration test, the sensor testing accuracy is affected by the environmental noise and the location of measuring points and so on. This paper presents a data level fusion method for multi-sensor vibration response of high arch dam based on improved variance contribution rate. The method is modified by defining the percentage of the total vibration energy of the signal at different time points relative to the measured point signal. The fusion coefficient of each measuring point is kept at the same energy level when it is distributed dynamically. The feasibility of the algorithm in practical engineering is verified by combining the numerical simulation test of an arch dam and the prototype project of Ertan high arch dam. Compared with the fusion algorithm of correlation function, this method has obvious advantages in extracting information of vibration test of high arch dam. And has better dynamic fusion and applicability. (2) study on characteristic level information fusion method for damage diagnosis of drainage structure. Take the main frame structure of hydraulic arc gate as the research object. The dynamic damage identification method based on flexibility sensitivity is used to identify the single position, asymmetric position and symmetrical position unit respectively. It is found that in the case of asymmetric position element damage of symmetrical structure, the method has some defects such as large deviation of unit misjudgment and damage degree identification. Using the static damage identification method based on strain change rate, the damage degree identification effect is still not good. In this paper, D-S evidence theory and Bayes theorem are used to fuse the dynamic and static force damage sensitive characteristic. The comparison of recognition results shows that the multi-source information is more accurate than the single information and can reflect the overall characteristics of the structure. The feature-level information fusion method can be applied to the damage identification of hydraulic structures. The fusion effect of D-S evidence theory and Bayes method is consistent with the actual damage, but the Bayes method has better engineering applicability.
【学位授予单位】：南昌大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TV31

【参考文献】