泄流结构多传感器振动响应数据级融合方法与参数辨识研究

发布时间：2018-05-15 04:02

本文选题：方差贡献率 + 泄流结构　；参考：《南昌大学》2014年硕士论文

【摘要】：泄流结构的模态参数识别与动力损伤诊断是近20年来研究热点，频率和振型为结构动力损伤诊断的两大整体损伤因子，，其中结构的固有频率是最易测得的动力参数。多测点泄流结构模态识别中，由于不同测点的振动信号频率成分和噪声水平不尽相同，结构固有频率识别结果仅能精确到一定区间范围，而不是一个准确值，这不利于基于频率（或与频率有关的损伤因子）变化的动力损伤诊断。数据级信息融合技术可根据一定的理论规则将多个信号融合为一个更加贴近真实值的信号。信息融合技术在结构工程中的应用尚处于信号模拟的初步探索阶段，为提供准确、全面的振动信号，实现泄流结构模态频率的准确识别，本文主要研究了以下内容：（1）针对传统数据融合算法对信号相似度要求高以及固定融合系数的缺陷，本文提出了基于方差贡献率的泄流结构动态响应数据融合算法。并通过数值模拟的方式，从三个方面（准确性、全面性以及密频结构融合）探讨了基于方差贡献率的泄流结构动态融合算法的可行性，并以蜀河闸墩结构、二滩拱坝及信江枢纽泄流闸墩的原型观测数据验证了该算法的可行性。研究表明：该算法能够有效利用不同测点信号的相关性、互补性和冗余性，使融合后信号具有结构整体振动特性，其动态融合模式较传统加权融合算法更加灵活。（2）研究了基于概率密度函数的泄流结构谐波响应模态剔除方法。多测点模态识别难以分辨信号中的结构响应与谐波响应，针对泄流结构振动响应中混入的谐波信号（如兼发电与泄流的河床式水电站发电机组运转产生的谐波信号），基于方差贡献率的泄流结构多测点动态响应融合算法具有较强的信息提取能力，无法剔除原测点混入的谐波响应，常规去噪方法亦无法将其剔除。基于概率密度函数的泄流结构谐波响应模态剔除方法从统计学的角度分析谐波响应与结构响应的差别，先对多测点信号频域分解，再基于概率密度函数拟合各个频率成分的振动信号，最后通过模拟一组结构响应与谐波响应的混合信号研究了该方法的可行性，并且通过预先在某泄流结构原型观测信号中加入谐波信号研究其识别效果，结果表明：该方法简单而实用。
[Abstract]:Identification of modal parameters and dynamic damage diagnosis of discharge structures have been a hot research topic in recent 20 years. Frequency and mode type are the two major global damage factors for dynamic damage diagnosis of structures, and the natural frequencies of structures are the most easily measured dynamic parameters. In modal identification of multi-point discharge structure, due to the different frequency components and noise levels of vibration signals at different measuring points, the identification results of natural frequencies of structures can only be accurate within a certain range, not an accurate value. This is not conducive to dynamic damage diagnosis based on changes in frequency (or frequency-related damage factors). Data level information fusion technology can fuse multiple signals into a signal closer to the real value according to certain theoretical rules. The application of information fusion technology in structural engineering is still in the initial exploration stage of signal simulation. In order to provide accurate and comprehensive vibration signal and realize accurate identification of modal frequency of discharge structure, this paper mainly studies the following contents: 1) aiming at the shortcomings of the traditional data fusion algorithm which requires high similarity and fixed fusion coefficient, a dynamic response data fusion algorithm based on variance contribution rate is proposed in this paper. Through numerical simulation, the feasibility of dynamic fusion algorithm of discharge structure based on variance contribution rate is discussed from three aspects (accuracy, comprehensiveness and dense frequency structure fusion), and the structure of Shuhe sluice pier is used. The prototype observation data of discharge pier of Ertan Arch Dam and Xinjiang River Project verify the feasibility of the algorithm. The results show that the proposed algorithm can effectively utilize the correlation, complementarity and redundancy of different measurement points, and make the fusion signal have the whole vibration characteristics of the structure. The dynamic fusion mode is more flexible than the traditional weighted fusion algorithm. (2) the method of eliminating harmonic response modal of leakage structure based on probability density function is studied. It is difficult to distinguish the structural response from the harmonic response in the multi-point modal identification. Aiming at the harmonic signals in the vibration response of discharge structure (such as the harmonic signal generated by generating and discharging generating units of river bed hydropower station), the fusion algorithm of dynamic response of multiple measuring points of discharge structure based on variance contribution rate is presented. Has a strong ability to extract information, The harmonic response of the original measurement point can not be eliminated, nor can the conventional denoising method. Based on probability density function, the method of eliminating harmonic response modal of leakage structure is presented. The difference between harmonic response and structural response is analyzed from the point of view of statistics, and the frequency domain decomposition of multi-point signal is carried out. Then the vibration signals of various frequency components are fitted based on the probability density function. Finally, the feasibility of the method is studied by simulating a set of mixed signals of structural response and harmonic response. The recognition effect of harmonic signal is studied by adding harmonic signal into the prototype observation signal of a discharge structure in advance. The results show that the method is simple and practical.
【学位授予单位】：南昌大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TV31

【参考文献】