基于测点关联分析的结构监测方法研究

发布时间：2018-05-19 23:39

本文选题：关联度 + 键能算法　；参考：《哈尔滨工业大学》2014年硕士论文

【摘要】：结构健康监测作为结构安全评定的一个重要手段，在各类大型复杂工程结构中应用越来越广泛。考虑到建筑结构在外部荷载作用时，不同测点处的结构响应存在着相互关联，充分认识和利用测点之间的关联性规律，以对健康监测系统进行有效的数据挖掘及数据融合，为结构安全评定提供更全面及可靠的监测信息。本文依据测点响应信息之间的关联性分析，对传感器优化布置及故障诊断进行了相关的研究。为降低测点之间的冗余性，使不同位置处的监测信息尽可能保持独立，让有限测点发挥最大作用，为结构安全评定提供全面信息，本文研究了基于关联度的传感器优化布置方法。根据有限元分析结果及关联度计算公式计算所选待定测点之间的关联度，建立关联度矩阵；对关联度矩阵进行二元处理，设定关联度阈值，以得到等价关联度矩阵；采用键能算法对等价关联度矩阵进行矩阵变换，以达到对待定测点进行分组的目的；根据两两之间平均关联度最小的传感器优化布置原则，确定传感器数量及位置分布；采用衡量冗余度的相关信息熵对传感器优化布置结果进行验证，以验证方法的有效性。准确的测量信息才能有效地反映结构的响应，为保证结构安全评定的可靠性，本文研究了基于关联度的传感器故障诊断方法。选取历史正常监测数据作为故障诊断前的参考数据，计算测点的关联度矩阵，采用二元变换及键能算法对测点进行分组，获得测点之间的关联性信息数据库；选取适当长度的滑动时间窗，利用参考数据计算测点之间随滑动时间步变化的关联度向量，进而求得偏离平均关联度的偏离率向量，以确定所研究测点任意两点间的偏离率限值；以关联度偏离率限值为依据建立故障诊断函数，采用有限元模型仿真故障的识别实验验证了方法的有效性，并讨论了噪声对测点关联性的影响。选用深圳湾体育中心结构健康监测系统中的三类应力测点为研究对象。对每类应力测点进行关联性分析，，获得了测点间关联性强弱的数据信息，分别在三类应力测点中的其中一个测点，加入输出恒定值、漂移及周期性干扰信号以仿真故障测点，采用本文提出的基于关联度的故障诊断方法验证了对于含有噪声的实测数据分析的有效性及实用性。
[Abstract]:As an important means of structural safety assessment, structural health monitoring is more and more widely used in various large and complex engineering structures. Considering that the structural responses at different measuring points are interrelated when the building structure is subjected to external loads, it is necessary to fully understand and make use of the law of correlation between the measured points in order to carry out effective data mining and data fusion for the health monitoring system. Provide more comprehensive and reliable monitoring information for structural safety assessment. Based on the correlation analysis of the response information of measurement points, the optimal arrangement and fault diagnosis of sensors are studied in this paper. In order to reduce the redundancy between the measuring points, to keep the monitoring information at different positions as independent as possible, to allow the limited observation points to play the greatest role, and to provide comprehensive information for structural safety assessment, In this paper, the optimal arrangement method of sensor based on correlation degree is studied. According to the results of finite element analysis and the calculation formula of correlation degree, the correlation degree matrix is established and the correlation degree matrix is established, and the threshold value of correlation degree is set up to obtain the equivalent correlation degree matrix. The matrix of equivalent correlation degree is transformed by the key energy algorithm, and the number and position distribution of sensors are determined according to the principle of optimal arrangement of sensors with minimum average correlation degree between two pairs, and the number of sensors and the location distribution of sensors are determined according to the principle of optimal arrangement of sensors with minimum average correlation degree. In order to verify the effectiveness of the proposed method, the optimal sensor layout is verified by the entropy of information to measure redundancy. The accurate measurement information can effectively reflect the response of the structure. In order to ensure the reliability of structural safety assessment, the method of sensor fault diagnosis based on correlation degree is studied in this paper. Selecting the historical normal monitoring data as the reference data before the fault diagnosis, calculating the correlation degree matrix of the measuring points, grouping the measured points with binary transformation and key energy algorithm, and obtaining the correlation information database between the measured points. By selecting a sliding time window of appropriate length and using the reference data to calculate the correlation degree vector between the measured points with the sliding time step, the deviation rate vector deviating from the average correlation degree is obtained. The fault diagnosis function is established by determining the limit value of deviation rate between any two points of the measured points, and the validity of the method is verified by using the finite element model simulation experiment to identify the faults, and to establish the fault diagnosis function based on the limit value of the correlation degree deviation rate. The effect of noise on the correlation of measurement points is discussed. Three kinds of stress measurement points in the structural health monitoring system of Shenzhen Bay Sports Center were selected as the research object. The correlation analysis of each kind of stress measuring points is carried out, and the data information of the correlation between the measuring points is obtained. In one of the three kinds of stress measuring points, the output constant value, drift signal and periodic disturbance signal are added to simulate the fault measuring points. The method of fault diagnosis based on correlation degree proposed in this paper is used to verify the validity and practicability of the analysis of measured data with noise.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TU317

【相似文献】