Research Of Structural Monitoring Methods Based On Correlation

Structural health monitoring is widely applied in kinds of large and complexstructure, which is significiant to structural safety assessment. Considering theassociation between different position under external loads, understanding and makingfull use of the relevant laws, the dissertation aims to conduct effective data mining anddata fusion for health monitoring system, and provide more comprehensive and reliablemonitoring information for structural safety evaluation. The optimal placements ofsensors and their fault diagnose based on the analysis of correlation are proposed.In order to reduce the redundancy of measurements from adjacent monitoringpoints, the optimal placements of sensors based on correlation is proposed. Theproposed method aims to obtain independent monitoring information in differentposition, which the maximum effective information can be obtained from these limitedmeasured points and comprehensive information can be provided for structural safetyevaluation. Firstly, the correlation degree for the potential measured points arecalculated and the correlation matrix is established. Secondly, binary processing forcorrelation matrix based on different correlation threshold is operated to get theequivalent correlation matrix. Futhermore, in order to classify the undetermined pointsin several groups, the bond energy algorithm (BEA) is used to transform the correlationmatrix. Subsequently, the number and placements of sensors are determined based onthe principle of optimal placement of sensor. The information entropy is used to validatethe reasonable of the result for optimal placement of sensor.Accurate measurement information can effectively reflect the response of thestructure, in order to ensure the reliability of structural safety evaluation, the applicationof correlation to sensor fault diagnosis is studied. Normal history monitoring data isselect as the reference data for fault diagnosis, the correlation matrix between themeasuring point is calculated to get correlation information database between differentmeasuring points; the appropriate sliding time window is chosen and the correlationvector between different measurement points based on reference data is calculated, thenthe deviation rate vectors of deviation from the average correlation can be obtained inorder to make sure of the threshold value of correlation deviation rate; the faultdiagnosis function based on the threshold value of correlation deviation rate isestablished the proposed method is proved its effectiveness by simulating fault in afinite element model.Three kinds of stress measuring point in structure health monitoring system ofShenzhen Bay Stadium are selected to analyze the correlation of measuring points in each type of stress points, and correlation data information between different points canbe obtained firstly. The different types of fault signals are simulated and add to one ofthe points, the effectiveness and practicalness of the proposed method are proved.