The method of optimal sensor placement based on data correlation: Application to cable group of Nanjing No.3 Yangtze River Bridge

To evaluate the states of cables effectively and accurately, all cables in Nanjing No.3 Yangtze River Bridge are installed sensors monitoring cables under multiple load. However, a large redundancy is found during the analysis of the monitoring data that many cables show great similarity in the change of cable force, which means the strategy that sensors are arranged on all of cables is not only increases economic investment during the construction of system, but also increases more time to analyze monitoring data in the operational phase. Therefore, the question arises of how to construct a cost-effective health monitoring system by selecting the representative sensor positions considering the similarity existing in cable group, which can obtain maximum amount information of structure by those limited sensors. The tendency of time-varying cable force measured by anchor dynamometer is chosen as research object in this paper, which is acquired by an extraction method based on “B-spline interpolation”. The mutual information(MI) based on kernel density estimation, as a general measure of dependencies between variables is applied to detect potential relationship mot only linear in cable group. Then a clustering technology-the bond energy algorithm(BEA) is taken for the optimal placement of cable sensors. Several optimal placements are discussed with different correlation threshold for cable group with 84 stay-cables in the upside of Nanjing No.3 Yangtze River Bridge and verify the effectiveness of this method. Compared with the results that correlation is measured by Pearson Coefficient, the mutual information makes more sense to identify general association existing cable group and to get more reasonable arrangement of sensors.