Damage Detection of Bridges Considering Environmental Temperature Effect by Using Cluster Analysis

It is difficult to detect the damage of practical bridges by using the variation of monitoring modal parameters directly since the varying environmental conditions may mask the change of modal parameters induced by the damage of bridges. In this study, Gaussian mixture model (GMM) combining with novelty detection was proposed to eliminate the effect of environmental temperature on vibration frequencies of bridges. Firstly, GMM was applied to classify the monitoring modal parameters, obtained by using the long-term monitoring data of bridges, into different clusters. The monitoring vibration frequencies of bridge satisfying the same probability distribution were classified into the same cluster, which means that these vibration frequencies were acted by the similar environmental temperature load. Secondly, at each cluster, the novelty detection was implemented to detect the damage of bridges. Finally, the effectiveness of proposed method was demonstrated by using a numerical example.