Mitigation of unmeasured environmental and operational variability in long-term bridge health monitoring by unsupervised statistical learning

Environmental and operational variations are significant challenges in long-term bridge health monitoring due to their potential to cause serious confounding influences and wrong decisions emerging as false alarms and mis-warnings. To mitigate these negative consequences, this paper aims at proposing an innovative unsupervised statistical learning method involving three steps of data clustering, unsupervised feature selection, and novelty detection. Firstly, the k-means clustering is applied to categorize dynamic features (bridge modal frequencies) into five types of clusters, reflecting the influences of five key factors that can alter bridge dynamic responses; that is, temperature, humidity, wind, traffic, and damage. Secondly, the clustered features are then processed through an unsupervised feature selection algorithm based on reconstruction independent component analysis to extract reduced features. Thirdly, a novelty detection model that leverages the Mahalanobis-squared distance is used to generate anomaly indices essential for decision-making. A steel arch bridge is utilized to verify the proposed method. Results demonstrate that this method can effectively handle the demanding issue stemming from unmeasured environmental and operational variability conditions.