Model-Class Selection Using Clustering and Classification for Structural Identification and Prediction

Structural identification using physics-based models and subsequent prediction have much potential to enhance civil infrastructure asset-management decision-making. Interpreting monitoring information in the presence of multiple uncertainty sources systematic bias a model is computationally expensive task. The computational cost this task exponentially proportional number parameters updated data. In paper, novel model-class selection method proposed obtain optimal identifiable classes. Unlike traditional sensitivity methods for selection, method, responses at sensor locations are clustered identify underlying trends response datasets. K-means clustering used determine relevant clusters Cluster indices then as labels classification. Support-vector machine classification forward variable with sequential search select that help classify result trade-off curve comparing error class. This helps practical near-optimal paper compared linear regression-based analysis full-scale bridge. Identification classes obtained both two configurations suggests model-based an efficient minimum remaining fatigue life bridge predicted 720 years represents fatigue-life extension 10 times design predictions prior measurements. approach provides good support asset managers when they interpret measurement