Comparative Analysis of Random Forest and Support Vector Machine for a Bridge Damage Detection

The need for effective structural health monitoring of bridges has increased due to population growth, transportation development, and the aging of existing bridges. This study compares ensemble-based machine-learning techniques, namely, the Random Forest (RF) method, with the Support Vector Machine (SVM) in the damage detection of bridges. The updated finite element (FE) model of the Swanston bridge is used as a baseline model for damage detection. The results obtained from Interferometric Radar are used to update the FE model of the bridge. The damage is modeled as stiffness reduction on elements. Incomplete modal parameters are used for damage detection. The results indicate the capability of Interferometric in model updating. Moreover, the RF method outperforms SVM in damage detection, comparing time and accuracy.