Classification of in-plane failure modes for reinforced concrete frames with infills using machine learning

The failure modes of reinforced concrete frame structures with masonry infill panels have strong implications to their overall seismic performance. This paper explores a data-driven approach to classifying the in-plane failure modes of infill frames by employing machine learning methods. To this end, an experimental database consisting of 114 infill frame specimens is constructed. Six machine learning algorithms are implemented and evaluated for failure-mode classification using nine structural parameters as input variables. The validation results indicate that most of the models are able to achieve more than 80% prediction accuracy, with the highest accuracy of 85.7% achieved by the Adaptive Boosting and Support Vector Machine algorithms.