Machine Learning-Based Prediction of Seismic Failure Mode of Reinforced Concrete Structural Walls

Machine learning techniques have gained significant popularity within earthquake engineering for constructing predictive models to understand how structures will behave during seismic events. These models often employ complex methodologies to attain a high level of accuracy in decision-making. However, the comprehensibility of such predictive models is just as crucial as their accuracy. Engineers require insight into the model’s decision-making process to ensure its practicality. This research strives to simultaneously achieve both transparency and precision, introducing an intelligible classification model designed to forecast the potential seismic failure mode of reinforced concrete shear walls. To accomplish this, eight distinct machine learning methods are utilized where experimental failure modes of conventional shear walls were used and designated as outputs, whereas wall design parameters such as compressive strength of concrete, axial load ratio, etc. were used as the inputs (features). The findings reveal that the Decision Tree approach emerges as the most suitable classifier, effectively delivering both high classification accuracy and interpretability.