Structural monitoring of critical infrastructure, including hospitals, universities, and schools, is essential for ensuring safety during seismic events. Traditional monitoring methods, however, are complex and time-consuming, limiting effective emergency response capabilities. This paper presents an automated tool based on machine learning techniques for structural failure detection. The study implements Random Forest, XGBoost, and Artificial Neural Networks algorithms, which have demonstrated high effectiveness in predicting and detecting structural failures. The models were trained using intensity measurements extracted from seismic accelerogram data collected from sensors installed at various levels of steel-scale building prototypes. These signals were generated through shake table tests simulating both failure and non-failure conditions. The models achieved accuracy rates between 96% and 99% for both fault detection (binary classification) and fault localization (multiclass classification) during testing and training phases. These results validate the efficacy of machine learning techniques for continuous monitoring of critical structures and optimization of emergency response protocols. The proposed methodology and implemented ML models are available through a publicly accessible online repository.