Data-driven approach to fault detection for hospital HVAC system

Purpose – Hospital heating, ventilation and air conditioning (HVAC) systems are essential to patient safety and wellness. System malfunctions, however, may result in energy waste and even pose health dangers. This project aims to provide a fault detection and diagnostics framework designed primarily for HVAC systems in hospitals. Design/methodology/approach – In order to identify problems in hospital air handling units, the study uses a data-driven methodology that makes use of Long Short-Term Memory (LSTM) and Gated Recurrent Units (GRU) models. To address the problem of uneven data, the dataset is balanced. Other machine learning classifiers, such as Logistic Regression, Multilayer Perceptron, Support Vector Machine, Random Forest, Gradient Boosting and eXtreme Gradient Boosting, are compared to see how well the LSTM and GRU models perform. Findings – Regarding defect detection, the LSTM and GRU models outperform traditional classifiers in terms of both accuracy and computation speed, with high accuracy rates surpassing 90%. Due to its simpler design, GRU achieves higher accuracy and performs faster calculations than LSTM. These recurrent models work well to identify temporal relationships in time-series data, which is crucial for detecting HVAC system problems. Originality/value – This study closes a research gap by concentrating on issue identification in hospital HVAC systems using actual data. It illustrates how deep learning may increase the precision of fault identification and computational efficiency in medical settings by utilizing LSTM and GRU models.