Heating, Ventilation, and Air Conditioning (HVAC) systems are crucial in the hotel sector, accounting for a significant share of energy consumption, and their unexpected failures are costly. Predictive maintenance offers an effective strategy to reduce these costs. In particular, HVAC chillers contribute substantially to overall energy consumption in hotels, and their sudden failures can result in significant expenses. This study presents a hybrid framework comprising a Fuzzy-based layer and an Artificial Neural Network-based layer, providing a practical, flexible, and interpretable solution for the predictive maintenance of hotel HVAC chillers. The framework addresses challenges related to sensor limitations, data scarcity, and system heterogeneity. The fuzzy logic (FL) layer incorporates expert knowledge through rule-based reasoning, generating meaningful outputs even without historical failure data, which is critical given the infrequent occurrence of HVAC failures. However, an FL-only system cannot adapt to the specific context of each hotel. The ANN-based layer enables continual adaptation through retraining guided by technician feedback. Each output r j ∈ 0 , 1 represents the estimated risk of an upcoming failure in a chiller component and triggers an alert when r j exceeds its threshold T j . The primary contribution of this work is the development of a flexible hybrid framework that integrates expert-driven fuzzy reasoning with data-driven ANN adaptation to estimate failure risk in hotel HVAC chillers, even under sensor limitations and scarce fault data. The system demonstrated strong performance with synthetic data and was validated in a five-star hotel in Seville, Spain.
Ropero et al. (Tue,) studied this question.