Predictive maintenance of building facility: A digital twin framework using long short-term memory encode-decode model

In recent years, the predictive maintenance (PdM) approach has emerged as a cost-cutting balance between the action after failure (reactive maintenance) and routine maintenance before failure (preventive maintenance). The goal of digital twins (DT) in facility management is to constantly monitor the performance of a system and aid in intelligent decision-making for the optimal operation and maintenance of the facility. Integrating DT and PdM concept facilitates real-time monitoring and predicting the building facility's status. This chapter proposes a standard data-driven DT framework for the PdM of an air handling unit (AHU). It develops a framework for data-driven failure predictive DT of the AHU using LSTM and LSTM encode-decode models. The chapter tests the performance of the developed framework using real-time operational data of an AHU. The Earle Hall Building at University Park, Pennsylvania State University is used as the test platform.