Fault Prediction in a Smart Building Lighting System

With the advances in many areas such as sensing technologies, new connectivity options and improved IoT architectures, predictive maintenance is considered as a promising solution for the maintenance of Smart Buildings (SBs). However, because of the lack of failure data for these systems, the approaches in the literature, which are mostly data-based approaches, are not always applicable. Moreover, a SB is a system of systems where failures in one system can propagate and impact other systems, making maintenance decisions difficult. In this paper, we propose a fault prediction model for the smart building lighting system. This model is based on a Bayesian Network that is scalable according to the operating conditions of the system components. We solely rely on manufacturer’s data that characterize each component to build the failure probability distributions. We show that we are able to characterize and generate statistics of the impacts of a maintenance operation on the system and its components for different intervention scenarios.