Description

This project in collaboration with Buildwise and VMA considers the execution of interventions on big installations (heating, cooling and electrical systems) before a fault occurs, to prevent it from happening and hopefully reducing collateral damage and the total cost of maintaining the installation. The cost can be financial (replacing damaged components) or environmental (a faulty device may consume too much energy for its purpose before being replaced).

In this project, the VUB AI Lab formalizes the problem of preventive maintenance as a Markov Decision Process and applies various Reinforcement Learning algorithms to it. The learned agent will be used to suggest preventive actions by filling tickets to VMA’s ticketing system, hence perfectly integrating with existing workflows.

Research topics:  

• Reinforcement Learning