Deep learning for a customised head-mounted fault display system for the maintenance of wind turbines
Résumé
Virtual reality technology offers a new experiential learning environment that can revolutionise the predictive maintenance domain. These technologies have limitless potential as they can provide an infinite number of solutions and models to help the maintenance manager in his decision-making. In this paper, we propose a 3D learning model generated from data sampled and recorded by the supervisory, control and data acquisition system, to support the manager and operators in the maintenance of offshore wind farms. The application integrates a diagnoser based on a deep learning model to early identify and locate an anomaly. Coupling the virtual reality model with the deep neural diagnoser model allows us to customise the head-mounted fault display system to alert the maintenance operator to the severity or criticality of faults on wind turbines. This information enables him to validate or specify the action he needs to take. As a proof of concept, we applied the methodology to a set of wind farm data supplied by Energias De Portugal. The feasibility and performance are discussed in this paper.