Session: 09-01-16 Wind Energy: Structures 1

Paper Number: 126223

126223 - Diagnostic Digital Twin for Anomaly Detection in Floating Offshore Wind Energy 

Wind energy is increasingly moving offshore, and the potential for floating offshore wind in deep waters is substantial. However, these remote wind farms are more difficult to access and inspect. Furthermore, the response time to failures is much higher, and maintenance operations may require specialized service vessels that need to be rented up to half a year in advance. Unexpected failures can lead to months of turbine downtime thereby reducing both revenue and energy efficiency of the turbine. To avoid unexpected downtime, preventive maintenance strategies with large safety margins are employed. These maintenance strategies result in the replacement of components long before their lifetimes expire and ultimately lead to the decommissioning of turbines that could be operated for several more years. And still, despite the large safety margins, unexpected failures may occur. If the remaining useful lifetime of components could be estimated based on the actual locally experienced conditions and formerly unexpected failures could be predicted from small anomalies, the maintenance strategy could be changed from reactive and preventive to condition-based and predictive.
In the first part of this article, diagnostic digital twins and their application for offshore wind energy are discussed. A diagnostic digital twin is a virtual representation of an asset that combines real-time data and models to monitor asset fatigue throughout its lifecycle, detect anomalies during operation, and diagnose failures, thereby enabling condition-based and predictive maintenance. By applying diagnostic digital twins to offshore wind turbines, unexpected downtime can be alleviated, but the implementation can prove to be challenging. In the second part of this article, a diagnostic digital twin is implemented for a real floating offshore wind turbine, thereby demonstrating the feasibility of the concept. The condition of the asset is monitored based on measured data. Unsupervised learning methods are employed to build a normal operation model and identify anomalies. Real-time capabilities of the model are demonstrated and the results are visualized in a 3D interface and in virtual reality using the Unity engine.
In the final part of the work, the concept and implementation of diagnostic digital twins are discussed in the broader context of offshore engineering. While some details of the methods employed are specific to floating wind turbines, most of the presented methods can be generalized to other offshore assets and enable the shift from reactive and preventive to condition-based and predictive maintenance, thereby ultimately increasing the lifetime, efficiency, and sustainability of offshore assets.

Presenting Author: Florian Stadtmann Norwegian University of Science and Technology

Presenting Author Biography: Florian Stadtmann is a PhD candidate at the Department of Engineering Cybernetics at the Norwegian University of Science and Technology.
In his PhD, he focuses on enabling technologies for digital twins, including data integration, machine learning, hybrid modeling, and virtual reality.
He received his B.Sc. and M.Sc. in Physics at the RWTH Aachen, Germany.

Authors:

Florian Stadtmann Norwegian University of Science and Technology
Adil Rasheed Norwegian University of Science and Technology