Wind Project O&M and Safety Conference 2018

Towards Hybrid Domain/Model Data-Driven Approaches For Real-Time Diagnostics Of Big Cross-Platforms Fleets And Its Economic Impact On Operations & Maintenance.

27 Feb 18
11:30 AM - 12:00 PM

Tracks: Asset Management, Emerging Technology

Speaker(s): Gorka Parada, CEO, NEM Solutions USA Inc.

The early detection of symptoms of degradations and potential critical failures of big cross-platforms fleets is a key aspect of a condition-based operations and maintenance (O&M) strategy. With the emergence of digitalization and big-data technologies, data-driven approaches based on advanced statistical and machine learning methods have been targeted as the new paradigm in O&M. The pure model-driven analytics approach usually assumes little or no knowledge of how a turbine's components operate and it relies on advanced machine learning methods to model how a turbine fails upon available data. The domain knowledge driven approach aims to formulate the in-house domain O&M knowledge in mathematical terms. A combination of domain and model-driven approaches is the optimal approach to data-driven O&M. Advanced statistical and machine learning methods help to translate technicians’ knowledge into mathematical rules through visualizations and estimation of thresholds using operational data. On the other hand, domain knowledge helps model-driven approaches to guide the variables selection beyond correlation oriented analysis, and provide a better interpretation of the model representation at the different stages of the learning process. This paper demonstrates that a domain and model-driven hybrid approach has a positive financial impact on O&M. This paper shows real cases from various end users that provide evidence on the positive economic impact.