Posters

Abstract

We present an integrated approach that combines a real-time in‑oil wear‑particle sensor with interpretable SCADA‑based analytics to improve early detection and diagnosis of gearbox wear on wind turbines. The internally developed sensor delivers particle event signals and spectral features that act as proxies for particle size and material class. These direct measurements are fused with an analytics pipeline that (1) aligns heterogeneous status‑code logs to uniform 10‑minute SCADA records, (2) quantifies production impact using generalized additive models (GAM) to compute energy‑deficits per event, and (3) produces percentile‑based anomaly scores from a gradient‑boosting regression predicting operational parameters from environmental inputs. Deployed across multiple operator fleets (SCADA analytics applied to 70+ turbines; sensors installed across multiple wind turbine sites) and validated in laboratory test rigs and by endoscopic gearbox inspections, the combined method detected signs of bearing degradation earlier than vibration monitoring and conventional particle counters. In a documented gearbox failure case, a SCADA anomaly in gearbox oil‑pressure was flagged 14 days before an alarm from a conventional particle‑counter; trends from our in-house sensor system showed a sustained rise in particle counts over months prior to endoscopic confirmation in several turbines. The presentation will demonstrate how tilemaps, energy‑deficit summaries and anomaly timelines translate into actionable, prescriptive maintenance decisions, and will discuss practical lessons for piloting and scaling the approach across heterogeneous fleets.