NorwayEnergy and utilities2017 – ongoingStatkraft
Machine-learning health models for hydropower maintenance
Statkraft detected sensor changes before a bearing failed 12 hours ahead of its scheduled inspection
Problem
Aging hydropower equipment required costly maintenance. Risk assessments relied heavily on physical inspections and manual calculations.
Approach
Statkraft started a machine-learning pilot at Kvilldal in 2017 and tested a health model at Jostedal. The model detected small sensor changes, prompting an inspection scheduled 3 weeks later. A thrust bearing failed before that inspection. The team then reported installing extra sensors and maintaining weekly contact with site staff.
Results
Takeaway
Detecting deterioration did not establish how urgently maintenance staff needed to intervene.
Categories
- Industry
- Energy and utilities
- Business function
- Production and quality
- Innovation type
- Process innovation
- AI technology
- Predictive analytics and ML
- Adoption
- Built in-house
- Goal
- Quality and risk
- Evidence
- Company's own claim
Sources
Source summaries
Statkraft described maintenance pilots at Kvilldal and Jostedal. Sensor anomalies preceded a bearing failure, but the planned inspection was 12 hours too late.
Full details are in the original sources.
