Energy · Transformer oil
Transformer oil properties from one NMR scan, validated against your methods
For utility and grid operator oil laboratories. Rombo AI Platform turns a benchtop NMR scan into several insulating oil property estimates, each benchmarked against your IEC or ASTM reference method with metrics agreed before a pilot. Your methods stay the reference, and your engineers keep the decision.
Use case — Energy
A faster read on oil condition between full test panels
A transformer oil laboratory runs a panel, not a single test: water by Karl Fischer, breakdown voltage, acidity, interfacial tension, dissolved gases, furanic compounds and more, each with its own method and instrument. Rombo AI Platform is designed to add a screening layer to that panel: one NMR scan of the oil, several property estimates, and a flag when a sample moves away from its own history.
Samples that look different can go to the full method sooner. Which properties qualify is decided by validation on your samples, not by a datasheet. Our transformer oil testing guide maps the routine properties, the standard methods and the decisions each one supports.
The customer problem
Utilities and grid operators manage fleets of transformers where the oil is the most accessible window on condition. The tests are well established. The difficulty is capacity and timing: many methods, many samples, and results that arrive as periodic snapshots.
One property, one method
Water, acidity, interfacial tension and dissolved gases each need their own instrument, procedure and trained analyst, so a full panel takes laboratory time.
Periodic snapshots
Test intervals follow the equipment category in maintenance guidance such as IEC 60422. Between two samples, a change in oil condition can go unnoticed.
Trends that are hard to compare
Results from different laboratories, instruments and sampling temperatures are difficult to line up into one trend for each unit.
The Rombo AI solution
Rombo AI delivers validated properties, not another model to build and maintain. Rombo AI Platform starts from an NMR foundation model pre-trained on a large library of spectra, then calibrates it on your oil samples and reference results. It is designed to need fewer reference samples per property than a model trained from scratch, and to transfer across instruments and sites; both are checked during the pilot.
Several properties from one scan
NMR AI Analyzer acquires the benchtop NMR spectrum and returns estimates for the oil properties that passed validation in your pilot.
Validated against your methods
Each estimate is benchmarked on paired samples against your IEC or ASTM reference method, with acceptance metrics agreed before the pilot starts.
Trends and drift per unit
Track each transformer's oil fingerprint over time and flag samples that move away from their history. Drift monitoring is configured during the pilot.
Traceable results
Model versions and audit logs record which model produced each result, so every number can be traced back and re-checked against the reference method.
Main applications
What NMR does not replace
Dissolved gas analysis needs gas chromatography (IEC 60567, interpreted per IEC 60599), furanic compounds need liquid chromatography (IEC 61198), and breakdown voltage is an electrical test (IEC 60156). Published evidence on NMR for transformer oil properties is still limited, so we treat every property as a hypothesis until it passes validation on your samples. For water, the reference stays coulometric Karl Fischer: see our note on measuring water in transformer oil.
From feasibility to production
- Feasibility review, free: we look at your samples, reference data, sampling routine and the decision you want to improve, and tell you whether a pilot makes sense.
- Pilot: a model calibrated on your oil samples and reference results, with a validation report against your IEC or ASTM methods and the metrics agreed at the start.
- Production: deployment on your instruments, with model versioning and audit logs. Integration with laboratory or maintenance systems is scoped after the pilot.
Only the properties that meet the agreed metrics move to production.
How it helps your team
- Oil laboratory managers screen more samples with one scan and send the ones that changed to the full method first.
- Asset and maintenance engineers get more frequent condition data per unit to plan sampling and interventions. The decision stays with them.
- Analytical chemists keep the reference methods as the reference: the model is benchmarked against them, not the other way round.
- Utility leadership sees a measured pilot with acceptance criteria agreed up front, before any scale-up.