RONIN AI turns the DGA and oil data you already have into a CIGRE TB 761 health index for every asset, ranked by failure risk. In seconds. No sensors, no hardware, no integration project.
No sales calls · guided by AIPlay the live demo below, then create a free account for the guided demo on preloaded assets.
Free account · guided demo · no sales calls.
Arc furnace or transmission grid, a fault leaves the same signature in the oil. RONIN reads it before it becomes an outage.
RONIN AI is a transformer health index software. It calculates a health index built on CIGRE TB 761 from your existing DGA and oil test data, then ranks an entire fleet by failure risk so you fund the maintenance that matters first. The score comes from an interpretable machine-learning model anchored to the CIGRE standards, not a black box. No sensors to install, no IT rollout, results in seconds.
Transformer fleets are managed on assumptions, not data. Too many assets, too little time, and diagnostic reports that sit in spreadsheets instead of driving decisions. Ageing infrastructure and a shrinking expert workforce have made reactive maintenance untenable. "Old" is not the same as "at risk", but most fleets still spend as if it were.
Lab reports arrive as spreadsheets and PDFs, then stall. The signal you paid for never becomes a decision.
Reading DGA correctly takes a specialist most teams no longer have in the room.
Budgets follow age or failure, not actual risk. The wrong transformer gets the attention.
Every parameter in a health assessment has a cost: a sensor to fit, a lab test to run. RONIN's model distilled the 14 that carry the signal, out of 60+, so you get a dependable health index without paying to gather data you do not need. And when some are missing, RONIN still answers, at a stated, lower confidence, instead of stalling. The score returns in about 6 seconds.
Confidence is always shown, never silent. Fewer inputs means a lower, declared confidence, not a stall and not a false certainty.
Same substation data every incumbent already has. RONIN reads the DGA and oil, scores each asset on CIGRE TB 761, and settles the whole population worst-first. Switch the sort, read the verdict. This is the demo, running in your browser.
Free account · guided demo · no sales callsThe lab results you already have, turned into a fleet ranked by risk. The Prediction List is where a whole fleet becomes a ranked worklist: green is healthy, red is at risk, and every asset carries its health index, date, type and area. This is the actual RONIN interface, not a mock-up of one.
Every transformer you have scored, ranked so the at-risk units surface first.
Every card is a step the tool takes for you, from the lab results you already have to a ranked fleet you can share with a board.
Upload DGA, oil and nameplate data. A health index per unit, one at a time or across a whole batch.
A per-asset report with the health index, the Duval Triangle for DGA and a plain interpretation.
Standard recommendations tied to the result, so the next step is clear without an expert in the room.
Track an asset across testing dates and catch degradation before the next scheduled assessment.
Organise assets into fleets and rank the whole population by severity, worst first.
Share assets with their health indices as A4 or A3 PDF, or export to Excel for your own reports.
A plain verdict for the technician who is not a specialist. The sub-scores, Duval and standard clause one click away for the engineer who wants to check the work.
Download the guided template, fill in the DGA and oil results you already collect from any lab, and upload it. No sensor, no integration project, no IT ticket.
✓ Guided templateA single 0 to 100 score with band, trend and a recommended action. One colour, one verdict, one next step.
⏱ ~6s per assetThe whole population settles worst-first, so the budget conversation starts with the right asset.
⏱ ~30s · up to 500 unitsTwo anonymised programmes, told as the customer stated the problem and what the health index made possible. Sectors only, no names.
The challenge: a generation utility with an obsolete generator step-up fleet could not tell which units actually carried the risk, and kept spending on the wrong ones.
A 4-year health-index trend flagged the risk in time for a one-week planned outage, instead of an unplanned failure and a two-year replacement lead time.
The challenge: a steel group running electric arc furnaces (over 1.1M tonnes a year) needed to know where the reliability budget should go across a large, mixed fleet.
Over 100 assets classified by risk and priority, so testing went where it mattered and planned outages fell.
Figures from real Seetalabs deployments, reported by sector and anonymised. Your results depend on your data and your fleet.
Colour always comes with a label and the value, so a report stays readable in greyscale and no button is ever mistaken for a "Critical" asset.
RONIN's score comes from an interpretable machine-learning model anchored to CIGRE TB 630 and TB 761, with the standard clause shown next to every result. The trust comes from interpretability and standards, not from hiding the AI. An open box, not a black one.
The international reference for transformer health index and fleet prioritisation. RONIN's model is anchored to it, so the output maps back to a published method.
Developed in alignment with the standard for AI governance, transparency and lifecycle accountability.
AI used to manage critical infrastructure is classified as high-risk. RONIN is governed and documented against those obligations ahead of the applicable deadlines.
The 80% is the R² (coefficient of determination) of RONIN's health-index model, an interpretable machine-learning model anchored to CIGRE TB 630 and TB 761. It is high for a model trained on real, messy, cross-utility data. Forcing it toward 85% or more would mean overfitting the test set, a number that looks better on a slide and performs worse on your fleet.
The reason to trust it is not the number. It is that the model is interpretable and tied to published standards: every score carries its sub-scores, its Duval interpretation and its clause. The model compensates for missing parameters. It does not mask poor-quality data: good data in is still a precondition, not a variable we pretend to fix.
No dark patterns and no fine print. Here is exactly where your data lives, what it is used for, and how it is kept apart.
Your data sits on a VPS in the European Union and is handled under the GDPR.
Your DGA, oil and asset data are never used to train or tune the model. Your fleet stays yours.
No behavioural tracking. Account data is minimal: name, email and an optional company. No photos.
Each customer runs on a separate domain, accessible only to the owner. Asset data is non-sensitive by design.
Encryption at the highest standard, everything over HTTPS. The ML service is never exposed to the internet.
Access is firewalled and the scoring service is reachable only through the application, not directly.
If you already run oil treatment, testing or field service, RONIN becomes a diagnostic service you sell to your own clients. You bring the samples and the relationship; RONIN turns the data into a health index and a ranked fleet your customer can act on. A software add-on, no hardware to stock.
Plain-language verdicts for the industrial technician who is not a specialist; the depth on click for the asset manager and the engineer. You do not need to be a DGA expert to know which transformer is at risk.
Since 1999 I have had one job that changed name about ten times: I enter a field I do not know, learn it fast, and dig until I find what the specialists stopped seeing. Industrial compliance, then ecodesign, then artificial intelligence. In 2019 I pointed the same method at power transformers.
Court-appointed technical expert at the Court of Turin. Taught Industrial Standardisation and Engineering at Politecnico di Torino. Founder of Ecotp and Quantic. Independent expert in the Plenary of the European Commission's Code of Practice for general purpose AI.
Seetalabs is founder-led and lean by design, backed by a deliberate network of power engineers, transformer testing specialists and utility asset managers across Europe, Asia-Pacific and the Americas.
Our transformer expert, grounded in the standards and RONIN’s library, answers right here, in your language. This public preview reads the library, not your fleet.
HAKUTAKU (白澤) is the beast of East Asian myth that knows every ailment and names it - fitting for one that knows transformer faults and names them.
The full expert agent that reasons over your own transformers inside the RONIN platform, not just the library. In development now.
Usage-based and transparent. No hardware to buy, no hidden line items, and no opaque enterprise quote you have to chase.
You pay for the predictions you run, the assets you actually assess. No hardware, no seats you do not use, no hidden costs.
The more assets you assess, the less each one costs. Scoring a whole fleet is cheaper per asset than checking a single unit.
Larger volumes, partner and white-label terms are available on request, on the same transparent, usage-based basis.
No "contact us" wall. You see how it works, and try it free, before anyone asks you for a number.
Register once. The guided demo runs your first health index on preloaded assets in minutes, then on your own data when you are ready. No sales calls.
Free account · guided demo on sample data · 500 kVA to 800 MVA
Onboarding assistant. No sales calls, the AI guides you.
LinkedIn profile