Transformer Technology Tech Talks – Luiz Cheim, Hitachi ABB

The most dangerous job in the power industry doesn’t involve high voltage — it involves climbing inside a transformer. And one robot is making it obsolete.

In this Transformer Technology Tech Talk, Alan Ross sits down with Luiz Cheim, the inventor of the TXplore — the submersible inspection robot developed at ABB (now Hitachi ABB) that navigates oil-filled transformers, inspecting internal components without draining oil or risking human life. This conversation was first sparked at a Doble conference, where Cheim’s former colleague Craig Stegemeier demonstrated the TXplore remotely — and Ross couldn’t walk away.

What began as a bold concept inside ABB’s transformer service organization became a multi-million-dollar, four-year collaboration with ABB’s Robotics division. Cheim, who later won the 2018 CIGRE Paris Best Paper Award for his pioneering work on machine learning in transformer diagnostics, shares the untold story: the ball-shaped prototype that was too large to fit most transformers, the risky redesign to miniaturize it, the communications breakthrough through a Faraday cage environment, and the “no loose parts” mandate that kept the project alive.

Key Takeaways:

• How ABB’s Robotics division partnered with transformer engineers on a multi-million-dollar, four-year R&D bet with no guarantee of success
• The ball-shaped prototype that was too big — and the make-or-break redesign that miniaturized the TXplore
• Why a transformer full of oil acts as a Faraday cage — and how the team engineered a communications bridge through it
• Testing in tea: how the Robotics lab simulated progressively darker oil before taking the robot into real transformers in St. Louis
• The “no loose parts” design mandate — why a single dropped screw or washer would have killed the entire program
• Real-world deployments where TXplore inspections told utilities: “Nothing’s wrong — energize it,” saving millions in unnecessary teardowns
• Luiz Cheim’s transition from robotics to AI: building machine learning algorithms for transformer diagnostics and asset management
• The 2018 CIGRE Paris paper (“Machine Learning in Support of Transformer Diagnostics”) that won Best Paper — and why novelty mattered more than polish
• What’s next: self-training algorithms that ingest live site data to predict and prevent transformer failures across entire fleets