Luca Santoro — CEO of Therness and Assistant Professor at Politecnico di Torino (PoliTo) — co-chaired the Special Session on Machine Learning for Thermographic Process Monitoring in Non-Destructive Testing at ECNDT 2026, the 14th European Conference on Non-Destructive Testing, held in Verona. The session was organised and moderated together with Prof. Raffaella Sesana, and gathered researchers and industry experts to examine how artificial intelligence and advanced thermography are reshaping industrial quality assurance.
Key Takeaways
- Luca Santoro, CEO of Therness and Assistant Professor at Politecnico di Torino, co-chaired the ECNDT 2026 special session on machine learning for thermographic NDT, alongside Prof. Raffaella Sesana.
- The session focused on how AI, machine learning, and advanced thermographic techniques are transforming process monitoring, defect detection, and quality assurance.
- ECNDT 2026 — the European Conference on Non-Destructive Testing — ran 15–19 June 2026 in Verona, where advanced NDT research met industrial application.
- The themes align directly with Therness’s mission: turning thermography and AI into real-time, in-process quality monitoring for welding and advanced manufacturing.
In this article
- A special session where research meets industry
- Why machine learning is reshaping thermographic NDT
- From conference floor to production line
- Therness at the intersection of AI and thermography
A special session where research meets industry
Special sessions at ECNDT are where the conference’s broad NDT programme narrows onto a single fast-moving frontier — and few frontiers are moving faster than the convergence of machine learning and thermography. The session co-chaired by Luca Santoro and Prof. Raffaella Sesana brought together academic researchers and industrial practitioners to discuss process monitoring, defect detection, and quality assurance across critical applications, from composites to metal fabrication.
“A sincere thank you to all speakers, contributors, and attendees for the engaging discussions and the high scientific quality of the session,” Luca Santoro noted after the event, with special thanks to Prof. Raffaella Sesana for the collaboration in organising and moderating it.
Why machine learning is reshaping thermographic NDT
Thermography is one of the most information-rich non-destructive testing methods available — but that richness is also its historical bottleneck. A single inspection can produce thousands of thermal frames whose subtle gradients encode subsurface defects, bond quality, and process anomalies. Interpreting that data by hand is slow, operator-dependent, and difficult to scale onto a production line.
Machine learning changes the economics of that interpretation. Trained models can:
- Classify defects automatically, separating real flaws from thermal noise and surface artefacts.
- Compress inspection time from minutes of manual review to milliseconds of inference.
- Enable in-process decisions, moving thermography from an offline laboratory check to a live, closed-loop quality control signal.
This is the same shift Therness pursues every day. For readers who want the underlying methods, our technical library covers the practical foundations: a comparison of lock-in vs flash vs pulse active thermography, the industrial applications and standards for active thermography, and how thermography achieves subsurface weld defect detection.
From conference floor to production line
ECNDT 2026’s tagline for its special sessions — “where advanced NDT research meets industrial application” — captures exactly why this work matters. Under the European Federation for Non-Destructive Testing (EFNDT), the conference is the point where peer-reviewed research is pressure-tested against the realities of factories, pipelines, and aerospace structures.
For machine-learning thermography, that translation from lab to line is the whole game. A model that detects a lack-of-fusion defect in a curated dataset is a research result; a system that catches that same defect on a moving weld, in real time, with a traceable record for ISO 3834 documentation, is an industrial product. The discussions in Verona repeatedly circled this gap — data quality, model robustness, calibration, and standardisation — which is precisely where the next decade of NDT innovation will be decided.
Therness at the intersection of AI and thermography
Therness was built for this convergence. We develop AI-driven thermal monitoring systems that watch welding and additive manufacturing processes as they happen — detecting anomalies, flagging defects, and producing an automatic quality record before parts ever leave the cell.
- Explore the welding-camera in-process monitoring hub for how calibrated thermal imaging feeds real-time inspection.
- See HeatCore AI, our software for AI-based quality monitoring in robotic welding cells.
- Review the HeatCam IR-C thermal camera built for industrial weld monitoring.
Chairing a session like this one is part of the same mission: bringing the research community and industry into the same room so that machine learning for thermographic NDT moves faster from promising paper to deployed production capability.
Bring Machine-Learning Thermography to Your Production Line
Therness builds AI-driven thermal monitoring that detects weld and process defects in real time — with a traceable quality record. See it on your application.
Book a demoFrequently Asked Questions
What is ECNDT 2026?
ECNDT 2026 is the 14th European Conference on Non-Destructive Testing, held 15–19 June 2026 in Verona, Italy. Organised under the European Federation for Non-Destructive Testing (EFNDT), it is the largest European gathering of NDT researchers, inspectors, and industrial users, covering radiography, ultrasonics, thermography, eddy current, and emerging AI-driven inspection methods.
What was the special session chaired by Luca Santoro about?
The special session, co-chaired by Luca Santoro and Prof. Raffaella Sesana, focused on Machine Learning for Thermographic Process Monitoring in Non-Destructive Testing. It brought researchers and industry experts together to discuss how AI, machine learning, and advanced thermographic techniques are transforming process monitoring, defect detection, and quality assurance in critical industrial applications.
Who is Luca Santoro?
Luca Santoro is the CEO of Therness and an Assistant Professor at Politecnico di Torino (PoliTo). His work sits at the intersection of thermography, machine learning, and non-destructive testing, where he develops AI-based systems for in-process weld and material quality monitoring.
Why is machine learning important for thermographic NDT?
Thermographic NDT generates large volumes of high-dimensional thermal data that are difficult and slow to interpret manually. Machine learning automates defect classification, suppresses noise, and enables real-time, in-process decisions — turning thermography from an offline laboratory technique into a deployable production quality-assurance tool.
How does Therness apply machine learning to thermographic monitoring?
Therness builds AI-driven thermal monitoring systems that observe welding and additive processes as they happen, flagging anomalies and defects in real time and producing a traceable quality record. The approach combines calibrated thermal imaging with trained models for automatic defect detection and process control.
Where can I learn more about thermographic NDT methods?
Therness publishes detailed technical guides on active thermography, lock-in versus flash versus pulse techniques, and subsurface weld defect detection. See the linked articles in this post, or explore the welding-camera hub for thermal in-process monitoring resources.
