Physics-Informed Model Predicts Steel Cooling Transformation
Peter Hedström (submitted Nov. 21, 2025) introduces a physics-informed machine learning framework to model continuous cooling transformation (CCT) diagrams for steels, training on a dataset of 4,100 diagrams and validating against literature and experimental data. The model generates complete CCT diagrams with 100 cooling curves in under five seconds, delivers phase classification F1 scores above 88% and temperature MAE below 20°C for most phases.
Key Points
- 1Trains physics-informed ML on 4,100 CCT diagrams to predict phase transformations and full diagrams
- 2Achieves high accuracy: phase classification F1 scores >88% and temperature MAE <20°C (bainite 27°C)
- 3Enables rapid computation—100 cooling-curve CCT diagrams in under five seconds—facilitating fast design iterations
Scoring Rationale
Strong novelty and actionable results with rapid, accurate CCT modeling; limited by single-source preprint and vertical (steel) focus.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
