EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Interpretable screening of processing conditions and material properties for electromagnetic heating of carbon fiber reinforced composites

Sarmad Hassan, Christopher Salazar, Katarzyna Borowiec, Sami Park, Surabhit Gupta, Vipin Kumar, Ashis G Banerjee, Aniruddh Vashisth

Machine Learning: Engineering · 01.12.2026 · 2 (2) · 025016

DOI: 10.1088/3049-4761/aeb894 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Machine Learning: Engineering, online erschienen am 05.11.2026. Verfasst von Sarmad Hassan und 7 weiteren Autor:innen. Die Arbeit nennt 40 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Advanced Manufacturing Office.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Radio-frequency (RF) heating can enable energy-efficient, out-of-autoclave processing and manufacturing of carbon-fiber reinforced thermoplastics, but practical deployment is limited by strongly coupled electromagnetic and thermal transport across a high-dimensional design space. Here, we introduce an interpretable RF–machine learning screening framework that fuses multiphysics finite element simulations with infrared thermography and computer vision analytics to predict manufacturing-relevant thermal responses. Multi-physics simulations generate training data with anisotropic properties, while experiments provide spatiotemporal surface temperature fields; Segment Anything Model (SAM) segmentation extracts core and peripheral heating footprints. Across integrated experimental-simulation studies, tree-based surrogates (LightGBM/XGBoost) outperform linear and kernel baselines, achieving peak-temperature prediction up to R2=0.81 under balanced cross-validation, while Gaussian process and neural network models excel for selected simulation tasks with R2 exceeding 0.95. A model-agnostic interpretation framework (SHAP) identifies current/input power and frequency as the dominant drivers of thermal response, with the relative field orientation, conductivity tensor components, standoff gap, and heat-transfer coefficient providing secondary control. These results provide a practical route to rapid parameter screening and a foundation for closed-loop thermal control in RF-enabled welding and repair of thermoplastic composites.

Förderung

  • Advanced Manufacturing Office

Zitieren

APA

Hassan, S., Salazar, C., Borowiec, K., Park, S., Gupta, S., Kumar, V., Banerjee, A. G., & Vashisth, A. (2026). Interpretable screening of processing conditions and material properties for electromagnetic heating of carbon fiber reinforced composites. Machine Learning: Engineering, 2(2), 025016. https://doi.org/10.1088/3049-4761/aeb894

Vancouver

Hassan S, Salazar C, Borowiec K, Park S, Gupta S, Kumar V, et al. Interpretable screening of processing conditions and material properties for electromagnetic heating of carbon fiber reinforced composites. Machine Learning: Engineering. 2026;2(2):025016. doi:10.1088/3049-4761/aeb894

BibTeX
@article{Hassan2026,
  author = {Hassan, Sarmad and Salazar, Christopher and Borowiec, Katarzyna and Park, Sami and Gupta, Surabhit and Kumar, Vipin and Banerjee, Ashis G and Vashisth, Aniruddh},
  title = {Interpretable screening of processing conditions and material properties for electromagnetic heating of carbon fiber reinforced composites},
  journal = {Machine Learning: Engineering},
  year = {2026},
  volume = {2},
  number = {2},
  pages = {025016},
  doi = {10.1088/3049-4761/aeb894}
}
Kostenlos starten →