In-situ quality monitoring is increasingly important for the industrialization of polymer additive manufacturing, where certification-level quality assurance must align with the complex physics of powder sintering. In this context, Selective Laser Sintering (SLS) of thermoplastics such as PA12 has achieved widespread industrial adoption, yet it remains affected by a recurring set of process defects. Despite a substantial body of research on data-driven monitoring for SLS, comparatively few works consider the joint exploitation of visible and thermal signals, and there are few public datasets that support this direction. This paper proposes a Physics-Informed Multimodal framework fusing RGB and thermal (IR) imagery via a Featurewise Linear Modulation (FiLM) mechanism conditioned on process-level texture descriptors extracted from the powder bed. We also introduce a reservoir-based variant of the FiLM generator (ESN-FiLM) with a frozen nonlinear projection and a trainable linear readout. A multi-seed evaluation on a public SLS benchmark, using a physics-guided synthetic IR channel in lieu of a currently unavailable multimodal dataset, shows that the proposed configuration reaches macro- F1=0.989±0.003 over five seeds, exceeding the RGB-only baseline.

A Physics-Informed Multimodal Framework for Manufacturing Defect Detection in Polymer Selective Laser Sintering / Prist, M., Cicconi, P., Di Biase, A., Brunella, F., Colletta, M., Libofsha, A., Monteriù, A., Bonci, A.. - ELETTRONICO. - (2026), pp. 1-4. (31th IEEE International Conference on Emerging Technologies and Factory Automation, ETFA 2026 Wasteras, Sweden september 2026) [10.1109/etfa68377.2026.11724756].

A Physics-Informed Multimodal Framework for Manufacturing Defect Detection in Polymer Selective Laser Sintering

Prist, Mariorosario
Primo
;
Cicconi, Paolo;Di Biase, Alessandro;Brunella, Federico;Colletta, Matteo;Libofsha, Angjelo;Bonci, Andrea
Ultimo
2026-01-01

Abstract

In-situ quality monitoring is increasingly important for the industrialization of polymer additive manufacturing, where certification-level quality assurance must align with the complex physics of powder sintering. In this context, Selective Laser Sintering (SLS) of thermoplastics such as PA12 has achieved widespread industrial adoption, yet it remains affected by a recurring set of process defects. Despite a substantial body of research on data-driven monitoring for SLS, comparatively few works consider the joint exploitation of visible and thermal signals, and there are few public datasets that support this direction. This paper proposes a Physics-Informed Multimodal framework fusing RGB and thermal (IR) imagery via a Featurewise Linear Modulation (FiLM) mechanism conditioned on process-level texture descriptors extracted from the powder bed. We also introduce a reservoir-based variant of the FiLM generator (ESN-FiLM) with a frozen nonlinear projection and a trainable linear readout. A multi-seed evaluation on a public SLS benchmark, using a physics-guided synthetic IR channel in lieu of a currently unavailable multimodal dataset, shows that the proposed configuration reaches macro- F1=0.989±0.003 over five seeds, exceeding the RGB-only baseline.
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11566/366745
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact