19–21 Oct 2026
Frascati
UTC timezone
The deadline for the abstracts submission has been delayed to July 20, 2026

Machine Learning-Assisted Structural Diagnostics of the DTT Vacuum Vessel Through Modal Reconstruction

19 Oct 2026, 16:10
1h
Corridor (Frascati)

Corridor

Frascati

Poster AI in support of diagnostics Poster session A

Speaker

Gabriele Liuzzo (Department of Economics, Engineering, Society and Business Organization (DEIM), University of Tuscia, Largo dell’ Università, 01100 Viterbo, Italy)

Description

Structural diagnostics plays a key role in ensuring the integrity, safety, and reliability of complex engineering systems. A major challenge in structural health monitoring is reconstructing the global structural response from a limited number of displacement or strain measurements acquired at accessible locations. This challenge is particularly critical in fusion devices, where harsh operating conditions and geometric constraints severely restrict sensor deployment.
This work presents a structural diagnostic framework for the Vacuum Vessel of the Divertor Tokamak Test (DTT) facility. The proposed methodology combines modal reconstruction with machine learning to estimate the complete displacement and strain fields from a sparse network of displacement and/or strain sensors. By exploiting a reduced modal representation and data-driven inference, the framework provides a virtual sensing capability, extending the available diagnostic information from the instrumented locations to the entire structure. Additional validation sensors may be used to verify the reconstructed response without participating in the reconstruction process.
The framework is designed to monitor the Vacuum Vessel during both normal operation and off-normal events, such as plasma disruptions. For each measurement set, the algorithm identifies the modal combination that best represents the structural response and reconstructs the corresponding full-field displacement and strain distributions, providing a synthetic diagnosis of the structural state, including inaccessible regions. The reconstructed response also supports sensitivity analyses to estimate the minimum detectable structural defect associated with measurable variations in the monitored quantities.
The proposed methodology offers an effective structural diagnostic tool for fusion applications, demonstrating the potential of integrating modal reconstruction and machine learning for full-field monitoring, structural integrity assessment, and future condition monitoring strategies in next-generation fusion reactors.

Author

Gabriele Liuzzo (Department of Economics, Engineering, Society and Business Organization (DEIM), University of Tuscia, Largo dell’ Università, 01100 Viterbo, Italy)

Co-authors

Dr Gian Mario Polli (DTT S.C.a r.l., Frascati, Roma, Italy) Dr Mauro Dalla Palma (Consorzio RFX (CNR, ENEA, INFN, Università di Padova, Acciaierie Venete SpA), Padova, Italy) Prof. Pierluigi Fanelli (Department of Economics, Engineering, Society and Business Organization (DEIM), University of Tuscia, Largo dell’ Università, 01100 Viterbo, Italy)

Presentation materials