Stephen Lane-Walsh
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Stephen Lane-Walsh has a total of 55 co-authors. Here are the top co-authors by number of shared publications:
- Fernando Santoro
Plasma Science and Fusion Center, Massachusetts Institute of Technology, Cambridge, MA, USA
3 shared publications - Joshua Stillerman
Plasma Science and Fusion Center, Massachusetts Institute of Technology, Cambridge, MA, USA
3 shared publications - Amy Watterson
Plasma Science and Fusion Center, Massachusetts Institute of Technology, Cambridge, MA, USA
2 shared publications - Brian LaBombard
MIT Plasma Science and Fusion Center, Cambridge, MA, United States of America
2 shared publications - Daniel Nash
Thea Energy, Inc., Kearny, NJ, USA
2 shared publications - Jeffrey Doody
MIT Plasma Science and Fusion Center, Cambridge, MA, USA
2 shared publications
Latest Publications
- Open Access
- 24 August 2026
- Published by Elsevier in SSRN Electronic Journal
Latency Is Everything: Benchmarking Fusion Data Access for Machine-Learning Dataset Construction
Machine learning is an important tool in fusion research, enabling newapproaches to disruption prediction, control, and large-scale analysis.However, machine-learning workflows require data access patterns that differsubstantially from traditional experimental analysis. Rather than examiningindividual discharges in depth, machine-learning researchers typically extracta small set of signals across many shots to construct training datasets. Theefficiency of dataset construction directly impacts dataset size and scientificvalue.A large fraction of fusion data is stored in \MDSplus archives designed forinteractive, shot-centric workflows. While this model supports detailedanalysis of individual discharges, large-scale studies require efficient accessto selected signals across thousands of shots. Researchers increasingly rely oncurated datasets in formats and services such as \HDFfive, \Zarr, and \HSDS tosupport cloud-based analysis and FAIR data practices.In this work we present results from the MDSplusML project, which investigatesstrategies for constructing machine-learning datasets from fusion archives. Webenchmark data retrieval, transformation, and storage workflows acrosson-premises systems, private clusters, and cloud compute instances with object storage. Archived datafrom the Alcator C-Mod Tokamak~\cite{greenwald_20_2014} were used for these benchmarks. Our resultsshow that data access latency, rather than storage bandwidth alone, is oftenthe dominant factor determining dataset construction performance.
- 24 August 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessLatency Is Everything: Benchmarking Fusion Data Access for Machine-Learning Dataset Construction
Machine learning is an important tool in fusion research, enabling newapproaches to disruption prediction, control, and large-scale analysis.However, machine-learning workflows require data access patterns that differsubstantially from traditional experimental analysis. Rather than examiningindividual discharges in depth, machine-learning researchers typically extracta small set of signals across many shots to construct training datasets. Theefficiency of dataset construction directly impacts dataset size and scientificvalue.A large fraction of fusion data is stored in \MDSplus archives designed forinteractive, shot-centric workflows. While this model supports detailedanalysis of individual discharges, large-scale studies require efficient accessto selected signals across thousands of shots. Researchers increasingly rely oncurated datasets in formats and services such as \HDFfive, \Zarr, and \HSDS tosupport cloud-based analysis and FAIR data practices.In this work we present results from the MDSplusML project, which investigatesstrategies for constructing machine-learning datasets from fusion archives. Webenchmark data retrieval, transformation, and storage workflows acrosson-premises systems, private clusters, and cloud compute instances with object storage. Archived datafrom the Alcator C-Mod Tokamak~\cite{greenwald_20_2014} were used for these benchmarks. Our resultsshow that data access latency, rather than storage bandwidth alone, is oftenthe dominant factor determining dataset construction performance.
- 5 January 2025
- Published by Elsevier in Fusion Engineering and Design
- Vol. 211
Updating the build, test, and deploy systems for MDSplus
- 5 January 2025
- Published by Elsevier in Fusion Engineering and Design
- Vol. 211
Updating the build, test, and deploy systems for MDSplus
- Open Access
- 10 January 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Applied Superconductivity
- Vol. 34 (2) , 1-16
Building the Runway: A New Superconducting Magnet Test Facility Made for the SPARC Toroidal Field Model Coil
A new superconducting magnet test facility was created at the MIT Plasma Science and Fusion Center (PSFC) for the SPARC Toroidal Field Model Coil (TFMC) program. The facility was designed and constructed in parallel with the TFMC between 2019 and 2021, with capabilities and design approaches tailored to the needs of this project and its time line. The major components of the facility include a new cryostat (outer dimensions, 5.3 m × 3.7 m × 1.5 m) with open bore; a novel cooling system circulating supercritical helium in a closed loop to provide $\sim$ 600 W cooling power at $\sim$ 20 bar-a, $\sim$ 20 K; a 50-kA $\pm$ 10-V power supply with supporting nitrogen-cooled high temperature superconductor (HTS) binary current leads operating at record currents, as well as VIPER-cable HTS cold bus; and a new instrumentation and programmable-logic-controller-based control system handling $\sim$ 650 input and output signals distributed between the facility and the test article. Substantial legacy infrastructure inherited from the PSFC's Alcator C-Mod tokamak program, including liquid nitrogen facilities and 10 mW of ac power, was instrumental in the rapid deployment of these new systems. Immediately after initial commissioning, the facility was used successfully to test the SPARC TFMC, operating the magnet in a campaign achieving 20 T on the coil, as well as a second campaign performing quench testing. The facility has since undergone several upgrades and has been used in campaigns of other test articles, and it is expected that the facility will remain a resource for the community for the foreseeable future to develop fusion magnets and related technology.
- 10 January 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Applied Superconductivity
- Vol. 34 (2) , 1-16
Open AccessBuilding the Runway: A New Superconducting Magnet Test Facility Made for the SPARC Toroidal Field Model Coil
A new superconducting magnet test facility was created at the MIT Plasma Science and Fusion Center (PSFC) for the SPARC Toroidal Field Model Coil (TFMC) program. The facility was designed and constructed in parallel with the TFMC between 2019 and 2021, with capabilities and design approaches tailored to the needs of this project and its time line. The major components of the facility include a new cryostat (outer dimensions, 5.3 m × 3.7 m × 1.5 m) with open bore; a novel cooling system circulating supercritical helium in a closed loop to provide $\sim$ 600 W cooling power at $\sim$ 20 bar-a, $\sim$ 20 K; a 50-kA $\pm$ 10-V power supply with supporting nitrogen-cooled high temperature superconductor (HTS) binary current leads operating at record currents, as well as VIPER-cable HTS cold bus; and a new instrumentation and programmable-logic-controller-based control system handling $\sim$ 650 input and output signals distributed between the facility and the test article. Substantial legacy infrastructure inherited from the PSFC's Alcator C-Mod tokamak program, including liquid nitrogen facilities and 10 mW of ac power, was instrumental in the rapid deployment of these new systems. Immediately after initial commissioning, the facility was used successfully to test the SPARC TFMC, operating the magnet in a campaign achieving 20 T on the coil, as well as a second campaign performing quench testing. The facility has since undergone several upgrades and has been used in campaigns of other test articles, and it is expected that the facility will remain a resource for the community for the foreseeable future to develop fusion magnets and related technology.
- Top Cited PapersOpen Access
- 14 November 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Applied Superconductivity
- Vol. 34 (2) , 1-18
Experimental Assessment and Model Validation of the SPARC Toroidal Field Model Coil
In this article, the SPARC Toroidal Field Model Coil (TFMC) experimental tests are described. The tests include detailed comparisons to a hierarchy of electromagnetic and structural models of the coil. The tests confirm the ability of the no-insulation no-twist (NINT) configuration to provide highly stable dc operations with peak magnetic field in excess of 20 T at the rare earth yttrium barium copper oxide (REBCO) tape stacks. The advantages of the modular TFMC approach are validated including the ability to test and probe the REBCO and structural response of the coil in stages, a strategy that will be applied to large-scale coil production for SPARC. The electromagnetic models show high fidelity agreement to the dc and charging coil performance in terms of current distribution, voltages, and heating. The transient response of the TFMC to open-circuit events is examined. The coil exhibits the positive features of the NINT design with very low induced voltages and a predictable redistribution of current and resulting volumetric heating. However, highly localized damage occurs during a programmed open circuit at operational conditions similar to those used in the SPARC tokamak, in agreement with calibrated electromagnetic models. This reveals the cause and remedy to the thermal instability that gives rise to the damage. Therefore, the conclusion of the tests is that the NINT TFMC is confirmed as meeting the programmatic requirements for SPARC and, more generally, for high-field tokamaks.
- 14 November 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Applied Superconductivity
- Vol. 34 (2) , 1-18
Top Cited PapersOpen AccessExperimental Assessment and Model Validation of the SPARC Toroidal Field Model Coil
In this article, the SPARC Toroidal Field Model Coil (TFMC) experimental tests are described. The tests include detailed comparisons to a hierarchy of electromagnetic and structural models of the coil. The tests confirm the ability of the no-insulation no-twist (NINT) configuration to provide highly stable dc operations with peak magnetic field in excess of 20 T at the rare earth yttrium barium copper oxide (REBCO) tape stacks. The advantages of the modular TFMC approach are validated including the ability to test and probe the REBCO and structural response of the coil in stages, a strategy that will be applied to large-scale coil production for SPARC. The electromagnetic models show high fidelity agreement to the dc and charging coil performance in terms of current distribution, voltages, and heating. The transient response of the TFMC to open-circuit events is examined. The coil exhibits the positive features of the NINT design with very low induced voltages and a predictable redistribution of current and resulting volumetric heating. However, highly localized damage occurs during a programmed open circuit at operational conditions similar to those used in the SPARC tokamak, in agreement with calibrated electromagnetic models. This reveals the cause and remedy to the thermal instability that gives rise to the damage. Therefore, the conclusion of the tests is that the NINT TFMC is confirmed as meeting the programmatic requirements for SPARC and, more generally, for high-field tokamaks.