Verification, Validation and Uncertainty Quantification
The project aims to accelerate commercial fusion energy development by ensuring predictive models are robust, transparent, and experimentally validated—minimizing risk and building stakeholder confidence in fusion reactor design and operation.
As we approach the era of burning plasmas, the computation and data challenges in the field will become much larger. In order to support the next step of fusion data analysis, we have been started the construction of an Integrated Research Infrastructure (IRI) that seamlessly integrated data from large-scale experimental facilities like the DIII-D National User Facility with powerful computing resources that will help researchers keep pace with the ever-increasing influx of scientific data. This work is a large collaboration between scientists at General Atomics, Columbia University, Argonne National Laboratory, Princeton Plasma Physics Laboratory and Lawrence Berkeley National Laboratory, with additional contributions from many others. The goal of this framework is to empower researchers with world-class physics models, modern engineering tools, continuous data access and HPC resources to radically accelerate discovery and innovation in fusion pilot plant design. The effort emphasizes rigorous Verification, Validation and Uncertainty Quantification (VVUQ) methodologies to provide quantifiable confidence in reactor design decisions. By systematically validating models across multiple facilities and operational conditions while incorporating real-time stability assessments, this project aims to establish a robust, transparent foundation for future fusion power plant development that minimizes risk and accelerates the path toward commercial fusion energy.
Project Details
Research into accelerating fusion energy realization focuses on developing predictive models and control strategies validated through multi-facility experiments. Columbia researchers actively design and execute experimental campaigns across three major tokamak facilities—DIII-D in the United States, KSTAR in South Korea, and WEST in France—with examples of feed-forward trajectory optimization and scenario development pictured above. These techniques are essential for confident decision-making in the design and operation of future fusion power plants, including ITER, currently under construction in France.
Topics of active research include systematic Verification, Validation and Uncertainty Quantification (VVUQ) of physics and control models to ensure they are credibly anchored in experimental reality. Comparing results from different facilities allows understanding of model transferability across diverse reactor environments and operational conditions. Active research is also integrating MHD stability assessment tools into unified modeling frameworks to optimize actuator timing and avoid instabilities during plasma discharge startup.
Columbia researchers are also actively developing interpretable machine learning models for plasma control with systematically identified bounds of applicability. These efforts leverage a multi-institutional data platform to assess the reliability of real-time diagnostics in sparse data environments.
DIII-D Tokamak
Edge stability research on the DIII-D tokamak focuses on the use of small electromagnets to control the ELM, and the exploration of novel plasma scenarios naturally stable to the ELM.
International Tokamaks
Work at international research facilities enables targeted comparisons of ELM control phenomena under a wide range of plasma conditions.