(MENAFN- The Arabian Post)

A newly devised artificial intelligence system dubbed Diag2Diag is poised to reshape how physicists observe and control the seething plasma inside fusion reactors. The tool generates synthetic diagnostic data to compensate for gaps left by conventional sensors, particularly at the plasma's elusive edge, enhancing stability control and reducing hardware demands.

Diag2Diag emerges from a cross-institutional effort led by Princeton University and the Princeton Plasma Physics Laboratory, with partners in South Korea and Columbia University. The system was trained on measurements from the DIII-D National Fusion Facility, a major U. S. fusion experiment. The AI examines inputs from working sensors-monitoring properties such as magnetic fields, density, or temperature-and predicts what slower or spatially limited diagnostics would have recorded. The synthetic output aligns closely with real sensor data and often delivers finer resolution than hardware alone.

The plasma“pedestal,” or edge region, represents a critical frontier in fusion control: instabilities arising there can degrade performance or even damage reactor walls. Because some sensors cannot track rapid edge fluctuations, Diag2Diag fills those voids by cross-modality inference. As lead author Azarakhsh Jalalvand describes, the system“takes ... data from a bunch of sensors ... and generate[s] a synthetic version of the data for a different kind of sensor.” The result helps scientists monitor fast-evolving events that traditional diagnostics miss.

One diagnostic in particular, Thomson scattering, is limited in temporal resolution and struggles to capture the dynamics in the pedestal region. Diag2Diag augments the effective resolution of such diagnostics without requiring additional hardware. Egemen Kolemen, a senior researcher on the project, says the AI is“giving your diagnostics a boost without spending hardware money.” By reconstructing data streams at higher fidelity, Diag2Diag enables continuous insight into plasma behaviour-even when physical sensors fail or lag.

See also Autism's Genetic Edge: Evolution's Hidden Bargain

The AI also contributes to advancing understanding of edge-localized modes, abrupt energy releases that pose threats to reactor integrity. A prevailing theory holds that applying resonant magnetic perturbations can suppress ELMs by creating“magnetic islands” that flatten temperature and density variations at the plasma edge. Conventional sensors have struggled to validate that flattening conclusively. Using synthetic diagnostics, the team observed simultaneous flattening of both temperature and density in the pedestal, lending strong empirical support to the magnetic island suppression theory.

This insight has concrete implications for ITER and next-generation tokamaks, which will rely on ELM control schemes to operate safely over long periods. Diag2Diag could reduce the need for dense sensor arrays, simplify reactor design, and lower costs. As SangKyeun Kim from PPPL notes, future commercial reactors will likely“need to have far fewer” diagnostics to maintain space and reliability.

The research builds on the team's recent Nature Communications publication, which outlines the methodology for cross-diagnostic super-resolution. The approach exploits hidden correlations between sensor modalities-relations too complex for analytic derivation-to reconstruct what an underperforming diagnostic would have detected.

Beyond the fusion domain, Diag2Diag may find application in other systems where sensor degradation or incompleteness is a concern. The authors suggest use cases in spacecraft, robotic systems, or medical devices, where maintaining continuity of measurement is crucial. The team is already exploring expansions to additional fusion diagnostics and broader implementation across research facilities.

The emergence of Diag2Diag underscores a deeper shift in fusion science: diagnostics and control are becoming inseparable from data science. While earlier efforts emphasised materials, magnetic confinement, and plasma heating, the frontier now lies in interpreting and interpolating vast sensor networks in real time. By enabling synthetic“virtual diagnostics,” the AI becomes a new instrument in its own right.

See also Gen X Women Exhibit High Rates of Ultra-Processed Food Dependence

Notice an issue? Arabian Post strives to deliver the most accurate and reliable information to its readers. If you believe you have identified an error or inconsistency in this article, please don't hesitate to contact our editorial team at editor[at]thearabianpost[dot]com . We are committed to promptly addressing any concerns and ensuring the highest level of journalistic integrity.

MENAFN12102025000152002308ID1110183906