Multimodal Super-Resolution: Discovering hidden physics and its application to fusion plasmas

Abstract

A non-linear system governed by multi-spatial and multi-temporal physics scales cannot be fully understood with a single diagnostic, as each provides only a partial view, leading to information loss. Combining multiple diagnostics may also result in incomplete projections of the system's physics. By identifying hidden inter-correlations between diagnostics, we can leverage mutual support to fill in these gaps, but uncovering such correlations analytically is too complex. We introduce a machine learning methodology to address this issue. Unlike traditional methods, our multimodal approach does not rely on the target diagnostic's direct measurements to generate its super-resolution version. Instead, it uses other diagnostics to produce super-resolution data, capturing detailed structural evolution and responses to perturbations previously unobservable. This not only enhances the resolution of a diagnostic for deeper insights but also reconstructs the target diagnostic, providing a valuable tool to mitigate diagnostic failure. This methodology addresses a key challenge in fusion plasmas: the Edge Localized Mode (ELM), a plasma instability that can cause significant erosion of plasma-facing materials. A method to stabilize ELM is using resonant magnetic perturbation (RMP) to trigger magnetic islands. However, limited spatial and temporal resolution restricts analysis of these islands due to their small size, rapid dynamics, and complex plasma interactions. With super-resolution diagnostics, we can experimentally verify theoretical models of magnetic islands for the first time, providing insights into their role in ELM stabilization. This advancement supports the development of effective ELM suppression strategies for future fusion reactors like ITER and has broader applications, potentially revolutionizing diagnostics in fields such as astronomy, astrophysics, and medical imaging.

Figures (10)

• Figure 1: Main methodology. DiagA is essential to capture fast transient events near the edge of plasma. But due to its low temporal resolution and accuracy it fails to track the evolution of such events. Diag2Diag solves this problem by generating synthetic super resolution of DiagA by learning the correlation between DiagA data and other diagnostic measurements with higher resolutions and better accuracy.
• Figure 2: Reconstructing co2 spectrograms from ece spectrograms of DIII-D shot 170669 using convolutional neural networks. (a) The configuration of 4 ece and 40 ece probes at DIII-D . (b) A tensor of $(40 \times time \times frequency)$ is supplied to CNN. (c) The configuration of CNN. (d) Visual comparison of measured and reconstructed spectrograms (e) Comparison of the Alfven Eigenmode detector output Garcia2023comparison supplied with the measured and reconstructed spectrograms.
• Figure 3: (a) Comparison of the electron density by the measured ts and the synthetic srts, for the DIII-D shot 153761 diallo2015 near the edge ($Z=0.71m$). $D_\alpha$ with arbitrary units is plotted as an indicator of elms. An example of elm event captured by both diagnostics, and another example only captured by srts are highlighted in green and purple, respectively. (b-c) Aggregating the measured ts density and temperature in three locations of plasma near the edge for several elm cycles of the DIII-D shot 174832. The circle highlights the measures ts for one selected elm cycle and the solid lines present the srts which agreeably match the measures ts. $t=0$ represents the time when elm is identified by $D_\alpha$. (d-e) The evolution of srts between two consecutive measured ts near one elm cycle across the plasma location.
• Figure 4: Structure of 3D coils and islands by perturbed field (a), and the evidence in the simulation (b-d) and the srts diagnostic (e-g) for rmp-induced island mechanism on the plasma boundary in DIII-D shot 157545.
• Figure 5: (a) Time evolution of edge safety factor ($q_\text{95}$) and $D_\alpha$ emission at plasma edge. (b) Contour of electron pressure versus normalized plasma radius and time. The numerically derived width of the magnetic island at the pedestal top is illustrated as green contours. (c) Comparison of TS (blue), srts (red), and filtered srts (orange solid line).