Real-Time Equilibrium Reconstruction
Equilibrium reconstruction infers the plasma's magnetic shape and internal profiles from external measurements, fast enough to feed control.
The inference problem
The plasma's magnetic equilibrium, its shape, position, and internal current distribution, cannot be measured directly. It is reconstructed by finding the equilibrium that best matches external magnetic measurements and other diagnostics. This is an inverse problem solved many times per second.
The physics constraint
The reconstruction is not free-form; it must satisfy the equation of magnetohydrodynamic force balance, which relates plasma pressure and current to the magnetic field. Enforcing that constraint is what makes a reconstruction physically meaningful rather than just a curve fit to the sensors.
Fast enough for control
- Reduced formulations and precomputed structure cut the solve time.
- Surrogate models can approximate the reconstruction for the fastest loops.
- GPU acceleration handles the linear algebra within the latency budget.
Feeding control
The reconstructed equilibrium is the input to shape and position control and to disruption prediction. If the reconstruction is wrong, every decision built on it is wrong, so its accuracy and speed both matter.
For Hyperion
The Hyperion breeder is a spherical tokamak with negative triangularity of -0.30, a shape that reconstruction must capture faithfully. The tandem-mirror burner has no toroidal equilibrium of this kind, so this technique is specific to the breeder while the general idea of state estimation applies to both.
Verification
Reconstructions are validated against higher-fidelity offline solves and against synthetic cases with known answers, part of verification and validation.