Session GI3: Disruptions, Stellarators

Disruption Event Characterization and Forecasting in Tokamaks

High reliability disruption prediction and avoidance are critical needs for next-step tokamaks such as ITER. The Disruption Event Characterization and Forecasting (DECAF) code was written to automate analysis of tokamak data determining chains of events leading to disruptions and to forecast their evolution. This approach provides a flexible framework to evaluate the proximity of plasma states to disruption events by coupled physics analyses, model criteria, and machine learning techniques. An expansive list of events including MHD modes, density limits, off-normal plasma motion, and mismatch of plasma current feedback target are evaluated. The expanding tokamak database includes KSTAR, NSTX/-U, MAST, DIII-D, and TCV. Automated analysis of rotating MHD modes allows identification of coupling, bifurcation, locking, and potential triggering by other MHD activity. Resistive stability including Delta’ calculation by the Resistive DCON code is evaluated on long pulse KSTAR plasmas using kinetic equilibrium reconstructions with magnetic field pitch angle data to determine capability for instability forecasting. Greenwald density fraction and local island power balance theory are evaluated for disruption forecasting. As density increases towards or surpasses empirical / theoretical limits, the onset of MHD activity and subsequent disruption are observed. Insights are also gained connecting mode activity to density limit models by their coupling through plasma rotation. In an NSTX database exhibiting global MHD, resistive wall mode (RWM) and loss of boundary control events are always found and VDE events are found in over 90% of plasmas. A reduced kinetic RWM stability physics model computes the evolving proximity of discharges to marginal stability. Stringent marginal stability evaluation with a non-optimized model shows high success (greater than 85%) as a disruption predictor.   

Shattered Pellet Injection Simulations With NIMROD

Using the newly developed Particle-in-Cell (PiC) based Shattered Pellet Injection (SPI) model in the NIMROD code, nonlinear 3D extended MHD initial value SPI driven thermal quench simulations show that MHD activity triggered by injection of impurity fragments strongly impacts the thermal quench dynamics and alters the mixing and radiation efficiency and toroidal peaking factor.  Given the critical nature of the Disruption Mitigation System (DMS) in the operation of ITER, accurate simulations are essential to developing Disruption Mitigation Strategies for ITER.  SPI is the primary candidate for ITER's DMS.  Building upon prior work on Massive Gas Injection (MGI) simulations by V. A. Izzo (Nuclear Fusion 55, 073032 (2015)), we have developed a PiC based model to mimic the fragment plume of the injected shattered pellet.  An analytic expression has been developed to model the ablation of the frozen impurity mixture assuming each fragment is a isolated uniform sphere.  This discrete PiC based SPI plume model captures the essential features of the plume front shielding the fragments that follow and offers the flexibility to conveniently vary trajectory, size, and distribution of the fragments that make up the shattered plume.  D3D and ITER SPI simulations indicate that the thermal quench proceeds in two phases. Initially, the outer plasma is shed via interchange-like filaments while preserving the core temperature.  This results in a steep gradient and  triggers the second phase of the thermal quench, an external kink-like event that collapses the core. Based on these simulations we will comment on advanced mitigation scenarios such as multiple SPI injectors and the prospects for the ITER DMS.
  

Computing local sensitivities and tolerances of stellarators using shape gradients

Tight tolerances have been a leading driver of cost in recent stellarator experiments, so improved definition and control of tolerances can have significant impact on progress in the field. Here we relate tolerances to the shape gradient representation that has been useful for shape optimization in industry, used for example to determine which regions of a car or aerofoil most affect drag, and we demonstrate how the shape gradient can be computed for physics properties of toroidal plasmas. The shape gradient gives the local differential contribution to some scalar figure of merit by normal displacement of the shape. In contrast to derivatives with respect to quantities parameterizing a shape (e.g. Fourier amplitudes), which have been used previously for optimizing plasma and coil shapes, the shape gradient gives spatially local information and so is more easily related to engineering constraints. The shape gradient for any figure of merit can be computed using the parameter derivatives that are already routinely computed for stellarator optimization. Examples of shape gradients for plasma and electromagnetic coil shapes are given. We also derive and present examples of an analogous representation of the local sensitivity to magnetic field errors; this magnetic sensitivity can be rapidly computed from the shape gradient. The shape gradient and magnetic sensitivity can both be converted into local tolerances, which inform how accurately the coils should be built and positioned, where trim coils and structural supports for coils should be placed, and where magnetic material and current leads can best be located. Both sensitivity measures provide insight into shape optimization, enable systematic calculation of tolerances, and connect physics optimization to engineering criteria that are more easily specified in real space than in Fourier space.