Within this research we present a method to speed up the simulation of convective heat loads onto the plasma-facing components (PFCs) of the Wendelstein 7-X (W7-X) stellarator by a factor of approximately with the same statistical precision as compared to the previous standard. The geometric models developed for this are also designed to unravel the complex 3D PFCs onto a 2D picture-like input format which gives access to the full set of image analysis tools like for example wavelet analysis or the applicability of convolutional neural network (CNN) architectures.The significant speedup of heat load calculation allows to simulate a massive data set of heat loads for approximately magnetic configurations in edge rotational transform-radial axis shift space. To first order, plasma dynamic effects like toroidal current development as well as beta effects mainly influence this space which motivates the simulation in this scope.A criterion to evaluate the safety of a magnetic configuration with respect to the convective heat load onto the plasma facing components has been developed taking statistical fluctuations of the simulation into account. This criterion, applied to the introduced data set, provides a map relevant for discharge planning and machine safety.The methods and concepts introduced herein could contribute to the safety evaluation of magnetic confinement devices in general and are not specific for W7-X.
Machine learning prediction of plasma behavior from discharge configurations on WEST