Building database of 2D UEDGE simulations for the development of a surrogate model of divertor detachment control

A large set of 2D UEDGE simulations with currents and cross-field drifts based on a generic medium-size tokamak geometry are obtained for the development of machine learning surrogate models for detachment control. For the current 2D data set, three control parameters are varied: gas puff rate, power input and impurity fraction. In addition, the values of the perpendicular anomalous transport coefficients are scanned because they are the most important uncertainty input of UEDGE or other 2D transport codes. The range of the parameters covers the normal experimental control range. A spatial constant concentration of impurities (fixed-fraction) is assumed, and currents for solving electric potential and cross-field drifts are included since they are important in determining divertor plasma states. Compared to our previous work of building a large data set of 1D UEDGE simulations [1] that already captures the main SOL physics by solving plasma parallel transport equations in a flux tube with recycled atom neutrals, these 2D simulations have higher fidelity by including cross-field drifts and the uncertainty of perpendicular transport coefficients, which provides 2D plasma states toward both inner and outer target plates. The data set shows that for a certain power input and impurity fraction, the target plate ion saturation current roll-over appears as the puff rate increases, recovering the trend of experimental data. However, the roll-over appears earlier (at a smaller upstream density) than the results of our previous 1D simulations due to higher radiation. A surrogate model is under development using the 2D UEDGE simulations for detachment control [2]. Possible extensions of including geometry effects: varying divertor leg lengths and divertor target tilt angles, will also be discussed.