Tomographic reconstructions of the MAST U Fast Ion Loss Detector using iterative algorithms
Tikhonov regularization has so far been the method par excellence applied to Fast Ion Loss Detector (FILD) [1] measurements to obtain an accurate fast ion velocity distribution [2]. To address two major drawbacks of Tikhonov regularization, the computational expense and the critical reliance on the optimal value of a hyperparameter, iterative algorithms have been developed and validated using the ASDEX U pgrade FILD [3]. Here, we extend this work by evaluating Kaczmarz, Coordinate descent and Cimmino algorithms together with the resolution principle for the MAST U FILD [4] using a synthetic signal model. The performance of the algorithms is evaluated by calculating their reconstruction error evolution with the number of iterations (i.e. the error of the obtained velocity distribution for each as well as the computation time. Gaussian noise is added to the synthetic model to simulate the electronic noise induced by the experimental setup in real measurements. To characterize the reliability of the reconstructions, we develop a novel “fidelity map” based on reconstructing signals in each grid point. The Coordinate descent algorithm is applied to experimental measurements of the MAST U FILD [5] corresponding to prompt losses of an L-mode plasma heated by an on axis Neutral Beam Injector with an input power of 1.5 MW. This algorithm shows favorable performance when compared against zeroth order Tikhonov regularization. [1] M. Garcia-Munoz, et al., Review of Scientific Instruments 80 , 053503 (2009) [2] J. Galdon-Quiroga, et al , Plasma Phys. Control. Fusion 60 , 105005 (2018) [3] B. Schmidt, et al. al., Nucl. Fusion 64 , 076009 (2024) [4] J. F. Rivero-Rodriguez, et al. Review of Scientific Instruments 89 , 10I112 (2018) [5] L. Velarde et al, Plasma Phys. Control. Fusion (accepted)