Analysis of HL-2A charge exchange spectra using parallel genetic algorithm

Abstract

In this work, we present a new method based on parallel genetic algorithm (GA) for in-between shot data analysis of the Charge-Exchange (CX) spectra on the HL-2A tokamak. The neutral beam induced active CX spectra is a powerful ion diagnostic technique to provide spatially resolved ion temperature and rotation velocity measurements on fusion devices. Currently CX spectra obtained in HL-2A experiments are mainly analyzed by the CXSFIT code [A. D. Whiteford, et.al, 2007]. While the analysis itself is fast, its accuracy relies on proper setup of the initial values for the spectral fitting parameters. Time-consuming manual interventions are needed. In the new parallel GA code, a two-loop GA analysis is used to gradually update the fitting parameter search ranges, which enables automatic analysis. A parallel algorithm based on the Linux Message Passing Interface (MPI) cluster is adapted to speed up the process. In a test run, for a set of 1600 data slices, the total time elapsed with 8 CPU nodes is about 310 s (0.2 s per data slice), which is efficient for in-between shot analysis on HL-2A. The uncertainty calculations using virtual CX signals with a noise level up to 5% show that the accuracies for ion temperature and rotation velocity are better than 10.14% and 2.14%, respectively. The ion temperature and rotation velocity obtained by applying the new parallel genetic algorithm on experimental CX data show good agreement with the conventional CXSFIT results.