期刊
REVIEW OF SCIENTIFIC INSTRUMENTS
卷 89, 期 8, 页码 -出版社
AIP Publishing
DOI: 10.1063/1.5035416
关键词
-
资金
- EURATOM research and training programme 2014-2018 [633053]
Two methods for fast analysis of Collective Thomson Scattering (CTS) spectra are presented: Function Parametrization (FP) and feedforward Artificial Neural Networks (ANNs). At this time, a CTS diagnostic is being commissioned at the Wendelstein 7-X (W7-X) stellarator, with ion temperature measurements in the plasma core as its primary goal. A mapping was made from a database of simulated CTS spectra to the corresponding ion and electron temperatures (T-i and T-e). The mean absolute mapping errors are 4.2% and 9.9% relative to the corresponding T-i, for the ANN and FP, respectively, for spectra with Gaussian noise equivalent to 10% of the average of the spectral maxima in the database at 650 sampling points per GHz and within a limited parameter space. Although FP provides some insight into the information contents of the CTS spectra, ANNs provide a higher accuracy and noise robustness, are easier to implement, and are more adaptable to a larger parameter space. These properties make ANN mappings a promising all-round method for fast CTS data analysis. Addition of impurity concentrations to the current parameter space will enable fast bulk ion temperature measurements in the plasma core region of W7-X.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据