Here we present a new method that allows the determination of background neutral density profiles based on measurements from neutral particle analyzers (NPA). Bayesian optimization is used to obtain a reliable 5-parameter representation of the inferred profiles. The method has been benchmarked using forward modelling from FIDASIM against measured data. The systematic errors coming from assumptions made in the analysis were evaluated and they are lower than ≲5%. The new method has been tested using data from the ASDEX Upgrade tokamak. When using reconstructed neutral density profiles, good agreement is found between the measured NPA fluxes of neutralized fast ions with predictions based on TRANSP and FIDASIM. Moreover, a clear drop of neutral density is observed at the plasma boundary after edge localized mode (ELM) activity. As suggested by KN1D simulations, this drop is mainly due to an increase of scrape-off layer (SOL) ionization rate, resulting from higher SOL plasma densities and temperatures after the ELM crash. Moreover, the new method allows calculating the local plasma ionization source which will be important for future transport studies.

在这里，我们介绍一种新方法，该方法可以根据中性粒子分析仪（NPA）的测量结果确定背景中性密度分布。贝叶斯优化用于获得推断轮廓的可靠的5参数表示。该方法已使用FIDASIM的正向建模对测量数据进行了基准测试。对来自分析中的假设的系统误差进行了评估，且误差小于5％。新方法已经使用ASDEX升级托卡马克中的数据进行了测试。当使用重构的中性密度分布图时，在测得的中和的快速离子的NPA通量与基于TRANSP和FIDASIM的预测之间发现了很好的一致性。此外，边缘定位模式（ELM）活动后，在等离子体边界处观察到明显的中性密度下降。正如KN1D模拟所建议的，该下降主要是由于ELM碰撞后更高的SOL等离子体密度和温度导致刮擦层（SOL）电离速率的增加。此外，新方法允许计算局部等离子体电离源，这对于将来的运输研究将是重要的。