Data-driven sensitivity inference for Thomson scattering electron density measurement systems

Abstract

We developed a method to infer the calibration parameters of multichannel measurement systems, such as channel variations of sensitivity and noise amplitude, from experimental data. We regard such uncertainties of the calibration parameters as dependent noise. The statistical properties of the dependent noise and that of the latent functions were modeled and implemented in the Gaussian process kernel. Based on their statistical difference, both parameters were inferred from the data. We applied this method to the electron density measurement system by Thomson scattering for the Large Helical Device plasma, which is equipped with 141 spatial channels. Based on the 210 sets of experimental data, we evaluated the correction factor of the sensitivity and noise amplitude for each channel. The correction factor varies by ≈10%, and the random noise amplitude is ≈2%, i.e., the measurement accuracy increases by a factor of 5 after this sensitivity correction. The certainty improvement in the spatial derivative inference was demonstrated.

Estimation of plasma parameter profiles and their derivatives from linear observations by using Gaussian processes

“…In the case of the standard GPR, M in equation ( 2) is replaced by an identity matrix, and the elements of θ are defined at the observation locations. This regression predicts the values of θ where measurements are not available by using d [5,10,12,13]. On the other hand, the θ we consider in this GPR has sufficiently dense spatial points and is able to represent smooth profiles.…”

“…Since errors in the measurements and the choice of fitting functions typically affect the estimates of derivatives more than those of their original parameters, conventional fitting techniques cannot be employed reliably under many circumstances. To address these issues, Gaussian process regression (GPR) has been proposed and successfully applied to many profile measurement problems encountered in various scientific disciplines [3,5,[7][8][9][10][11][12][13][14]. GPR is a non-parametric regression technique in which each discrete point used to represent the parameter profile of interest has its own degree of freedom.…”

Robust Regression for Automatic Fusion Plasma Analysis Based on Generative Modeling

“…the noise variance (or its statistical quantity) is assumed to be uniform over all the data points. However, many advanced measurement data exhibit complex dependencies of the noise variance [15,16]. A typical example of this is Thomson scattering, which measures the electron temperature and density from the spectral shape of scattered light.…”

Developing an optimization algorithm for diagnostic modeling of optical emission spectroscopic measurement of non-equilibrium plasmas based on the argon collisional-radiative model

“…Fujii et al proposed a Thomson scattering measurement system for the large helical device by the experimental data analysis based on neural network. [18][19][20] In this way, various mathematical methods in information engineering will enlarge the strategy to diagnose plasmas in various states of non-equilibrium by OES examination, as the diagnostic modeling is, eventually, to solve the complex optimization problems. We must consider spectral intensities, detection specifications of photometric systems, plasma parameters of plasma sources, and so on.…”

“…The quantity, n i (T e−true , x true , N e−true )/g i was substituted into n g i i OES of Eq. (18). Further, the algorithm solved Eq.…”