Fast Electron Temperature Diagnostics on EAST Using W⁴⁵⁺ Spectral Line Ratios and a Convolutional Neural Network

摘要: Tungsten (W) spectral line emission from highly charged ions provides an important diagnostic for electron temperature in high-temperature fusion plasmas. In particular, W⁴⁵⁺ spectral line intensity ratios exhibit strong sensitivity to electron temperature. However, conventional spectroscopic methods for electron temperature diagnostics typically rely on iterative spectral fitting or collisional-radiative modeling, which can be computationally expensive and limit their applicability for fast analysis. In this work, a new fast electron temperature diagnostic scheme is developed based on W⁴⁵⁺ spectral line intensity ratios measured by the X-ray crystal spectrometer (XCS) on the Experimental Advanced Superconducting Tokamak (EAST). The temperature dependence of the selected line ratios is validated using Flexible Atomic Code (FAC) and FLYCHK calculations, confirming their suitability for temperature diagnostics. A convolutional neural network (CNN) is employed to establish a direct mapping from spectral features to electron temperature, enabling rapid inference without iterative modeling. The method is validated against electron cyclotron emission (ECE) measurements, showing good agreement in both temporal evolution and spatial profiles. Compared with conventional approaches, the proposed method significantly reduces computational cost while maintaining high accuracy. These results demonstrate that W⁴⁵⁺ spectral line ratios provide an effective and efficient observable for electron temperature diagnostics, with strong potential for real-time applications in fusion plasmas.