Soft X-ray tomography using the optimized regularization method in Alvand Tokamak

•Reconstructing two-dimensional plasma emissivity distribution from the synthetic line integrated radiation measurements.•Determination of plasma shape and position in tokamak using the regularization method.•Study the accuracy of reconstruction method as well as influence of noise, using several profiles of soft x-ray emissivity.•Discretization of soft x-ray emissivity in the poloidal cross section of tokamak on square pixels in order to solve the tomography problem.•Reconstruction three distribution functions in tokamak plasma, including Gaussian, Hallow and Banana. The detection of Soft X-ray (SXR) radiation emitted during plasma discharge gives valuable information on tokamak plasma physics, position and shape of the plasma. Intensity of SXR radiation is routinely measured with SXR detectors. In this research, the arrangement and performance of such a system was investigated in Alvand tokamak. According to the diagnostic ports available on the system, the optimal placement has been obtained for the maximum coverage of the detectors. By choosing three synthetic plasma emissivity profiles, including Gaussian, Hollow and Banana profiles, the lined-integrated data reached to detector were simulated. Two-dimensional reconstruction of plasma emissivity was done using the Tikhonov regularization method. In this method, the regularization term and parameter were calculated using the Laplace matrix and iteration method, respectively. The minimum number of pixels for discretizing the poloidal cross section of plasma was 400. For this number of pixels, the root mean square derivation of reconstruction emissivity was obtained for Gaussian, Hallow and Banana models 0.016, 0.029 and 0.031, respectively. In addition, the convergence and stability of the reconstruction method was investigated by adding random noise to the primary data. For Gaussian and Hallow profiles, adding noise up to about 20 % has not significantly affected on reconstruction results. Meanwhile, in the Banana model, the reconstruction process will lose its effectiveness only for noise more than 2 %.