Abstract
A back-propagation artificial neural network algorithm is applied to a Mo X-pinch to estimate plasma parameters from typical L-shell spectra in the keV region. The spectrum was generated by a very compact LC-generator (40 kV, 200 kA) for driving 25-μm Mo two-wire X-pinches with a current rise-time of 200 ns. The neural network was trained over a set of synthetic spectra generated using a previously developed L-shell non-LTE collisionally radiative model. As a result, electron temperature and density of the Mo plasma were estimated as Te = 1000 eV and ne = 9 × 1020 cm−3. Furthermore, effects of electron beams on plasma parameters have been investigated through the inclusion of hot electrons in the kinetic model. The small fraction of hot electrons resulted in a better fitting of spectrum and decreased parameters Te = 650 eV and ne = 3 × 1020 cm−3.
•Modeling of L-shell Mo X-pinch plasmas using Artificial Neural Network proposed.•Training of ANN is based on collisional radiative model produced spectrums.•ANN is a good classifier and correlates the features in synthetic spectrums.•Electron temperature and density of XP601 is modeled Te = 1000 eV and ne = 9 × 1020 cm3.