Deep Neural Network-Assisted Efficient Calculation on CO₂ Pulsed Discharges Under Martian Pressure for Various Pulse Rise Rates

In this study, a deep neural network (DNN) is developed to optimize the numerical simulation on <tex-math notation="LaTeX">CO₂</tex-math> pulsed discharges under Martian pressure for various pulse rise rates from 16 to 30 <tex-math notation="LaTeX">V/ns</tex-math> , and the training dataset is obtained from the fluid model or experimental observation. The validity and feasibility of the DNN are verified by comparing the prediction results of the DNN with the experimental measurements and fluid simulation results. The DNN takes only about 0.01 s to yield the discharge characteristics and plasma chemistry in <tex-math notation="LaTeX">CO₂</tex-math> pulsed discharges at any pulse rise rate in a given range with good prediction accuracy, resulting in about <tex-math notation="LaTeX">10⁷</tex-math> fold improvement in computational time. According to the predicted results, the peak discharge current and breakdown voltage increase almost linearly with the pulse rise rate. The growth of the pulse rise rate can significantly strengthen the electric field in the cathode sheath region. In addition, the densities of charged particles, neutral particles, and vibrationally excited <tex-math notation="LaTeX">CO₂</tex-math> molecules also increase with the pulse rise rate. This study indicates that the numerical simulation of pulsed discharges can be efficiently enhanced by introducing DNN even with a very sharp pulse rise rate and hundreds of reactions, to gain more insights on the discharge mechanisms with plenty of data from DNN.