PyTorch-based deep learning model for heat exchange of plasma actuator through the air thrusted flow

The activation energy of plasma discharge in dielectric barrier discharge (DBD) plasma actuators is consummated in two distinct ways, when the actuator is working as a momentum thruster, namely induced trust and heat generated. As a thruster the electric discharge must be maintained between the breakdown and the filamentary conditions. In this range the heat generated becomes an enormous part input energy. In recent years, due to the utilizing of plasma actuators as flow controllers, the idea of using them as a heat source has also received much attention. The heat generated from an active plasma thruster can be divided to two parts namely the heat transfer to neutral gas and the heat loss in dielectric. Regarding to recent studies the gas heat transfer is the major part of the total heat loss, about ۵۰% to ۹۰% depending on dielectric characteristics. In this work, an artificial neural network (ANN) model is developed in Python using Pytorch modules to predict the heat transfer rate to the passing gas upon some input variables namely applied voltage, dielectric thickness, dielectric dissipation, and the dielectric permittivity constant. For this purpose, the data set of the experiments of Rodrigues et al. [۱] is used to be clustered and modeled using deep artificial neural network. Trying to regulate the number of hidden layers, number of perceptron, seed number, and the learning rate gained to an optimum ANN model. The model undergoes validation by experimental tested results. The test results showed the adequate goodness of fit (GoF) upon multiple error criterions.