Real-Time Prediction of Characteristic Parameters of Inductively Coupled Plasma Based on Artificial Neural Network

Artificial neural networks (ANNs) were implemented to predict the plasma parameter distribution to actively control the plasma absorption frequency band. Taking the inductively coupled plasma (ICP) fluid dynamics simulation results as the initial data, the general regression neural network (GRNN) is introduced to predict the plasma characteristic parameters under any external conditions. The electron density and electron temperature are measured by the Langmuir probe and compared with the fluid simulation results and model prediction results. It is found that the fluid simulation result of the electron density is higher than the experimental diagnosis result but slightly lower than the model prediction result. The fluid simulation result of electron temperature is lower than the experimental diagnosis result, but it is basically consistent with the model prediction result. Comparing the fluid simulation results, model prediction results, and experimental diagnosis results of the plasma characteristic parameters, grasp the distribution law of the characteristic parameters of the ICP. A prediction model for real-time prediction of plasma characteristic parameters is obtained. The rapid and active adjustment of the plasma characteristic parameter distribution is realized, and then, the active adjustment of the plasma absorption frequency band can be quickly realized.