A novel on-line microwave diagnostic for an atmospheric pressure air plasma based on an artificial neural network

Abstract

Permittivity is an important aspect of the design of microwave devices. In order to design microwave devices that are more suitable for the excitation of air plasmas, this paper presents a method for the on-line measurement of the relative permittivity of an atmospheric pressure air plasma with 2.45 GHz microwaves. First, a ridged waveguide is designed and constructed to excite the air. Using the ridged waveguide, the plasma can be excited and measured simultaneously. An artificial neural network trained with simulated data is integrated into the system to invert the dielectric constant of the air plasma being measured. At a power of 786 W and an airflow rate of 1250 L/h, the real and imaginary parts of the permittivity are obtained to be −1244 and 60, respectively. The forecasted errors on the real part and the imaginary part of the relative permittivity are kept at around 5% and 3%. The Drude model is used to construct a relationship between the electronic and dielectric properties. This diagnostic method is simple to operate and has a high degree of generality. It should be useful in the design of microwave plasma devices and can greatly improve the efficiency of microwave plasma equipment.