Simulating nitrogen molecular plasma spectra using artificial neural networks

We introduce a method based on artificial neural networks (ANNs) to simulate molecular spectra under experimental conditions. The spectra of a nitrogen discharge were measured under variable pressure ranging from 0.3 to 2.0 Torr and discharge current varying from 5.0 to 50.0 mA and they have been used as a training set. Our procedure uses a system of two ANNs. The first one extracts the physical characteristics of the training set. The second one uses the information collected from the first to simulate the spectra under conditions that have not been presented in the training sample. The parameters used to assure the quality of the simulated spectra were the temperatures. Our simulated spectra have agreed with those measured experimentally. These results have indicated the robustness of the ANN in simulating molecular spectra. These facts stimulated us to investigate the ability of our system in generating spectra under different conditions of experimental resolution. The results show that our method could be used to improve the accuracy of experimental temperature whenever the experimental spectrum has a poor resolution.