Chemical decomposition by normalization of millimeter-wave spectra

The sharp and distinct absorption spectra of chemicals at low pressures in the millimeter wave range of frequencies become broadened at high pressures. As a result, detecting and quantifying different chemicals at high pressures from the broadened and merged spectra are difficult. This paper proposes a method of decomposition based on the low pressure spectra. Normalized low pressure spectral amplitudes are used as features to train a neural network. The network is tested using the peak spectra obtained for an unknown plume of chemicals at high pressure. Initial tests conducted on simulated and experimental spectra of selected chemicals show that the decomposition results of the proposed method are dependent on the dominance of the chemicals in the mixture common to conventional methods