Characterization of a photoacoustic system through neural networks to determine multicomponent samples

• New signal processing for measurement of concentrations of different gases in a mixture.
• Original application of artificial neural network.
• New contributions to photoacoustics signal processing.

Photoacoustic spectroscopy for trace gases detection, based on a CO2 laser, can be used in a wide range of applications. The tunability of this laser in the mid-infrared (9.4–10.6 μm) allows the quantitative determination of different substances in multicomponent samples. In general, at traces level, the total photoacoustic amplitude at a certain wavelength may be approximated by a linear superposition of the amplitudes given by each of the species absorbing at that wavelength. However, in some cases, the sum of the individual signals is no longer valid. In particular, it is known the presence of CO2 delays the acoustic signal in relation to the laser excitation due to the exchange of vibrational energy between CO2 and N2. This phenomenon generates a slow V-T energy relaxation from a metastable N2 vibrational level and the sum of individual contributions may no longer be valid. Moreover, the resolution of a linear equation system has limitations, so the possibility to determine concentrations in photoacoustics based on neural network is proposed in this work. This procedure is tried in a particular case of a volatile organic compound, such as C2H4, and CO2 in air. The results are compared with the ones obtained with a model based on rate equations.