Vibrating capacitor mapped chemical picture classification by artificial neural network

Potential maps of vibrating capacitor Kelvin measurements vary according to the gas surrounding the surface of the “artificial olfactory epithelia”. In the case of low gas concentration, the measured work function shifts, thus the measured potentials are proportional to gas concentrations around the surface. Maps of more different gas concentrations were characterised by 12-dimensional vectors, and neural networks were trained with them. The neural network was able to recognize which gas and concentration belong to a set of measured voltages for hydrogen (H2) and propane-butane (PB) gases. The two-layer backpropagation feedforward neural networks were simulated in Matlab.