Application of artificial neural networks to calculate the partial gas concentrations in a mixture

A way of improving the selectivity and sensitivity of semiconductor gas sensors is to use a multisensor array and analyze the whole response using pattern recognition methods, such as artificial neural network models (ANN). We use these models, not only to detect the individual components of the gas mixture (NO2 and CO), but also to measure the concentration of both gases with sufficient accuracy. We present a systematic study to enhance all parameters of the neural network, including pre-processing techniques. Finally, we have implemented the enhanced neural network into a 68HC11 micro, showing the NO2 and CO concentrations in real time on a digital display.