Artificial neural network modelling of a gas sensor for liquefied petroleum gas detection

Currently, low power Metal Oxide gas Sensors (MOXs) are widely employed in gas detection because of its benefits, such as high sensitivity and low cost. However, MOX presents several problems, as well as lack of selectivity and environment effect. In this paper, it is presented an Artificial Neural Network (ANN) that models an MOX sensor (TGS 2610) used in a operating environment. The structure and the learning of the neuronal model are optimized by the Genetic Algorithms (GA). TGS 2610 is a type of gas sensor based on thin film semiconductor that associate very high sensitivity to Liquefied Petroleum gas (LP gas) with low energy consumption and long duration. This model includes dependence in LP gas like ethanol, hydrogen, methane and iso-butane/propane. Sensor modelling is used to avoid accidents that may be generated in practice, studying and analyzing problems in the simulation to avoid them in practice. It was proved in this paper that ANN technique was a powerful tool for modelling LP gas sensor. The comparative study of the results from ANN model with the experimental data shows a good agreement which validates the proposed models.