A determinação de cada espécie que compõe uma mistura gasosa tem sido alvo de muitas pesquisas. A fim de viabilizar uma aplicação de baixo custo para a determinação da concentração de espécies em uma mistura gasosa, neste trabalho é proposto um aparato para sensoriamento de H 2 , CH 4 e CO encontrados em gases combustíveis provenientes de gaseificadores. ABSTRACTThe achievement of the content of each component of a gas mixture from gasifiers has been a matter of several studies. In order to allow a low cost application for obtaining the concentrations in a gas mixture, this study proposes a set up for sensing, H 2 , CH 4 and CO found in fuel gases produced by gasifiers. The sensing is performed by commercially avaliable chemiresistors of SnO 2 . The proposed set up has a gas dilution system before proceeding the sensing step. The 125 different gas mixtures were using nitrogen as the carrier gas. The samples were evaluated under two different methods for sensor recovery: forced and natural. Based on the results, it was established that: the cross sensitivity of the CO and CH 4 sensors is too high while the H 2 sensor presents selectivity almost 100% and the natural recovery method showed improved results because of the better thermal stability of the system. An artificial neural network was developed with the purpose of overcoming the problem of cross sensitivities and indicated a maximum error of 0.1% for H 2 , 23% for CH 4 and 29% for CO.