GAS TYPE DETECTION AND CONCENTRATION ESTIMATION USING THERMAL MODULATED RESISTIVE SENSOR AND NEURAL NETWORKS

In this paper, a new processing sensor data method base on neural networks and principal component analysis block is presented in order to identify the gas type and to estimate the gas concentration. Three gases in thirteen different concentrations have been examined including methanol, ethanol, and 2-propanol. For temperature modulation, the stair-case voltage was applied to the sensor heater at spans of 40s in 200s. In each of the obtained curves, at any span, transient and steady state responses were recorded. These recorded properties are analyzed using the usual methods of pattern recognition. Principal component analysis was used to increase the selectivity of the sensor and the neural network was used to recognize the type and estimate the gas concentration. In this study, we have achieved the separation of gases successfully as well as average estimation error concentration was calculated to be 0.00358%.