A gas sensor array was developed and evaluated using four high-frequency quartz crystal microbalance devices (with a 30 MHz resonant frequency in fundamental mode). The QCM devices were coated with ethyl cellulose (EC), polymethylmethacrylate (PMMA), Apiezon L (ApL), and Apiezon T (ApT) sensing films, and deposited by the ultrasonic atomization method. The objective of this research was to propose a non-invasive technique for acetone biomarker detection, which is associated with diabetes mellitus disease. The gas sensor array was exposed to methanol, ethanol, isopropanol, and acetone biomarkers in four different concentrations, corresponding to 1, 5, 10, and 15 µL, at temperature of 22 °C and relative humidity of 20%. These samples were used because human breath contains them and they are used for disease detection. Moreover, the gas sensor responses were analyzed using principal component analysis and discriminant analysis, achieving the classification of the acetone biomarker with a 100% membership percentage when its concentration varies from 327 to 4908 ppm, and its identification from methanol, ethanol, and isopropanol.

The source localization of gas leaks is important to avoid any potential danger to the surroundings or the probable waste of resources. Currently there are several localization methods using robotic systems that try to find the origin of a gas plume. Many of these methods require wind velocity information involving the use of commercial anemometric systems which are extremely expensive compared to metal oxide gas sensors. This article proposes the validation of the Gaussian plume model inside an empty room and its application to localize the source of a gas plume without employing anemometric sensors, exclusively using concentration data. The model was selected due to its simplicity and since it easily admits variants closer to reality, explaining the behavior of pollutants transported by the wind. An artificial gas source was generated by a conventional fan and liquid ethanol as contaminant. We found that the physical fan, far from making the model impossible to implement, enriched the information and added realism. The use of a robotic system capable of autonomously mapping the room concentration distribution is described. The results showed that the Gaussian plume model is applicable to localize our experimental gas source. An estimated position of the source with a deviation of 14 cm (6.1%) was obtained.