Resumen de Low-power techniques for wireless gas sensing network applications: pulsed light excitation with data extraction strategies
About 1 out of 8 yearly death are attributed to diseases related to air pollution worldwide. Having gas sensors available to monitor pollutant gases is mandatory to establish control strategies to reduce air pollution. Since 90 % of the world population is exposed to air pollution levels that provoke a significant health impact the society needs to have available environmental monitoring systems worldwide. Low-cost and portable sensors technologies are needed to develop monitoring solutions with enough spatial resolution. Nevertheless, gas sensing systems based on the most accurate operating principle such as spectroscopy analysis and gas chromatography normally present high cost and difficulty for their miniaturization. Moreover, some of these techniques require laboratory analysis of a previously acquired sample, which is a limitation for the scalability of monitoring systems and the data availability.
On the other hand, other gas sensors with operating principle based on optical, acoustic, or electrical properties variation have been used in the development of gas sensing systems. Non-dispersive infrared and resonant acoustic wave gas sensors have been designed obtaining small size and MEMS sensors which can be used in portable systems. Although these techniques provide long lifetime sensors with good sensitivity, the fabrication process require the use of expensive techniques which influence in the sensor cost. In addition, other technologies, such as electrochemical and chemiresistive sensors have been studied because these can be employed to develop portable and low dimensions analysers due to its relatively simple miniaturization, which can be endowed with low power consumption and communication capabilities, thus making possible the development of monitor gas sensing systems remotely. In this sense, chemiresistive sensors, such as metal oxide (MOX), perovskite oxide, Transition Metal Dichalcogenides (TMDC), and carbon nanomaterials sensors have been used to develop portable and even wearable gas sensing systems due to their simple preparation, low cost, simplicity of measurement systems, and relatively high performance. Chemiresistive sensors operating principle is based on the variation of their electrical resistance in presence of target analytes in the surrounding environment.
One of the two principal research line of this thesis was focused on the development of a concentration quantification methodology based on a pulsed light modulation of the sensor resistance and the use of numerical method to perform the prediction models. First, a Fast Fourier Transform (FFT) analysis is performed on the time domain sensor signal. It is found that fundamental frequency from the light switching and its even order harmonics have a relevant magnitude in the FFT spectrum. This frequency components are used to build the matrix used to do a Principal Component Analysis (PCA). Principal Components scores and loadings obtained in the PCA are used to carry out both qualitative and quantitative analysis of the tested gases. The PCA scores plot have been used to identify different gases since observation from different species are spatially separated in clusters. PCA scores and loadings biplots have been used to optimize the number of frequency component used to perform the linear regression mechanisms. Furthermore, linear regression methods as Principal Component Regression (PCR) and Partial Least Squares (PLSR) are employed to quantify the gas concentration.
The pulsed light modulation methodology was applied on different n-type and p-type chemiresistive sensors, using different materials, specifically MOX (WO3), metal transition dichalcogenides (WS2), perovskite oxide (SrTiO3@WO3), and carbon nanomaterials (Au@CNT). It was demonstrated that using n-type sensor led to obtain better quantification performance towards oxidizing gases, while p-type exhibit higher accuracy to quantify reducing gases. Light sources with wavelength from the UV to the visible spectrum (325, 365, and 410 nm) were used to activate the sensing layer of the sensors employed, obtaining better performance of the prediction models when the system operates under visible light modulation. This suppose an improvement in the system cost and power consumption since generally UV light sources are more expensive and power consumer than visible ones. The sensor surface activation mechanism used in this thesis represents a reduction in power consumption of about 90 % as compared with the traditional heating mechanism working at an operating temperature of 250 ºC. This power reduction is applicable to sensors developed on the commercial alumina sensors used in this work. Models developed using FFT analysis, PCA and linear regression techniques present high prediction performance with R-squared values up to 0.98 and RMSE values lower than 10 % of the total concentration range measured. These results open an opportunity for using non-MEMS chemiresistive sensors in real gas sensing applications, being part of low-cost, low-power, and portable monitoring systems.
The second research line addressed the design and implementation of a LoRa-based gas sensor network for Air Quality Monitoring (AQM) and gas leakage detection. The system is composed of low-cost and low-power LoRa nodes with sensing capabilities (temperature, humidity, and oxidizing and reducing gases), and a LoRa internet gateway. The sensor data management was also implemented (data transmission, storage, monitoring). The sensor data is sent periodically to a cloud server where the data is stored and replicated to a local server. The data monitoring system was developed using a web service developed using open-source software and hosted in a Raspberry Pi. The inclusion of commercial and lab-synthesized sensors allowed to properly detect oxidizing and reducing gases. Having a resistance measurement channel to monitor the lab-synthesized sensor behaviour opens the opportunity of using this system in a wide range of applications where leakage of pollutant or hazardous gases are prone to occur by selecting the sensing material according to the target gas monitored. The sensing nodes present hot-plug capabilities in the network since these only need to be registered on the cloud server, which favors the scalability of the sensor network. Hence, this Wireless Sensor Network (WSN) structure can be used for AQM purpose in crowded places as airports or train stations where air pollution control or even the prevention of chemical threats can save lots of human lives.
