Development of new applications in industrial backgrounds based on Middle Infrared spectroscopy (MID IR) using low cost and uncooled sensors

Air pollution due to the greenhouse emissions of pollutants by vehicles, is an issue whose gravity is increasing, and it gains more significance for big cities such as Madrid. Controlling the composition of vehicle fuel can reduce pollutant emissions from combustion and it can also improve the performance of the vehicle engine. The intentional addition of off-specification fuel or contaminants can occur during the distribution of fuels and it is difficult to detect, unless the fuel is analyzed at every stage or the fuel quality is carefully monitored. Establishing a fuel effective compliance program regarding specifications is important to ensure that fuels sold at gas stations meet all the requirements, and at the same time it is challenging because fuel is manipulated by many different agents along the distribution chain. It may also result costly, as the equipments used to measure the fuel composition are based on laboratory chromatographs and spectrometers. These techniques need of highly qualified staff, and large amount of samples, time and laboratory reagents consumption. Consequently, it becomes necessary the use of simple, fast, and accurate techniques to carry out this quality control process. Although optical methods (non destructive) such as Infrared and RAMAN Spectroscopy have been widely applied and studied, they still pose drawbacks such as the lack of in-field applicability and the high cost of the equipment currently employed for it. In this PhD thesis a new device is proposed and developed for the quality control of pure fuel and its blends with alcohols (additives and adulterants). It consists on complementing or replacing current laboratory techniques by an innovative compact MID IR sensor based on a combination of a linear-array of detectors made of Vapour-phase-deposited Lead Selenide (PbSe) coupled to a linear variable filter of wavelengths selection between 2.9 and 4.5 microns (3450-2222 cm-1). This device provides a clear and quick analysis of fuel, either pure or blended with alcohols, using the specified wavelength range. A simple, fast and accurate method has been developed to determine fuel composition in three single steps: the development of the device set-up; the recording of the spectra with optimized parameters; and a fast processing of the data based on chemometrics as to being able to easily get implemented at any stage of the fuel distribution chain. The results obtained in the first part of this thesis demonstrated the prospective of this device as a tool to differentiate pure chemical substances such as diesel oils, gasoline, and alcohols with increasing hydrocarbon chain length from 1C (Methanol) to 6C (n-Hexanol) by applying an appropriate pre-processing method in order to remove the noise and the drift from the spectra. A Principal Components Analysis provided a reduction in the dimensionality of the data and a preliminary differentiation between diesel oils, gasoline and alcohols when plotting PC1 (retaining 95.35 % of variance) against PC2 (4 %), PC3 (0.21 %) and PC5 (0.07 %). The most remarkable fact regarding these results was the relevancy of the information gathered by the PC2, PC3 and PC5; whilst the higher amount of information was retained by PC1 (95 % of variance). Afterwards, a MANOVA analysis was applied on the scores from PC1 to PC5, and the two first canonical variables gave a neat segregation among pure substances when graphically represented in a Cartesian diagram. A total of 75 spectra were used for doing this calibration of the method. These results were validated using the projection of new data sets (420 spectra) achieving a 100 % of successfully classified individuals. The key wavelengths used for this analysis were 3035, 3101 and 3145 cm-1 (3.3-3.2 microns) by the variable canonical 1, and 3035 and 3145 cm-1 by the canonical 2. The second part of this thesis was based on the experiments conducted to assess the capacity of the device and method, set up for simultaneously quantifying and identifying the type of alcohol in gasoline and the corresponding alcohol mixtures in a 10, 20, 30 % (v/v). After the spectra pre-processing by a smoothing and baseline correction, two different approaches were followed in order to quantify and determine the type of alcohol contained in the blend. Quantitative analysis was done by applying a PLS regression using 253 and 134 spectra for the train and test sets respectively. The results obtained for the training and validation data sets gave a goodness of fitting of 0.95 in both of them, with a Root Mean Square Error of Calibration and Cross Validation of 7.4 and 8.5 % respectively, using 5 latent variables to build the model. These results demonstrated the ability of the sensor for quantitatively predicting concentrations of alcohol up to 10 %. In parallel, a PCA was performed on the pre processed spectra followed by a MANOVA analysis that gave a segregation triangular pattern, where each vertex of the triangle corresponded to a type of alcohol and it was qualitatively determined by the angles values (173° for methanol, 57.1° for ethanol and 304° for n-butanol in the calibration data set, and 177.4°, 55.05° and 306.3 ° respectively for the validation data set) and the radial distance measurement of each cluster centroid to the centre of the triangle gave a parallel way to determine the concentrations of each alcohol. This method was validated with the projection of a new data set (97 spectra) on the loadings obtained from the calibration set (181 spectra). For this analysis, the spectra of pure gasoline were not taken into account, and neither the 40 % methanol and ethanol mixtures, as they did not result of practical interest for the analysis. The experimental set up implemented in this thesis together with the search of the most appropriate data analysis methods provided a new fast and accurate optical method to determine and quantify pure chemical substances and their blends with a prospect of being applied in the near future, at the fuel distribution chain