Master Degree / Yüksek Lisans Tezleri
Permanent URI for this collectionhttps://hdl.handle.net/11147/3008
Browse
8 results
Search Results
Master Thesis In-Depth Investigation of the Effects of Different Preprocessing Strategies on Infrared Spectroscopic Data(01. Izmir Institute of Technology, 2021) Deniz, Elin İlayda; Özdemir, DurmuşWhenever collecting experimental data, they may contain several spurious sources of variability which can hinder the extraction of the desired relevant information so that it is rarely the case that they can be processed as such by chemometrics approaches. In recent years, pre-processing techniques has become an integral part of chemometrics modeling with the purpose of providing better endmodels through fundamental knowledge for Near-Infrared (NIR) users. The aim of pre-processing techniques is to improve success of multivariate regression by reducing undesired physical phenomena in the spectra. This thesis describes the theory of present pre-processing techniques and compares the qualitative and quantitative results of their application. Mean centering, scatter-correction methods and spectral derivatives are the instances of those pre-processing techniques used in this thesis. Those techniques and combinations of them have been applied in order to find the best pre-processing strategy. To be able to observe the results and compare the effects of the applied methods, Partial Least Squares and Genetic Inverse Laast Squares are carried out as multivariate calibration methods. When comparing the calibration results of raw data with pre-processing techniques applied data, decreasement standard eror of prediction (SEP) values observed after appliying those techniques, which is good manner. However, better in comparison, t-Test: Paired Two Sample for Means applied. The results demontrates that there is no significant difference within 95% confidence level.Master Thesis Development of Chemometrics Method Based on Infrared Spectroscopy for the Determination of Cement Composition and Process Optimization [master Thesis](01. Izmir Institute of Technology, 2021) Tepeli, Dilek; Özdemir, DurmuşCalcium, silicon, aluminum, iron oxide-containing raw materials are used in controlled portions to manufacture cement. (How cement Is Made, n.d.) This mixture is first converted to clinker, obtained by heating the mixture to 1500oC; some additives are added and ground to obtain cement. Depending on the purpose, various types are produced, and therefore, the determination of cement composition is an essential task for the quality consideration and the sustainability of the production processes. The quantitative analysis of cement is performed with X-ray fluorescence spectroscopy. However, XRF generally requires tedious and lengthy analysis times. In this study, quantitative determination of the raw materials, intermediate products, and types of cement by using Fourier transform infrared spectroscopy coupled with chemometrics multivariate calibration method is aimed, which could be an alternative for the current XRF technique. Samples were collected from a local cement factory that has been in the sector for several years. Reference analyses of the samples were performed at the quality control laboratory of the same factory. The same samples were analyzed by the FTIR-ATR spectrometer. The resulting FTIR spectra combined with XRF reference composition data were used to construct calibration models using the partial least squares method (PLS). Based on the obtained results, the proposed method could generate quite successful results for the quantitative determination of all types of products used to produce cement. The regression coefficients (R2) of the PLS models vary from 0.95 to 0.99. The standard errors of cross-validations were found as from 0.21 to 1.42 (w/w%).Master Thesis Determination of Crucial Parameters in Gasoline Blends by Using Infrared Spectroscopy Coupled With Multivariate Calibration Methods(01. Izmir Institute of Technology, 2021) Sakallı, Fatma Nur; Özdemir, DurmuşIn petroleum refineries, converting the manual gasoline blending system to an automatic inline blending system provides the most economical blending in gasoline production, increasing efficiency, and reliability. The most important requirement for an automatic inline blending system is the determination of gasoline parameters in a short time with high reliability. For this purpose, fast and simple analytical methods have been developed to determine crucial parameters of gasoline blends by using Fourier Transform Infrared Spectroscopy (FTIR) coupled with multivariate calibration methods which are Partial Least Squares Regression (PLSR) and Genetic Inverse Least Squares Regression (GILS) for this study. Turkey Petroleum Refinery Incorporated Company (TUPRAS) Izmir Refinery collected all gasoline samples and tested them using reference test methods at Quality Control Laboratory. Since commercial product samples were used in this study, the data ranges of the parameters were quite narrow. The Standard Error of Cross-Validation (SECV) and Standard Error of Prediction (SEP) values were acceptable, although the determintion coefficient (R2) value of some parameters was below the expectation. It has been observed that the prediction results of GILS are better in these parameters, whose R2 value is low because the data range is very narrow. In the comparison made with the reproducibility values specified in the reference measurement methods, it was determined that the calibration model results of most parameters were acceptable. Collecting more samples in a longer time interval to expand the data range of the parameters, or preparing a data set with experimental design can improve the prediction performance.Master Thesis Estimation of Low Sucrose Concentrations and Classification of Bacteria Concentrations With Machine Learning on Spectroscopic Data(Izmir Institute of Technology, 2019) Mezgil, Bahadır; Baştanlar, Yalın; Baştanlar, YalınSpectroscopy can be used to identify elements. In a similar way, there are recent studies that use optical spectroscopy to measure the material concentrations in chemical solutions. In this study, we employ machine learning techniques on collected ultraviolet-visible spectra to estimate the level of sucrose concentrations in solutions and to classify bacteria concentrations. Some metal nanoparticles are very sensitive to refraction index changes in the environment and this helps to detect small refraction index changes in the solution. In our study, gold nanoparticles are used and we benefited from this property to estimate sucrose concentrations. The samples in different low sucrose concentration solutions are obtained by mixing the sucrose measured with precision scales with pure water and then the UV-Vis spectrum of each sample is measured. For the bacteria concentration solutions, spectra for six different bacteria concentrations are captured. Spectra of the same solutions are also captured before adding the bacteria. For each of these solutions, four sets are prepared where gold nanoparticles are not grown (minute 0) and grown for 4 minutes, 10 minutes and 12 minutes. After the dataset preparation, these spectrum measurements are transferred into MATLAB environment as sucrose concentration dataset and bacteria solution dataset. Then the necessary preprocessing steps are performed in order to get the most informative and distinguishing information from these datasets. The raw measurement values and processed spectrum measurements are trained with shallow Artificial Neural Networks (ANN) on MATLAB Deep Learning Toolbox and Support Vector Machine (SVM) on MATLAB Statistics and Machine Learning Toolbox. When the results of the conducted machine learning experiments are examined, success rate is promising for the estimation of sucrose concentrations and very high for classification of bacteria concentrations in pure water solution.Master Thesis The Development of Chemometric Methods Based on Molecular Spectroscopy for the Standardization of Production Processes and Product Traceability of Personal Care and Cleaning Products(Izmir Institute of Technology, 2019) Çiftçi İlmek, Berfu; Özdemir, DurmuşPersonal care and cleaning products are the main consumer goods. Changes in our heath caused by all of the chemicals that we exposed to everyday if these products are not produced according to the regulations and determined formulations. Because of this reason, quality control of the product formulation quantitatively is very important. There are some analytical methods for the determination of anion active matter, nonionic matter and total active matter in the product mixture. However, these techniques are expensive and do not give accurate results. The purpose of this thesis principally based on development of rapid, accurate and practical infrared spectroscopic technique based on multivariate chemometrics data analysis methods for the standardization of production processes and product traceability of personal care and cleaning products. In this thesis, two different products are studied which are namely liquid soap and shower gel. Fourier Transform Infrared spectroscopy coupled with Attenuated Total Reflectance accessory based chemometrics multivariate calibration models were developed for the quantitative determination of liquid soap and shower gel compounds. Genetic Inverse Least Squares was used as the chemometrics method for the development of multivariate calibration models in the quantitative determination of liquid soap and shower gel compositions. Standard error of cross validation and standard error of prediction values for content of the liquid soap samples were found 0.26% and 0.21 % (w/w %), respectively. Standard error of cross validation and standard error of prediction values for content of the shower gel samples were found 0.27 % and 0.30 % (w/w %), respectively.Master Thesis Development of a New Infrared Spectroscopic Method Based on Multivariate Calibration for the Determination of Aluminum and Magnesium Oxid Thickness on Aluminum Foil and Sheets Surfaces(İzmir Institute of Technology, 2016) Meşe, Ayten Ekin; Özdemir, DurmuşSurface oxidation is a general problem for certain industrial applications such as coating and painting of the finished rolled products. A detailed understanding for the oxide growth mechanism as well as the development of a simple analytical method to measure this oxide thickness is very important in aluminum rolling industry and this study aims to develop a spectroscopic method to determine the oxide thicknesses on the surface of the aluminum by using multivariate calibration and infrared spectroscopy. Two main series of different aluminum alloys (3005 and 3003BZ) were selected in this study to develop a proposed methodology which is based on the combination of Fourier Transform Infrared Spectroscopy (FTIR) with Grazing Angle ATR accessory and chemometrics multivariate calibration techniques. In order to obtain oxide thickness values, X-ray Photoelectron Spectroscopy (XPS) was used and aluminum oxide (Al2O3) and magnesium oxide (MgO) thicknesses determinations were carried out by two different multivariate calibration models which are Genetic Inverse Least Squares (GILS) and Genetic Partial Least Squares (GPLS). These models were able to predict Al2O3 and MgO thicknesses using FTIR that is faster, easier and cheaper to operate as well as from XPS. The correlation coefficients of XPS reference oxide thickness values versus FTIR-GATR based GILS and GPLS predicted values were better than 0.919 in range of 0 to 25 nanometers for Al2O3 and 0 to 35 nm for MgO. These results suggest that grazing angle FTIR-ATR spectroscopy may offer a simple and nondestructive alternative for quick determination of oxide layer thickness.Master Thesis Prediction of Extractives and Lignin Contents of Anatolian Black Pine (pinus Nigra Arnold. Var Pallasiana) and Turkish Pine (pnus Brutia Ten.) Trees Using Infrared Spectroscopy and Multivariate Calibration(Izmir Institute of Technology, 2008) Karaman, İbrahim; Özdemir, Durmuş; Özdemir, DurmuşDetermination of quality parameters such as extractives and lignin contents of wood by wet chemistry analyses takes long time. Near-infrared (NIR) and mid-infrared (MIR) spectroscopy coupled with multivariate calibration offer fast and nondestructive alternative to obtain reliable results. However, due to complexity of multi-wavelength spectra, wavelength selection is generally required. Turkish pine and Anatolian black pine are the most growing pine species in Turkey. Forest products industry has widely accepted use of these trees because of their ability to grow on a wide range of sites and their suitability to produce desirable products. Determination of extractives and lignin contents of wood provides information to tree breeders when to cut and on how much chemical is needed in pulping and bleaching process. In this study, 58 samples of Turkish pine and 51 samples of Anatolian black pine were collected to investigate the correlation between NIR and MIR spectra of these samples and their extractives and lignin contents which were determined with reference methods. Genetic inverse least squares (GILS) was used for multivariate calibration. Standard error of calibration (SEC) values were less than 1.86% (w/w) for lignin and 1.19% (w/w) for extractives whereas standard error of prediction (SEP) values were less than 3.81% (w/w) for lignin and 2.04% (w/w) for extractives. Resulting R2 values for calibrations were larger than 0.8. Classification for Turkish pine and Anatolian black pine samples was performed by genetic algorithm based principal component analysis (GAPCA) and these two pine species were classified by using NIR and MIR spectra.Master Thesis Varietal Classification and Prediction of Chemical Parameters of Turkish Wines by in Frared Spectroscopy(Izmir Institute of Technology, 2010) Öztürk, Burcu; Özen, Fatma Banu; Özen, Fatma BanuThis study was performed with the aim of varietal classification of mono-varietal Turkish wines and development of models to predict basic enological parameters from mid-IR spectra with the use of chemometric methods. Mid-infrared (MIR) spectroscopy combined with multivariate data analysis was employed to make a varietal classification of commercial Turkish wines (Boğazkere, Cabarnet Sauvignon, Çalkarası, Kalecik Karası, Merlot, Öküzgözü, Papazkarası, Shiraz, Emir, Misket, Narince, Sultaniye and Chardonnay) from 2006 and 2007 vintages. Wine samples (n.79) including red, rose and white wines were scanned in the mid-IR region (4000-650 cm-1) and three spectral regions (965-1565 cm-1, 1700-1900 cm-1 and 2800-3040 cm-1) were used to classify wines on the basis of grape variety. The principal component analysis (PCA) was applied to the spectral data of the wine samples. Although a clear classification could not be achieved according to varieties, almost complete classification of red and white wines was observed. For the quantification analysis, a total of eleven enological parameters, including total phenol and anthocyanin content, pH, brix, titratable acidity, colour intensity (CI), tint, yellow%, red%, blue% and the proportion of red colour produced by anthocyanins (dA%) were determined with analytical reference methods. Correlation between the results of the reference methods and MIR spectral data was tested with partial least square (PLS) regression analysis and prediction models were developed with the use of these correlations. The calibration and validation sets were established to evaluate the predictive ability of the models. As a result of PLS analysis, the best models were developed for total phenols and CI with excellent predictions (R2.0.93 and 0.89, respectively and residual predictive deviation RPD.3.68 and 3.83, respectively). The model of pH determination and yellow% gave a good prediction (R2.0.85 and 0.85, respectively and RPD.2.7 and 2.04, respectively).