Multivariate Statistical Analysis of Spectroscopic Data Obtained From Food Samples

Abstract

Various types of additives are widely used in food processing. However, both the type and dosage of these additives need to be closely monitored in order to ensure minimal health risk to consumers. Carrageenans and Gums are the most commonly used additives in food industry. However, despite their popularity in food processing industry there have been few research studies reported that have aimed to characterise Carrageenans and Gums in the mixtures containing both of these types of food additives. Most of the work previously reported in literature has focused on analysis, either qualitative or quantitative analysis, of either Carrageenans or Gums only. However, these two types of food additives are frequently blended together, necessitating the need to investigate appropriate techniques to characterise them in mixtures containing both Carrageenans and Gums. Therefore, the work reported in this thesis has focused on characterising both Carrageenans and Gums using vibrational spectroscopic techniques to generate data that is subsequently analysed using various chemometric techniques. In order to characterise Carrageenans and Gums, their spectral samples are firstly collected using three different spectroscopic techniques, namely Fourier Transform (FT)-Raman, near-infrared (NIR) and mid-infrared (MIR) spectroscopy. Appropriate pre-processing methods are applied to the collected spectral data. Subsequently, chemometric methods are applied to the pre-processed data for the purpose of classification and quantification of three types of Carrageenans and three types of Gums. Qualitative models based on the principal component analysis (PCA) and partial least squares-discriminant analysis (PLS-DA) are developed using the pre-processed data. These qualitative models are shown to be able to correctly identify the dominant Carrageenans with level of accuracy for each standard food additive mixture and using data obtained from any of the three spectroscopic techniques. However, this is not the case with Gums. In fact, only NIR spectroscopy in conjunction with PLS-DA can identify the existing type of Gums in the mixtures with relatively level of accuracy. Quantitative models based on partial least squares (PLS) and support vector machine (SVM) are also deployed to estimate the concentrations of each type of ingredients in the mixture containing both Carrageenans and Gums using the spectral data obtained from the three types of spectrometers. Results indicate that NIR spectroscopy when used in conjunction with linear regression method, namely PLS, is able to accurately predict the concentrations of Carrageenans, whereas NIR spectroscopy coupled with non-linear regression method, namely, SVM, is able to provide accurate predictions of the concentrations of Gums. Generally, NIR spectroscopy coupled with chemometric methods can classify and quantify all six types of food additives with relatively high accuracy, whereas Raman spectroscopy is not recommended for the analysis of mixtures that contain both Carrageenans and Gums.

Date of Award	1 Aug 2019
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Barry Lennox (Supervisor) & Ognjen Marjanovic (Supervisor)