Rapid Fuel Analysis with Machine Learning and Infrared Spectroscopy

This study presents a novel approach that integrates Fourier-transform infrared (FT-IR) spectroscopy with advanced machine learning for rapid identification of contaminants in marine fuels. By combining deep feature extraction using convolutional neural networks (CNNs) with dimensionality reduction techniques such as principal component analysis (PCA) and partial least squares (PLS), and employing support vector machine (SVM) classifiers, our framework effectively handles high-dimensional spectral data. The CNN-PCA-SVM model with a radial basis function (RBF) kernel demonstrates robust performance in distinguishing between high-quality and substandard fuels, although its ability to accurately classify specific contaminant types remains limited. Future work will focus on refining feature extraction methods, optimizing hyperparameters, and exploring alternative deep learning architectures to improve impurity classification, ultimately supporting the development of portable, real-time monitoring systems.