Geochemical biodegraded oil classification using a machine learning approach

Posted on February 27, 2024 By LabLynx Journal articles

The use of chromatography methods in characterizing and interpreting phenomena related to oil has been around for a long time. However, reading chromatograms—particularly large quantities of them—can be time-comsuming. Bispo-Silva et al propose a more rapid process using modern deep learning and artificial intelligence (AI) techniques, increasingly being used in large companies. The authors land on using convolutional neural networks (CNN) for this work, particularly in the use of discriminating biodegraded oils from non-biodegraded oils. After presenting background on the topic, as well as their materials and methods, the authors present the results of using CNN and a variety of algorithms to classify chromatograms with both accurate and misleading training materials. The authors conclude that "CNN can open a new horizon for geochemistry when it comes to analysis by" a variety of gas chromatography techniques, as well as "identification of contaminants (as well as environmental pollutants), identification of analysis defects, and, finally, identification and characterization of origin and oil maturation."