Determination of alcohols-diesel oil by near infrared spectroscopy based on gramian angular field image coding and deep learning

Abstract

Alcohols blended with diesel, as a renewable and clean substitute fuel of diesel engine, is a potential solution to alleviate the scarcity of fossil fuels and the worsening environmental pollution in transportation and industry. In this study, a novel near infrared (NIR) application for determination of alcohols-diesel is presented. Using the strategy of combining NIR spectroscopy with gramian angular field (GAF) image coding and deep convolution neural network (CNN), the proposed approach successfully realized the qualitative classification of different diesel (methanol diesel, ethanol diesel and pure diesel) and alcohols content detection despite the spectra were highly similar and collinear. To further verify the practical performance of the proposed method, it was briefly compared with 1DCNN, support vector machine (SVM) and BP neural network based on the identical data. The proposed GAF-CNN method can accurately distinguish diesel, methanol diesel and ethanol diesel. In addition, satisfactory results have been achieved with the smallest MSE and MAE and the highest R2 in the quantitative detection of methanol and ethanol content. This study explores the conversion of 1D spectra into 2D images through the mathematical of GAF, which not only opens up a new perspective for more intuitive display of spectral features, but also makes it possible to introduce the powerful advantages of deep learning image processing into the field of NIR analysis. The proposed method provides a new idea for the intelligent determination of alcohols-diesel.