Applications of an electronic nose in the prediction of oxidative stability of stored biodiesel derived from soybean and waste cooking oil

Igor G Vidigal; Adriano F Siqueira; Mariana P Melo; Domingos S Giordani; Maria L C P da Silva; Eduardo H S Cavalcanti; Ana L G Ferreira

doi:10.1016/j.fuel.2020.119024

Applications of an electronic nose in the prediction of oxidative stability of stored biodiesel derived from soybean and waste cooking oil

Fuel (Lond). 2021 Jan 15:284:119024. doi: 10.1016/j.fuel.2020.119024. Epub 2020 Aug 24.

Authors

Igor G Vidigal¹, Adriano F Siqueira², Mariana P Melo², Domingos S Giordani¹, Maria L C P da Silva¹, Eduardo H S Cavalcanti³, Ana L G Ferreira²

Affiliations

¹ Departamento de Engenharia Química, Escola de Engenharia de Lorena, Universidade de São Paulo, 12602-810 Lorena, SP, Brazil.
² Departamento de Ciências Básicas e Ambientais, Escola de Engenharia de Lorena, Universidade de São Paulo, 12602-810 Lorena, SP, Brazil.
³ Laboratório de Corrosão e Proteção, Instituto Nacional de Tecnologia (LACOR-INT), 20081-312 Rio de Janeiro, RJ, Brazil.

Abstract

Waste cooking oil (WCO) is a valuable feedstock for the synthesis of biodiesel but the product exhibits poor oxidative stability. Techniques available for assessing this parameter are generally expensive and time-consuming, hence the purpose of this study was to develop and validate a rapid and reliable predictive system based on signals from the sensors of a commercial hand-held e-nose instrument. Biodiesels were synthesized from soybean oil and six samples of WCO, and their physicochemical characteristics and oxidative stabilities determined before and after storage in different types of containers for 30 or 60 days at room temperature or 43 °C. Linear regression models were constructed based on principal component analysis of the signals generated by all 32 e-nose sensors and stochastic modeling of signal profiles from individual sensors. The regression model with principal components as predictors was unable to explain the oxidative stability of biodiesels, while the regression model with stochastic parameters (combining signals from 11 sensors) as predictors showed an excellent goodness of fit (R² = 0.91) with a 45-sample training set and a good quality of prediction (R² = 0.84) with a 18-sample validation set. The proposed e-nose system was shown to be accurate and efficient and could be used to advantage by producers/distributors of biodiesel in the assessment fuel quality.

Keywords: ANP, Agência Nacional do Petróleo, Gás Natural e Biocombustíveis; Electronic nose; FAEEs, fatty acid ethyl esters; FAME, fatty acid methyl esters; FFAs, free fatty acids; HDPE, high density polyethylene; Induction time; MLR, multiple linear regression; Oxidative stability; PCA, principal component analysis; Predictive model; SBO, soybean oil; Storage conditions; VOCs, volatile organic compounds; WCO, waste cooking oil; Waste cooking oil.