Document Type : Original Article

Authors

• AmirHossein BasiratPour ¹
• Seyed Mohamamd Safieddin Ardebili ¹
• Mostafa Mostafaei ²
• hasan Zakidizaji ³

¹ Shahid Chamran University of Ahvaz, Ahvaz, Iran

² Biosystems Engineering Department,Faculty of Agriculture, University of Tabriz, Tabriz, Iran

³ Department of Biosystems Engineering, Shahid Chamran University of Ahvaz, Ahvaz, Iran

Abstract

In this study, diesel, biodiesel, and fusel oil blends were meticulously prepared in different ratios. Subsequently, an electronic nose system equipped with eight metal oxide semiconductor sensors was employed to analyze these fuel mixtures. Various analytical techniques, including linear discriminant analysis (LDA), quadratic discriminant analysis (QDA), and support vector machine (SVM) analysis, were utilized. Notably, sensors such as MQ2, MQ5, MQ138, MQ8, and MQ135 demonstrated significant sensitivity in detecting the fuel compositions. Results indicated a detection accuracy of 66.35% for the LDA method, while the QDA method achieved a remarkable 99.98% classification accuracy on the training dataset. The SVM analysis showed impressive classification and separation accuracies, with SVM1 and SVM2 methods achieving 100% and 98.65%, respectively. Evaluation based on desirability function values revealed that SVM2 outperformed other methods, scoring the highest value of 334.0. Consequently, SVM2 emerged as the optimal model for accurately classifying and separating the 15 distinct fuel combinations with exceptional precision in this study.

Keywords

• Diesel
• Sensors
• Electronic nose
• Fusel oil
• Support Vector Machine