Detecting Fuel Adulteration for Otto-Cycle Vehicle Using Onboard Diagnostics Data and Supervised Machine Learning

Fuel adulteration affects operating costs, vehicle efficiency, and air pollution. Published estimates suggest it accounts for at least 10% of global sales. The Brazilian National Petroleum Agency (ANP) reported noncompliance in about 23% of inspections in 2023, including 4.3% confirmed adulteration. Quality verification requires laboratory equipment, and sensor-based approaches are often inaccessible to end consumers. This article proposes a sensorless (software-only) method that detects water adulteration in hydrated ethanol from standard Onboard Diagnostics (OBD) data using supervised machine learning, enabling on-vehicle fuel quality monitoring without additional hardware. The proposed approach is evaluated on real-world driving data from two production vehicles with three water adulteration levels in hydrated ethanol (0.0%, 2.5%, and 5.0%), achieving 84.85%–95.85% multiclass classification accuracy. These results indicate that software-only, OBD-based monitoring can provide a practical solution for in-use fuel quality control.