Parametric Model for Estimating Pollutant Emissions in M1 Otto Cycle Vehicles with OBD-II

• Published in: 2023 IEEE Seventh Ecuador Technical Chapters Meeting (ECTM) pp. 1 - 6
• Main Authors: Rivera, Nestor, Jimenez, Edisson, Cardenas, Joel
• Format: Conference Proceeding
• Language: English
• Published: IEEE 10.10.2023
• Subjects: Data models; Estimation; Machine learning algorithms; Neural Networks; OBD; Parametric Model; Parametric statistics; Temperature; Training; Urban areas; Vehicle Emissions
• DOI: 10.1109/ETCM58927.2023.10308992

This article presents a proposed parametric model for estimating pollutant gas emissions in vehicles with Otto cycle engines. This model is based on the acquisition of on-board diagnostic data and machine learning algorithms. The data are collected through portable devices during Real Driving Emissions (RDE) road tests, and subsequent analysis allows for the training and validation of neural networks to calculate emission factors of various pollutants (CO, CO 2 , THC, NOx). In addition, classification learning is considered to assess the behavior of each pollutant in each gear. The model is trained based on three vehicles that followed three different routes, complying with RDE conditions. The obtained emission factors were compared with the IVE model and values close to the latter were found. This model provides crucial information for creating an emissions inventory that reflects the real conditions of the vehicle fleet in the city of Cuenca.