Estimating Instantaneous Fuel Consumption of Vehicles By Using Machine Learning And Real-Time On-Board Diagnostics (OBD) Data

• Author(s) / Creator(s)

  • Ansari, Amir
  • Abediasl, Hamidreza
  • Patel, Parth Rakeshkumar
  • Hosseini, Vahid
  • Koch, Charles Robert
  • Shahbakhti, Mahdi

• Estimation of instantaneous fuel consumption of fleet vehicles to identify the causes of high fuel consumption and determine the optimum vehicle type for different applications and driving cycles is essential for the design of an intelligent fleet management system. Developing a practical and reliable method to estimate instantaneous fuel consumption of fleet vehicles is the focus of this study. The proposed method uses real-time on-board diagnostics (OBD) data from a vehicle and applies machine learning models that are trained based on actual fuel consumption measurements. Two machine learning models, including random forest and artificial neural network (ANN), are developed for fuel consumption estimation based on OBD and fuel consumption data. The data are collected during real-world urban and highway driving in a 100-km route for a Ford Escape PHEV and a Ford F-350. The OBD data used for machine learning models include engine load, engine speed, intake manifold absolute pressure, air-fuel equivalence ratio, and throttle position. The validation results show that the random forest method is more accurate than the ANN method, with a estimation accuracy of 99% for the two tested vehicles.

  Part of the Proceedings of the Canadian Society for Mechanical Engineering International Congress 2022.

• Date created

  2022-06-01

• Subjects / Keywords

  • Artificial Neural Networks
  • Random Forest
  • Instantaneous Fuel Consumption
  • Machine Learning
  • Fleet Management

• Type of Item

  Article (Published)

• DOI

  https://doi.org/10.7939/r3-4qnt-2997

• License

  Attribution-NonCommercial 4.0 International