Estimating Fuel Consumption, CO2 Release and Energy Consumption for Scania CBE1 Powertrain

Sammanfattning

Monitoring fuel consumption is essential for modern vehicles due to environmentalregulations, performance demands, and energy efficiency needs. A notable gap existsbetween laboratory-certified fuel consumption and real-world performance,prompting regulations like the EU's On-Board Fuel Consumption Monitoring(OBFCM). Traditional direct measurement methods are often impractical or costlyfor mass production vehicles.This thesis investigates indirect methods to estimate fuel consumption by aggregatingdata from standard engine sensors already present in the Scania CBE1 EURO VIpowertrain. The approach focuses on developing a fuel consumption monitoringsystem using sensor fusion. One fuel consumption estimate is derived from estimatedmass airflow and the λ sensor data. This estimate is then fused with the existing fuelconsumption signal from Scania's Engine Management System (EMS).The thesis evaluates weighted signal fusion using noise covariance and fuzzy logiccontrollers as fusion algorithms. The methodology involves implementation andevaluation in MATLAB and Simulink using raw data from Scania CBE1 EURO VIengine testing. The primary objective is to develop a robust model for heavy vehiclepowertrains using only existing hardware, aiming to meet or exceed OBFCM accuracythresholds. The research computes relative instantaneous and average errors inestimating fuel consumption, energy consumption, and greenhouse gas release,specifically carbon dioxide (CO2) calculated via the carbon balance method,compared to a laboratory reference.Results indicate that sensor fusion, particularly the noise covariance method with a'boxed λ' approach (where the model follows the EMS signal when λ is below athreshold), significantly improved accuracy. Monte Carlo simulations incorporatingsystematic sensor errors confirmed that the boxed fusion approach using noisecovariance met OBFCM accuracy thresholds (-5% to +10% over a test cycle). Thismodel demonstrated robustness in transient cycles (WHTC) where the EMS signalalone exceeded the OBFCM error threshold. The investigation also explored the effectof fusing multiple lambda signals, which influenced the precision and multiple σrange of the fused estimates. CO₂ emission estimation using the carbon balancemethod showed good nominal accuracy, calculated using the fused fuel consumptionsignal. In total, 16 cases (four different cycles, four different fuel consumptionestimates) were analyzed for single, double and triple λ signals, out of which thesensor fusion model based on weighted average using noise covariance was the onlymodel to consistently exceed OBFCM requirements for all the cases, through ananalysis of the multiple σ range of the results from Monte Carlo simulations. Theiiistudy demonstrates that a sensor-fusion approach utilizing existing hardware canprovide a viable solution for meeting OBFCM requirements for heavy-duty vehicles.