Diesel emissions prediction from dissimilar cycle scaling

Abstract

A method for prediction of heavy heavy-duty diesel vehicle emissions on a test cycle using data from dissimilar test cycles is examined and presented. Four dissimilar modes of a new California test schedule were used to generate emission predictions for the US heavy-duty urban dynamometer driving schedule (UDDS). Intensive properties of each mode and of the UDDS, including average speed, stops/mile, percentage idle and average kinetic energy, were chosen for further study. The four dissimilar modes were the idle, creep, transient and cruise modes, created by the California Air Resources Board (CARB) in a prior effort. The predictive weightings were applied to emissions from 11 heavy-duty vehicles measured in units of grams/second (g/s), grams litre of fuel consumed (g/l) and brake specific emissions (g/kW h). Predictions of emissions for the UDDS in units of grams second of NOx and CO2were acceptable, but particulate matter (PM) deviations were substantial. Errors for prediction of NOx did not exceed 7 per cent for any case, and errors for CO2 prediction did not exceed 15 per cent. PM errors in the g/s case varied substantially, depending on the weighting case, indicating instabilities in the predictions. Errors for g/l and g/kW h predictions were higher. A series of corrections was applied to each of the predictive cases for each emissions unit studied. These corrections are shown to improve the predictive ability of the weightings, but the fundamental nature of the prediction was eroded. A weighting using the intensive properties of average speed, percentage idle and average kinetic energy was found to yield the best uncorrected prediction for every case, regardless of the emissions unit considered. The best predictive method was also shown to work acceptably when applied to unseen data from another 12 heavy-duty vehicles.