During a fuel economy or emissions test vehicles have to follow the drive trace within a certain speed range. The tolerance band of the legal drive cycle is relatively small but even in this small tolerance band individual deviations of the test driver are inevitable. These deviations strongly influence the accuracy and repeatability of the test results. These deviations can vary from test to test and from driver to driver. Therefore it is difficult to evaluate their impact on the results in relation to small design changes. But more and more stringent emission regulations require more accurate measurements of emissions and fuel consumption. Other important factors that influence the variability of these test results are temperatures of engine oil and engine coolant at the start of the test. To estimate and compensate the effect of drive trace deviations an analytical model of the vehicle has been developed empirically by using a dynamic neural network. Advantages are better comparability of fuel economy and emissions test results that can enable to reduced fuel consumption and exhaust emissions. (a)
Abstract