The Influence of Physical Input Parameter Uncertainties on Multidimensional Model Predictions of Diesel Engine Performance and Emissions

Multidimensional models require physical inputs about the engine operating conditions. This paper explores the effects of unavoidable experimental uncertainties in the specification of important parameters such as the start of injection, duration of injection, amount of fuel injected per cycle, gas temperature at IVC, and the spray nozzle hole diameter. The study was conducted for a Caterpillar 3401 heavy-duty diesel engine for which extensive experimental data is available. The engine operating conditions include operation at high and low loads, with single and double injections. The computations were performed using a modified version of the KIVA3V code. Initially the model was calibrated to give very good agreement with experimental data in terms of trends and also to a lesser degree in absolute values, over a range of operating conditions and injection timings. This paper reports on the effect of input parameter uncertainties on predictions of BSFC, and soot and NOx emissions trends for these baseline cases. The results show that accurate input parameter specification is essential if multidimensional modeling is to be used for engine design and trend analysis. Indeed, erroneous conclusions about emissions trends are shown to be possible if physical model input uncertainties are not considered.