Transient NOx emission prediction of gasoline engine based on steady-state correction model

To accurately predict the transient NOx concentration of gasoline engines, a new transient emission prediction model is established by combing a series of dual implicit layer BP neural networks and the transient engine operation parameters. In the new model, steady-state engine data are adopted first for the unsteady neural network training and prediction. Then the output emissions are combined with vehicle control parameters as input parameters for transient emissions correction. All the steady and unsteady data are obtained from the engine bench and whole vehicle hub tests. The emission concentrations under various transient conditions (WLTC and RDE) were explored based on the new method. The results show that the improved transient model can well predict the transient NOx concentration of gasoline engines based on steady-state data with a correlation coefficient of 0.97 and a cumulative emission of error less than 3.5%. In addition, the new transient correction model has high prediction accuracy in extrapolation in terms of NOx, CO, and THC emissions under WLTC and RDE conditions.