Development of a Self-Adjusting Engine Out NOx Prediction Model Using Tailpipe O2 Measurements

On-Board Diagnostic (OBD) strategies utilize a predictive model to estimate engine out NOx levels for a given set of operating conditions to ensure the accuracy of the Nitrogen Oxides (NOx) sensor. Furthermore, this model is also used to determine urea dosing quantities in situations where the NOx sensor is unavailable such as cold starts or as a reaction to a NOx sensor plausibility failure. Physics-based NOx prediction models guarantee high levels of accuracy in real-time but are computationally expensive and require measurements generally not available on commercial powertrains making them difficult to implement on controllers. Consequently, manufacturers tend to adopt a mathematical approach by estimating NOx under standard operating conditions and use a variety of correction factors to account for any changes that can influence NOx production. Such correction factors tend to be outcomes of base engine calibration settings or outputs of models of other related sub systems and may not accurately capture the effect of component level drifts that directly influence NOx production such as mass air flow (MAF) sensor drifts, humidity variations, exhaust gas recirculation (EGR) rate variations, etc. Since mathematical approaches approximate physical phenomena, errors in any of the inputs tend to be compounded, thereby diminishing the accuracy of the final output. The method presented in this material focuses on using lambda values derived from tailpipe O2 measurements as the sole input to adjust the NOx model since it is a direct representation of the quality of combustion and accounts for variations in operation that can influence NOx production. This produces an easy-to-implement “self-adjusting” NOx model strategy that is consistent with the physics of NOx formation without requiring any additional hardware and ensures high levels of accuracy required to guarantee OBD robustness.