A structured approach to uncertainty analysis of predictive models of engine-out NOx emissions

Abstract

Predictive models of engine-out NOx emissions continue to be a topic of active research in the diesel powertrain community. Renewed interest in the use of in-cylinder pressure sensor has presented an opportunity for improving these models through the availability of direct information about the in-cylinder combustion process. Despite several research efforts in this domain seen in open literature, there is a significant paucity of work that reports on the uncertainty analysis of such models. These models are typically designed for real-time implementation, and the model prediction is often used for control and/or diagnostic functions. It is therefore critical to establish the expected error bounds or an error probability distribution and the model prediction for consideration during the design of relevant control and/or diagnostic algorithms.