Axially Distributed NH3 Storage Estimation for a Control-Oriented SCR Model

Cory  Hendrickson, Ford Motor Company

Impending stringent automotive tailpipe nitrogen oxide (NOx) emissions standards will require increasingly complex control-oriented selective catalytic reduction (SCR) models to improve SCR controller performance. However, ammonia (NH3) storage is not directly measurable and, under typical operating conditions, SCR models are poorly observable. Any SCR control strategy incorporating these models depends on accurate state estimation for robust operation, specifically with respect to NH3 storage. Advances in accuracy and increasing complexity demand a direct understanding of the limits and fundamentals of potential storage estimation approaches.

This study explores NH3 storage estimation for an axially distributed SCR model. Notably, two state estimation approaches, a heuristic but computationally simple observer and an unscented Kalman filter, are evaluated. Employing consistent one-dimensional structures, the control-oriented model’s NH3 storage distribution is compared to a plant model’s distribution. Careful attention is paid to the effect of sensor and actuator biases and initial storage error on the observer performance. The NH3 storage distribution errors, within and across each state estimation approach, are evaluated over repeated FTP75 and NEDC drive cycles. Generally, both state estimation approaches improve NH3 storage distribution errors but lack robustness to feedgas temperature and tailpipe sensor biases. The unscented Kalman filter outperforms the heuristic approach but carries a significantly greater computational load.

Download Presentation: