Removal of NOxNOx sensor ammonia cross sensitivity from contaminated measurements in Diesel-engine selective catalytic reduction systems

• A new model for engine NOx sensor cross-sensitivity.
• Linear parameter-varying model for SCR systems.
• Mixed H2/H∞H2/H∞ filtering for linear parameter-varying systems.
• Improved algorithm for the estimation problem.

In this paper, we study the robust observer design problem for removal of NOxNOx sensor ammonia cross-sensitivity from contaminated measurements in Diesel-engine selective catalytic reduction (SCR) systems via the mixed H2/H∞H2/H∞ approach. Due to the fact that the NOxNOx sensor is cross-sensitive to gaseous ammonia, the reading of NOxNOx sensor is contaminated severely by the ammonia concentration. Moreover, the cross-sensitivity factor is not constant but varying with respect to the temperature. As the real NOxNOx concentration in the downstream tailpipe is useful for the closed-loop control or fault diagnosis, it is necessary and meaningful to estimate the actual value of NOxNOx concentration and non-constant cross-sensitivity factor from the application perspective. Therefore, we aim to design an observer to achieve the objectives. To design the observer, it is assumed that the variation of cross-sensitivity factor is bounded and the bounds are known in advanced. This assumption is practical since the factor is non-negative and generally smaller than 2.5. With the assumption, a design approach for a robust gain-scheduling mixed H2/H∞H2/H∞ observer is proposed. The gain-scheduling strategy is adaptive to compensate for the existing nonlinearities and the involved time-varying parameters of the SCR model. At the end, simulations and comparisons via an experimentally-validated full vehicle simulator cX-Emission are carried out to show the efficacy and the advantages of proposed approach over the existing extended Kalman filter (EKF).