Robust Separation of Signal Domain From Single Channel Mixed Signal Output of Automotive Urea Based Selective Catalytic Reduction Systems

There is a class of sensor constrained, uncertain, chemical reactor systems that pose unique challenges with regard to the feedback signal. We refer specifically to the urea based selective catalytic reduction (SCR) of nitrogen oxides (NOx) in the engine exhaust of diesel powertrains. These catalysts rely on adsorbed ammonia (NH3), produced from aqueous urea, for the catalytic reduction of NOx to N2. Typically, underinjection of urea will result in the slip of NOx, whereas overinjection will induce NH3 slip. The ideal control objective of such a plant is, therefore, to regulate urea injection such that the net slip over the catalyst is minimized. Meeting these control objectives is made difficult due to the presence of an output sensor that is cross sensitive to both NOx and NH3, thereby producing a mixed feedback signal. This signal confounding poses significant challenges with regard to the stability and robustness of both closed loop control as well as on board diagnostics. In the absence of a robust NH3 sensor, it becomes necessary to create alternate methods of signal disambiguation. However, so far in open literature, there has not been a detailed discussion of this problem nor has a concrete solution been proposed to robustly and continuously identify the nature of slip as NOx or NH3. In this paper, we discuss the systematic development of a new method that allows a robust and continuous determination of the slip regime from the mixed signal output of a standard NOx sensor. The full scope of the practical problem is discussed and the performance of the proposed method is shown via experimental data.