Inlet NO_x and NH₃ Concentration Estimation for Diesel-engine SCR Systems by Combining Data-Driven Model and Unbiased FIR Filter

Selective catalytic reduction (SCR) systems have been widely used in diesel engine applications. In an SCR system, input NO_x and NH₃ concentration information are of critical importance for the urea dosage controller design and system fault detection. Generally, the NO_x and NH₃ concentration are obtained by physical sensors. However, the physical sensors do not only increase the cost of overall system, but also induce measurement delays. To deal with this issue, an input observer combining a data-driven model and an unbiased finite impulse response (FIR) filter is proposed. The structure of data-driven model is auto-regressive exogenous (ARX) model and partial least square (PLS) is utilized to identify the parameters in the ARX model. Nevertheless, fuzzy c-means (FCM) is also employed to partition the data and obtain multiple local linear models for describing the nonlinearities of the system. At last, an unbiased FIR filter is adopted to estimate the input NO_x and NH₃ concentration simultaneously due to its strong robustness against the noise. The comparisons between the unbiased FIR filter algorithm and Kalman filter algorithm are carried out in MATLAB/SIMULINK. The simulation results demonstrate that the performance of proposed estimator is outstanding.