Oxygen storage modeling of a three-way catalyst based on a NARX network

Abstract

The oxygen storage in a Three-Way Catalyst (TWC) influences the removal efficiency of pollutants when the exhaust gas concentration deviates from the equivalence ratio. To calculate the oxygen storage, a neural network model is established to characterize and simplify the oxygen storage of the TWC. Firstly, the TWC chemical reaction model is established to accurately reflect the downstream excess air coefficient changes and calculate the Relative Oxygen Level (ROL). Secondly, to reduce the complexity of the TWC model and accelerate the calculation procedure, a TWC model based on the Nonlinear Auto-Regression with eXogenous input (NARX) dynamic neural network structure is founded. The NARX-TWC model is trained and verified by the calculation results of the chemical reaction model. The results show that the NARX-TWC model can accurately reflect the change of ROL in the TWC, and the calculation time is greatly shortened, which is 2.5% of the time taken by the chemical reaction computation.