Methodology for Reduced-Order Multichannel Modeling of a Catalytic Converter

Numerical simulations of catalytic exhaust after treatment systems are critical for evaluation and optimization. While transient 3D-computational fluid dynamics (3D-CFD) simulations are computationally expensive, single channel models (SCM) are robust and computationally effective, but flow maldistribution that compromises the accuracy of SCM is unavoidable in realistic geometries. We present a novel multichannel model (MCM) by applying chemometric techniques on transient 3D-CFD solutions. The performance of an oxidation catalyst with flow maldistribution is simulated and variables of interest are collected at the catalyst outlet for calibration and assessment of the MCM. Channel locations and weights are determined using D-optimal design on the scores from a principal component analysis, followed by weighted least-squares regression based on two objective functions (heat content and species conversion). MCM exhibits lower residuals than a SCM and closely approximates the computational fluid dynamics data across various test cases, opening up for rigorous control applications with improved accuracy in real-world systems with intricate geometries.