Hybrid Dynamic Modelling of Engine Emissions on Multi-Physics Simulation Platform

This article introduces a hybrid dynamic modelling approach for the prediction of oxides of nitrogen (NOx) emissions for a Diesel engine, based on a multi-physics simulation platform coupling a one-dimensional (1D) air path model (GT-Suite) with in-cylinder combustion model (Computational Modelling Cambridge Ltd. (CMCL) Stochastic Reactor Model [SRM] Engine Suite). The key motivation for this research was the requirement to establish a real-time stochastic simulation capability for emissions predictions early in engine development, which required the replacement of the slow combustion chemistry solver (SRM) with an appropriate surrogate model. The novelty of the approach in this research is the introduction of a hybrid approach to metamodelling that combines dynamic experiments for the gas path model with a zonal optimal space-filling design of experiments (DoEs) for the combustion model. The dynamic experiments run on the virtual Diesel engine model (GT-Suite) were used to fit a dynamic model for the parameters required as input to the SRM. Optimal Latin Hypercubes (OLH) DoE run on the SRM was used to fit a response surface model for the NOx emissions. This surrogate model was then used to replace the computationally expensive SRM simulation, enabling real-time simulations of transient drive cycles to be executed. The performance of the proposed approach was validated on a simulated New European Drive Cycle (NEDC) against experimental data collected for the engine case study, which proved the capability of the methodology to capture the transient trends for the NOx emissions. The significance of this work is that it provided an efficient approach to the development of a global model with real-time transient modelling capability based on the integration of dynamic and local DoE metamodelling experiments.