Sources of CO emissions in an HCCI engine: A numerical analysis

In this paper, we investigate the factors influencinga reliable prediction of CO emissions in a homogeneouscharge compression ignition (HCCI) engineusing an improved probability density function(PDF)-based engine cycle model. The PDF-based stochasticreactor model (SRM) as validated in our previouswork [1] is utilized to identify critical sourcesof CO emissions numerically. The full cycle modelincludes detailed chemical kinetics, accounts for theinhomogeneities in temperature and composition, andhas been demonstrated to provide sufficiently reliablepredictions of the combustion and engine parametersand emissions [1–3].The single-zone, multizone, and CFD-based enginecycle models have been widely used to gaininsight into HCCI combustion; however, in general,the predictions of CO emissions have been poor [4].The underprediction of CO emissions using a singlezone-based model results from its inability to accountfor the inhomogeneities. A closed volume, 10-zonemodel faced intrinsic difficulty in predicting the COemissions, underpredicting the measurements by anorder of magnitude [5]. Studies involving a sequentialmultizone model [6] and the segregated solver [7]approaches also showed an error of around 70% inpredicting the CO emissions. Such underpredictionwas attributed to the lack of mass and energy trans-fer between the zones while emphasizing the need formore detailed description of the in-cylinder temperaturedistribution [8]. A 9-zone-based full cycle modelwith mass exchange between zones required modificationof rate constants for the CO oxidation reactionto obtain a good agreement against measurements [9].A CFD-coupled cycle model located the cylinder linerwall as a major source of CO emissions, but the COemissions were underpredicted by 80% [10]. In anotherstudy, the segregated multizone-based modelwas improved by incorporating heat transfer and mixingduring the post-main heat release part of theengine cycle and the CO emissions were predictedas 50% of the measurements for a specific operatingpoint [11]. It has been pointed out that implementinga certain inhomogeneity in the perfectly stirred individualzones of the multizone model could improvethe prediction capability [9].In our previous works, the integrated SRM-basedengine cycle model was demonstrated to predict COemissions at 70–80% of the measurements. Here, weemploy the model to investigate the factors responsiblefor the calculation of CO emissions in such aPDF-based modeling approach.