Multi-Dimensional Computational Combustion of Highly Dilute, Premixed Spark-Ignited Opposed-Piston Gasoline Engine Using Direct Chemistry With a New Primary Reference Fuel Mechanism

Abstract

This work presents a modeling approach for multidimensional combustion simulations of a highly dilute opposed-piston spark-ignited gasoline engine. Detailed chemical kinetics is used to model combustion with no sub-grid correction for reaction rates based on the turbulent fluctuations of temperature and species mass fractions. Turbulence is modeled using RNG k-ε model and the RANS-length scales resolution is done efficiently by the use of automatic mesh refinement when and where the flow parameter curvature (2nd derivative) is large. The laminar flame is thickened by the RANS viscosity and a constant turbulent Schmidt (Sc) number and a refined mesh (sufficient to resolve the thickened turbulent flame) is used to get accurate predictions of turbulent flame speeds. An accurate chemical kinetics mechanism is required to model flame kinetics and fuel burn rates under the conditions of interest. For practical computational fluid dynamics applications, use of large detailed chemistry mechanisms with 1000s of species is both costly as well as memory intensive. For this reason, skeletal mechanisms with a lower number of species (typically ∼100) reduced under specific operating conditions are often used. In this work, a new primary reference fuel chemical mechanism is developed to better correlate with the laminar flame speed data, relevant for highly dilute engine conditions. Simulations are carried out in a dilute gasoline engine with opposed piston architecture, and results are presented here across various dilution conditions.

The challenges of using detailed chemistry model for simulating direct injection spark ignition engine combustion during cold-start

“…Such methods of optimizing the rate constants for an improved mechanism behavior have been performed in the past. 11,41…”

“…34 Since the domain is initialized with laminar conditions (represented by the point in Figure 5), the flame propagation speed given by the SAGE model could be compared to the laminar flame speed from Chemkin. It should be noted that the propagating flame speed obtained from equation (11) implicitly accounts for the effects of stretch as explained by Chen, 35 whereas Chemkin gives the unstretched laminar flame speed. Such a comparison was made because of the challenges associated with estimating the stretched laminar flame speeds at engine relevant conditions provided in Table 7.…”

“…In spite of this, the model has been successfully used for capturing premixed gasoline engine combustion as evident from the multiple publications from academia and industry. [7][8][9][10][11][12][13][14] However, most of the validations have been provided for hot operating conditions, where the flame travel is assisted by a significant amount of in-cylinder turbulence. For example, Yang 10 performed extensive validations of the SAGE model in two different spark-ignited engines.…”

Machine Learning Approaches to Learn HyChem Models

“…While the current work focuses on developing HyChem models, such optimization techniques can be readily applied to other computational tasks. For example, optimizing overly reduced chemical models is a technique that has been pursued in practical engine simulations [39,40]. In such tasks, high dimensional kinetic systems with hundreds of model parameters are usually present.…”

Arrhenius.jl: A Differentiable Combustion SimulationPackage

“…There has also been increasing interest in optimizing the kinetic parameters in overly reduced reaction models to compensate for the error introduced by over-reduction [32,33]; this has the potential to produce a smaller model with higher fidelity than the ones obtained via traditional reduction methods.…”