A pressure coupled Representative Interactive Linear Eddy Model (RILEM) for heavy-duty truck engine combustion simulations

Licentiate thesis, 2022

Internal combustion engines (ICE) are frequently debated due to their environmental
consequences. Although the switch to electromobility is currently happening for light
and medium-duty vessels, the transition for ships and heavy-duty trucks is more complex.
The primary problem in the shift is the incapability of matching the range of ICE.
Maintaining the ICE for large vessels is inevitable, requiring intensive research to improve
the combustion engine technology to reduce their impact on the environment.
Operating engines in nonstandard conditions, such as low-temperature ranges or utilizing
partially premixed mixtures, reduces emissions such as NOx or soot. Simulating these
scenarios requires special modeling techniques that can advance finite rate chemistry
and simulate mixed combustion modes. The Linear Eddy Model (LEM) is a potential
candidate to simulate these methods. LEM is a regime and mode-independent mixing
model initially developed for nonreactive flows [1] and later extended to include reactive
flows [2]. LEM advances finite rate chemistry, solves turbulence on a one-dimensional line,
and advances molecular diffusion and heat convection. It offers the possibility of capturing
the turbulence-chemistry interaction directly, which is essential for predicting pollutants
formation. In this thesis, the coupling of LEM to a CFD code that simulates the spray
chamber geometry is referred to as the Representative Interactive LEM (RILEM). RILEM
describes the combustion chamber key parameters and interacts with the CFD in real-time
as a combustion model to update its chemical state. RILEM shares similarities with the
Representative Interactive Flamelet model (RIF) [3]. RIF is essentially the coupling of a
CFD domain to laminar flamelets embedded in a turbulent flow; the coupling is based on
the scalar dissipation rate χ for non-premixed [4], and on combustion progress variable c
or level set approach for premixed cases [5]. RILEM features distinct differences compared
to RIF, mainly the possibility of simulating partially premixed combustion modes and the
ability to represent combustion on a physical one-dimensional line. A previous version
of RILEM was implemented where the CFD and LEM were coupled based on a volume
constraint. The volume-coupled RILEM was successfully validated against experiments
[6]. A recent version of RILEM using pressure coupling is the objective of the presented
thesis. It has been tested using a stand-alone code operating with driving parameters
extracted from a reactive case [7]. The fundamental advantage of a pressure-coupled
RILEM is the intrinsic inclusion of latent heat of evaporation effects and wall heat losses
in the pressure trace communicated from the CFD. These effects need separate modeling
on the LEM side in the case of volume coupling.
The recently developed RILEM utilizes a spherical formulation of the LEM that allows
the line to maintain the characteristic length and the volume effects constant between
the CFD and the LEM. An additional model was developed in this project’s scope titled
multiple Representative Interactive Linear Eddy Model (mRILEMs). mRILEMs advance
multiple LEM lines in parallel, where different turbulent statistics are enforced in each
line to illustrate the stochastic behavior of the eddies. Both RILEM and mRILEMs utilize
scalars conditioned on mixture fraction Z and combustion progress variable space c to
update the CFD chemical state. The local temperatures on the CFD are iterated byutilizing the species mass fractions and the solution of the energy equation in its total
enthalpy form.
RILEM and mRILEMs were tested on a single cylinder case of a heavy-duty truck
engine. The pressure trace was compared against the experimental data and reached a
reasonable agreement. The two models allowed the quantification of intermediate species
such as CO and OH. RILEM provided accurate results by manipulating the probability
density functions of the turbulent scalars combined with different initialization techniques,
where utilizing mRILEMs ensured that the LEM side exclusively controls the combustion
process.