Numerical and experimental analysis of diesel spray dynamics including the effects of fuel viscosity
Bong, CH (2010) Numerical and experimental analysis of diesel spray dynamics including the effects of fuel viscosity. PhD thesis, University of Tasmania.
The maritime transport industry carries the majority of global trade. Large ships use diesel
engines that are significantly larger than automotive diesel engines and consume low quality
heavy fuel oil (HFO). This project aimed to better understand the dynamics of long duration,
high fuel viscosity diesel sprays that are typical of a marine engine and to improve the accuracy
of CFO models used for optimisation of engine design. The project utilised both numerical
methods and experimental methods. The numerical softw<U'e package called "Star-CO v3.26"
was used for the numerical simulation part of the project. For the gas phase, the Large Eddy
Simulation turbulence model was employed throughout. New sub-models were added to the
Star-CO software to improve the simulation accuracy. Experiments were conducted using a
custom-built High Pressure Spray Chamber (HPSC). The experimental results provided
verification of the simulation results.
Literature reviews showed that limited research has been done on long duration l-IFO sprays.
Most literature is focused on high speed diesel sprays with shott injection durations using low
viscosity fuel with light components. The low fuel viscosity has negligible effect on the droplet
breakup process and numerical modelling requires only a single component fuel. Such studies
can not be applied in HFO due to the high viscosity, complex molecular structure, presence of
liquid phase soot, and variable fuel density. As a result, this project aimed to perform a thorough
study of the fundamental dynamics of such sprays.
Throughout this project, the spray was studied at the gas density equivalent to combustion
chamber density of approximately 35 kg/m3 and at room temperature. This allowed the
dynamics of the spray to be studied in the absence of combustion and evaporation. In the case of
the HFO combustion spray, the presence of high molecular weight components and the
formation of carbonaceous residue means that the spray remains as a multiphase flow
thJoughout the combustion period.
The first phase of the project involved evaluating Lagrangian-Eulerian multiphase numerical
models. In each evaluation, the models were isolated so that only the effects from the model of
interest were shown in the results without unwanted influence of other models. The evaluation
found that the inter-droplet collision model required corrections which resulted in the
development of the mesh independent O'Rourke collision model (MIOC). The collision model
studies showed that the nozzle exit region and the initial droplet cluster region contained the
highest collision rate. This was because the droplet population was dense and the droplet velocities were high in these regions. The Kelvin-Helmholtz/Rayleigh-Taylor (KH-RT) hybrid
breakup model and Tnylor Analogy Breakup (TAB) based dynamic droplet drag model were
programmed into Star-CD v3.26.
The second phase of the project involved conducting experimental analysis of the spray. The
experiment was broken into three parts, namely: spray penetration and cone angle, light sheet
Particle Image Velocimetry (PIV) and dropsize shadowgraphy using a long-distance
microscope. The macro spray structure (penetration experiments) and PIV experiments showed
evidence that surface instabilities from the shearing of the jet against the surrounding gas, and
the in-flow of air into the low pressure region of the spray jet, forms the overall !'pray structure.
The experiment confirmed that increasing gas density results in a lower penetration rate and
larger cone angle. The increase of fuel viscosity had a negligible effect on the spray penetration
but increased the cone angle. PIV measurements were performed on the spray droplets because
it was difficult to externally introduce seeding particles. As a result, it was not possible to adjust
the seeding particle population density directly. The measured droplet velociti es can be
considered equivalent to the gas velocities in the sparse region of the spray because the
numerical results suggested that there is minimal difference between the gas and droplet
velocity. The spray images and the macro-PJV analysis showed the presence of high velocity
regions and the presence of droplet clusters. The dropsize shadowgraphy experiments provided
point velocity measurements that could be analysed using Particle Tracking Velocimetry (PTV)
and f..I.-PJV methods. The results were compared with the macro-PIV resu lts. The dropsize
shadowgraphy re~u lts provided valuable dropsize data and confi1med that high fuel viscosity
had a significant effect on the dropsize of the spray.
The third phase of the project involved the full numerical simulation of the diesel spray and
validation with the HPSC experimental results. The validation confirmed that the KH-RT
breakup model was able to reasonably predict the effects of fuel viscosity on dropsize, which
was not the case with the Reitz-Diwakar breakup model. The penetration and cone angle
validation showed very similar penetration rates in both numerical and experimental results, but
the cone angle of the numerical simulation was narrower compared to the experimental results.
The PIV measurements and simulation velocity proriles showed similar velocity patterns and
velocity magnitudes at the sparse region of the spray. In the dense region, the velocity patterns
remained si milar but the magnitude of the predicted spray velocity was higher. This was
because the experimental results were erroneous in the dense region, due to higher particle
density leading to multiple scattering. The drop10ize validation showed that the numerically
predicted dropsize was slightly larger than the experimental results but consistent under all conditions. It was concluded that the set-up using Large Eddy Simulation (LES) with Blob
atomisation, Kl-1-RT breakup model, MIOC model. TAB based droplet drag model and vertex
based interpolation, produced the most accurate simulation when validated against the
The full spray simulation suggested that the spray structure could be divided into two regions.
The disintegration region showed that most of the breakup process and momentum transfer
(from droplets to gas) occurred here. The stable region showed the formation of droplet clusters
and volumetric expansion of the spray. The numerical simulation result~ showed that a high
viscosity fuel spray contained significantly different internal structures compared to a low
viscosity fuel spray. This was also partly supported by the experimental results. The high
viscosity fuel spray droplet dispersion rate was significantly lower and the formation of droplet
clusters occurred much further away from the nozzle, when compared to low viscosity fuel
spray. The results also showed that an increase in gas density shortened the length between
The outcome of this project is improved understanding of long duration, high fuel viscosity
diesel sprays. It is concluded that the use of LES as the turbulence model produces good
qualitative internal spray structures that predict the instantaneous turbulent jet instability and the
formation of droplet clustering. The project highlights the limitations in the current state of
numerical prediction methods and recommendations are made for future work to improve
|Item Type:||Thesis (PhD)|
|Additional Information:||Copyright 2010 the Author - Embargoed until November 2012|
|Deposited By:||ePrints Officer|
|Deposited On:||11 May 2011 16:45|
|Last Modified:||25 Jul 2012 11:56|
|ePrint Statistics:||View statistics for this ePrint|
Repository Staff Only: item control page