IN THE FIELD OF TECHNOLOGY DEGREE PROJECT
ENGINEERING PHYSICS
AND THE MAIN FIELD OF STUDY MECHANICAL ENGINEERING, SECOND CYCLE, 30 CREDITS
,
STOCKHOLM SWEDEN 2018
CFD-simulations of urea-water
spray in an after-treatment system
using Star-CCM+
EMELIE TRIGELL
Acknowledgements
Abstract
The legislation has forced the vehicle industry to reduce tail-end emissions. The air pol-lutant nitrogen oxide (NOX) has been shown to have a negative impact on human health
and the environment. One of the key technologies to reduce the levels of NOX emitted
from a vehicle is by implementing an after-treatment system. The after-treatment system includes catalysts, a particle filter and an evaporation system. In the evaporation system a liquid jet containing a urea-water solution known as AdBlue is injected into the hot exhaust gases to evaporate into gaseous ammonia NH3and water H2O. Then NH3enters
the Selective Catalytic Reduction (SCR) catalyst where it chemically reacts with NOXto
form N2 and H2O. Problems can arise if an excessive amount of AdBlue is injected and a
fluid film is formed on evaporation surfaces. At certain operating conditions the fluid film can crystallise and form solid deposits. The solid deposits can cause high back-pressure, material deterioration and ammonia slip.
This project is done in collaboration with Scania CV AB. Scania is a world-leading manufacturer of heavy-duty vehicles, busses and engines. Scania works continuously to develop new simulation methods to capture the complex phenomena of AdBlue spray, wall film dynamics and risk of solid deposits, to use in the development process of new components. The aim of this project is to implement and evaluate a new method to predict the risk of crystallisation of urea (AdBlue) using the software Star-CCM+. Two di↵erent geometries are studied, a test rig and a Scania silencer. Di↵erent operating con-ditions, parameter settings and a speed-up method are analysed. During the project a base-line model has been created and the results have been compared with measurement results and the software AVL Fire.
The results on the test rig show the e↵ect of altering the mesh and important model parameters. Injected particles are grouped into parcels with the same properties. The number of parcels is a crucial factor for the wall film formation and should be sufficiently high to get a statistical representation of the droplet size distribution. The results from the real silencer show strong evaporation and thin wall film formation with the suggested method.
The method is shown to be stable and the software is user-friendly. A speed-up method was investigated to decrease the computational time. The computational time was re-duced by a factor 20. The outcome of this project is a guide for set-up of AdBlue spray and wall film simulations. Recommendations to future work includes further validation of the settings, investigation of the evaporation rate and droplet size distribution and the application to other cases. The next step is also to tune the critical thresholds for deposit risk assessment.
Sammanfattning
Lagstiftning har tvingat fordonstillverkare att minska avgasutsl¨appen. Luftf¨ororeningen kv¨aveoxid (NOX) har visat sig ha en negativ inverkan p˚a m¨anniskors h¨alsa och p˚a milj¨on.
En viktig teknik f¨or att minska utsl¨appen av NOX ¨ar att implementera ett
efterbehand-lingssystem. Efterbehandlingssystemet tar hand om avgaserna genom substrat, filter och ett f¨or˚angningssystem. I f¨or˚angningssystemet sprutas en urea-vattenl¨osning, som kallas AdBlue, in i de heta avgaserna d¨ar den f¨or˚angas till ammoniak NH3 och vatten H2O.
Ammoniakgasen leds d¨arefter in till SCR katalysatorn d¨ar den kemiskt reagerar med NOX och bildar kv¨avgas N2 och vatten˚anga. Problem kan uppst˚a om fel m¨angd AdBlue
sprutas in, d˚a kan v¨atska byggas upp p˚a f¨or˚angsningsytor, kristallisera och bilda avlag-ringar. Avlagringarna kan bygga upp en solid klump som kan orsaka ett h¨ogt mottryck, nedbrytning av material och ammoniakslip.
Detta arbete ¨ar ett samarbete med Scania CV AB som ¨ar en v¨arldsledande producent av lastbilar, bussar och motorer. Scania arbetar kontinuerligt med att utveckla nya simu-leringsvertyg f¨or att beskriva uppkomsten av Urea avlagringar f¨or att anv¨anda i utveck-lingen av nya komponenter. Syftet med detta arbete ¨ar att implementera och utv¨ardera en ny metod f¨or att prediktera klump mha simuleringsverktyget Star-CCM+. Tv˚a olika geometrier ¨ar anv¨and i arbetet: en testrigg och en av Scanias ljudd¨ampare. Olika drifts-punkter, parametrar och en uppsnabbad metod ¨ar studerade. Under projektets g˚ang har en modell byggts upp och j¨amf¨orts med m¨atningar och simuleringar fr˚an programvaran AVL Fire.
Resultatet fr˚an simuleringarna p˚a testriggen visar e↵ekten av att variera olika parametrar. Partiklarna som sprutas in i systemet ¨ar grupperade i paket med liknande egenskaper. Antalet paket p˚averkar uppbyggnaden av v¨aggfilm och det rekommenderas att denna pa-rameter h˚alls h¨og f¨or att statistiskt beskriva droppf¨ordelningen av partiklar. Resultaten p˚a ljudd¨amparen visar en stark f¨or˚angning och en tunn v¨aggfilm f¨or samtliga driftspunk-ter.
Den implementerade metoden har visat sig vara stabil och anv¨andarv¨anlig. En uppsnab-bad metod har utv¨arderats f¨or att minska ber¨akningstiden. Ber¨akningstiden kunde mins-kas med en faktor 20. Resultatet av arbetet ¨ar en guide f¨or hur metoden implementeras och b¨or anv¨andas. Rekommendationer till framtida arbete ¨ar en fortsatt unders¨okning av parametrar, utv¨ardering av f¨or˚angningsmodellen, validering av droppstorleksf¨ordelningen och till¨ampningen p˚a andra geometrier. N¨asta steg i utvecklingen skulle vara att kalibrera tr¨oskelv¨arden f¨or prediktering av klump.
Contents
1 Introduction 1
1.1 European strategy for low-emission mobility . . . 1
1.2 The diesel engine after-treatment system . . . 1
1.3 Engine after-treatment development at Scania . . . 2
1.4 Computer resources . . . 4
1.5 Thesis aim and delimitation . . . 4
2 Evaporation system and deposit formation metrics 5 2.1 Modelling the evaporation system . . . 5
2.2 Method . . . 8
2.2.1 Standard method suggested by Siemens . . . 8
2.2.2 Speed-up method . . . 8
2.3 Evaluation metrics . . . 10
2.3.1 Deposit risk assessment . . . 11
3 Models and methods 13 3.1 Flow modelling . . . 13
3.2 Numerical method . . . 14
3.2.1 Discretization . . . 15
4 Investigated cases and set-up 15 4.1 Test rig . . . 15
4.1.1 Measurement data . . . 15
4.1.2 Computational set-up . . . 16
4.1.3 Operating conditions . . . 17
4.1.4 Validation and verification tests . . . 17
4.2 Medium silencer . . . 20 4.2.1 Performed tests . . . 21 4.2.2 Measurement data . . . 21 4.2.3 Computational set-up . . . 21 4.2.4 Operating conditions . . . 21 5 Results 22 5.1 E↵ect of parameters on the test rig . . . 22
5.2 E↵ect of operating conditions on the Medium silencer . . . 29
5.3 Evaluation of the speed-up method . . . 30
1
Introduction
Emissions of air pollutants from internal combustion engines have been shown to have nega-tive e↵ects on public health and the environment. Example of air pollutants are hydrocarbons (HC), carbon monoxide (CO), nitrogen oxide (NOX) and particulate matter (PM). The ve-hicle industry aims at developing solutions for improved air quality through reduction of the amount of emitted air pollutants. This is done through after-treatment systems, which reduce the emissions from combustion engines.
The first section of this chapter starts with an overview of the legislation of exhaust emis-sions and how to control the levels of emisemis-sions. It is followed by a brief explanation of the after-treatment system and problems that can arise in the system. Finally, this chapter ends with a description of the purpose and the limitations of the thesis.
1.1 European strategy for low-emission mobility
During the past years, stringent legislation has forced the vehicle industry to reduce tail-pipe emissions. A European strategy for low-emission mobility aims at reducing the emissions pro-duced by combustion engines and speed-up the development towards zero-emission transport solutions. The Euro VI legislation defines the acceptable limits for exhaust emission in the EU and EEA member states. The current legislation, Euro VI, regulates the allowed emission limits for di↵erent vehicle categories such as passenger cars, light commercial vehicles, trucks and busses. According to Euro VI, the critical emission components for heavy duty diesel en-gines are nitrogen oxides (NOX) 0.4 g/kWh and particulate matter (PM) 0.01 g/kWh. This is a large reduction compared to Euro V emission limits of NOX2 g/kWh and PM 0.2 g/kWh [1]. In order to reduce the nitrogen oxides produced by the engine an exhaust gas re-circulation system (EGR) can be implemented. The purpose of the EGR is to recirculate the engine exhaust gases into the intake manifold. The EGR system dilutes the oxygen levels in the incoming flow, resulting in a reduction of the peak in-cylinder temperature. This system reduces the levels of NOX produced by the diesel engine operating at high cylinder temper-atures and pressures [2]. However, to fulfil Euro VI regulations for heavy duty trucks EGR is not sufficient. Thus, an after-treatment system is needed. Consequently, with a good after-treatment, EGR is no longer needed.
1.2 The diesel engine after-treatment system
Figure 1.1: A typical after-treatment system for an internal combustion engine. The focus of the thesis is the evaporation system illustrated by a circle, called AdBlue.
The next subsystem is the evaporation system. In this system a liquid urea-water solution (UWS) is injected into the hot exhaust gases. The most common UWS used is called AdBlue and contains 32.5% urea. The evaporation system aims at introducing the gaseous reductant ammonia (NH3) that can react with the NOX originating from the combustion engine. The injected liquid urea ((NH2)2CO) undergoes a couple of chemical reactions: thermolysis and hydrolysis, which convert urea into ammonia NH3 and isocyanic acid (HNCO) [3]. The reactions are defined in Equation 1.1. Then the conversion of NOX produced by the engine takes place in the Selective Catalytic Reduction (SCR) catalyst.
Thermolysis reaction: (NH2)2CO ! HNCO + NH3 Hydrolysis reactions: HNCO + H2O ! NH3+ CO2
NOXconversion: NOX+ NH3 ! N2+ H2O
(1.1)
The last step is the ammonia slip catalyst (ASC) that provides a gaseous ammonia (NH3) oxidation function. This function is important, because ammonia is a regulated emission and should not be emitted into the environment.
Issues can arise for both an excessive and an insufficient amount of injected urea. A wall film of liquid urea can be formed on evaporation surfaces. If the operation conditions are favourable a crystallisation of components in the liquid film can result in white solid deposits. These deposits can cause high back-pressure, material deterioration and ammonia slip. Fig-ure 1.2 shows the result of an excessive dosing where solid deposits have been formed. An insufficient amount of injected urea means that there is not enough ammonia available for the conversion of NOX needed.
1.3 Engine after-treatment development at Scania
Figure 1.2: Figure of solid deposits formed on a mixer blade and an injector [4].
Figure 1.3: Schematic figure of the after-treatment system used in Scania trucks called the Medium silencer.
1.4 Computer resources
The PDC Center for High Performance Computing at KTH is a leading provider of high per-formance computing (HPC) for academic research in Sweden. Scania has its own cluster part at PDC that enables Scania to access the HPC resources and perform benchmark tests for larger systems than on the in-house computer cluster resources. In this project simulations were performed on resources provided by the Swedish National Infrastructure for Computing (SNIC). The cluster used was the Beskow cluster which is a CRAY XC40 with 2,060 compute nodes. This cluster is currently the fastest academic supercomputing system in Scandinavia. Beskow is designed for running large parallel jobs and is thus suitable for large CFD simula-tions. The in-house clusters at Scania were used in the early state of the thesis and for serial meshing.
1.5 Thesis aim and delimitation
The objective of this Master thesis is to investigate a method to simulate the urea wall film formation in an after-treatment system. Furthermore, the thesis aims to quantify the risk factors for the formation of solid deposits in the system. A new method is tested and evalu-ated in the CFD software Star-CCM+ able to predict the risk of crystallisation of urea. This project was made in collaboration between KTH and Scania CV AB.
The main focus of this study was to create a base-line model for the prediction of deposits. The project focuses on the evaporation system. The substrates and filters will not be studied in detail. The developed method cannot capture the formation of solid deposits. The spec-trum of inter-phase reactions within the fluid film is very complex and there are uncertainties regarding the di↵erent reaction rates. Additionally, the chemical reactions would contribute to longer simulation times. Therefore, this thesis focuses on the formation and dynamics of the fluid film.
The project aims to answer if it is possible to use Star-CCM+ as an e↵ective simulation tool for after-treatment simulations and the prediction of deposit formation, compared to other software. The work flow is to investigate parameters such as flow conditions, meshing strategies etc; perform and compare the results with experimental data and other software; implement a speed-up method; and compare computational time. The outcome of the project is a cleaned simulation file containing the recommended parameter settings and a guide for set-up of simulations (found in Chapter 6).
Goal To create a validated base-line model for the prediction of urea deposits in an after-treatment system, which can be used by engineers.
2
Evaporation system and deposit formation metrics
This chapter reviews the implications of urea sprays in the evaporation system and the current state-of-the-art in after-treatment urea-spray research. Then, the chapter discusses the metrics used to evaluate the mixing of ammonia and the formation of wall film. Finally, a risk assessment of solid deposits based on metrics from literature is presented.
2.1 Modelling the evaporation system
This section describes the main physics inside the urea evaporation system. An illustration can be found in Figure 2.1.
Figure 2.1: Illustration of the complex physics inside the evaporation system [5]. The Lagrangian approach describes the flow field by tracking individual particles in space and time. In comparison, an Eulerian specification of the flow field describes the motion of the fluid at a specific location in space. The spray modelling utilize a Lagrangian-Eulerian framework. The dispersed phase (spray droplets) equations are solved using a Lagrangian approach, and the continuous phase (flow) is solved using the Eulerian approach. A multi-phase flow model describes the interaction between the distinct thermodynamic multi-phases solid, liquid and gas. The model tracks the flow path of the dispersed particles in the continuous phase, including droplet evaporation. A multi-component gas describes the exhaust gases from the engine, and can contain several di↵erent components such as oxygen and NOX. In the simulations all engine out gases are lumped into one gas species (air), to reduce simulation time. A two-way coupling is implemented in order to couple the particle-flow interactions. Momentum that is lost by the particles is gained by the continuous phase.
A Lagrangian injector is used to introduce UWS-particles into the continuous phase. The multi-component droplets consist of more than one material, in this system water and urea. In the simulation particles are grouped into packages of particles with the same physical properties, called parcels. The number of parcels must be high enough for the statistical distribution of droplet diameters to represent the distribution and not only single events. Evaporation characteristics of UWS (AdBlue) droplets show the relation between droplet diameter and initial temperature, see Figure 2.2. In the work by Wang [12] a single droplet is injected through a needle into a heated chamber and the evaporation characteristics is classified using di↵erent ambient temperatures. The process is complex since UWS is a multi-component liquid and the components influence each other. Urea influences the evap-oration rate of water and at the same time urea can decompose and react to form other substances [12]. The evaporation model used in the simulations is calibrated according to these measurements.
Figure 2.2: The droplet evaporation characteristics of AdBlue. Reprinted from T.J Wang [12], Copyright (2018) with permission from John Wiley and Sons.
When urea evaporates it is directly decomposed into ammonia (NH3) and isocyanic acid (HNCO). To reduce the computational cost a lumped evaporation approach is used. The approach is validated with [3], [12] and full scale cases. When a particle impinge on a wall surface it can rebound, spread, break-up, splash or decompose into a liquid film. The Bai-Gosman Multi-Regime impingement model describes how a droplet is resolved based on two parameters; the temperature of the wall TW and the incident droplet Weber number W e.
The Weber number is defined using the density of the droplet ⇢, the diameter of the incident droplet d, the velocity u and the surface tension , according to Equation 2.1. The di↵erent regimes can be found in Figure 2.3.
W e = ⇢u
2d
Spread
Rebound
Break-up and spread
Break-up Break-up and spread Splash, no deposition Deposition (100% liquid film) Splash and deposition (50%, 35%film) We [-] TW [K] TS TL We1 We2 Wea R eb ou n d Wes 5
Figure 2.3: The Bai-Gosman wall impingement regimes [5].
field, the droplets, the fluid film, and the solid wall in order to understand the formation and evaporation of the fluid film. The solid wall components are resolved and the heat conduction is computed to predict the correct metal temperatures. The conjugate heat transfer between the solids, flow and the droplets are computed. If the heat transfer to or through the solid is not captured correctly the fluid film can be over, or under predicted. The heat transfer between the droplets and the solid wall is described by Equation 2.2. This equation captures the cooling of a solid wall due to droplet impingement, using the e↵ective area A the droplet interact with, the contact time for the energy exchange tdc, the thermal conductivity ,
the density ⇢, and the specific heat cp. The indices d and w represent droplet and wall,
respectively. Qw d= A 2ptdc p ⇡ bwbd bw+ bd (Tw Td), A = 4 ⇡D 2 d,ef f, bi= q ( ⇢cp)i, i = d, w (2.2)
The conjugate heat transfer between the solid and fluid domain is described by: d dt ˆ V ⇢cpT dV = ˛ A qqq00· daaa + ˆ V sdV (2.3)
and wall is scaled by setting the heat penetration coefficient (HPC multiplier) equal to the value pSF . However, the build-up of wall film does not follow the same simple pattern. Therefore, the time data can only be scaled when analysing the temperature and not when the wall film is considered. Evaluations of the e↵ect of the scaling factor will be presented in Section 5.1.
2.2 Method
The method used to simulate the after-treatment system was suggested by Siemens, who own and develop Star-CCM+, and implemented by the author. This section briefly discusses the method used to simulate the wall film formation and dynamics. Two di↵erent methods will be presented, the standard method suggested by Siemens and a speed-up method used to get data for longer periods of time. A full set-up guide for the simulations can be found in Chapter 6.
2.2.1 Standard method suggested by Siemens
In the standard method suggested by Siemens the physics set-up includes defining the physics continuum, the fluid film, the Lagrangian multi-phase and the multi-phase interactions.
• Physics Continuum: unsteady, three dimensional, ideal gas, Multi-component gas: Air (NOX is not included), NH3, H2O, non-reacting, gravity, realisable K-Epsilon Two-Layer, Two-layer All y+ Wall Treatment
• Fluid Film: multi-component liquid H2O, N2H4CO, Bai-Gosman wall impingement,
droplet evaporation, impingement heat transfer, turbulent dispersion, two-way coupling • Lagrange multiphase: multi-component liquid N2H4CO and H2O, two-way coupling
• Multi-phase interactions: Impingement, film evaporation/boiling, droplet evaporation The droplet break up can be included in the simulation. This feature is disabled in the stud-ied simulations since no significant break-up is believed to occur when the injected droplet distribution has been formed.
Fluid droplets can be removed from a liquid film due to wave stripping across a surface or edge stripping at sharp corners. The wave stripping and edge stripping models do not compute the size of the stripped droplets that are added to the dispersed phase. This means that the droplets produced by the film phase have a user-defined diameter. These features are disabled in the studied simulations, but need to be activated in some cases.
The method does not include inter-phase reactions within the fluid film. Therefore, a risk assessment is made to predict the formation of solid deposits. The criteria used to evaluate the risk of solid deposits is found in Section 2.3.1.
2.2.2 Speed-up method
di↵er-Steady flow simulation Spray simulation (one injection cycle) Export source terms to Film simulation Run Film Simulation (⇠100 s) Export source terms to Spray simulation
Figure 2.4: A flow-chart over the speed-up method. The method is initialised using a steady state simulation. Then the simulation file is divided into two files. The spray simulation is run for one injection cycle and source terms during this time are exported to the film simulation. The film simulation is run for 100 s or more time. However, during these 100 s the spray simulation might have to be re-run a couple of times with a new wall temperature (indicated by a dashed arrow), exported from the film simulation.
ent problems; the spray simulation is run with a short time-step as required by the physics, while the film simulation can be run with a large time-step enabling results from longer time periods, without compromising the accuracy. The coupling between the two simulations is presently done manually by exchanging source terms, but will be integrated in future Star-CCM+ versions. A schematic flow chart of the method is presented in Figure 2.4.
The coupling between the two simulations is achieved by sampling source terms and data, with field functions in one simulation and then exporting this as tables. The tables are then imported and applied as source terms and boundary conditions in the other simulation. The method starts by setting up the simulations in agreement with the standard method proposed by Siemens, found in Section 4. A steady state simulation is run to find an initial condition. The simulation file is then divided into two files, one for the dispersed phase (spray) and one for the film. The spray simulation is run for one injection cycle with a small time-step, capturing the two-way coupling between the spray and the flow. If the spray is pulsating the time from the start of one injection to the next is called one injection cycle. The source terms of the components, momentum and energy are exported to the film simulation. The film simulation then captures the build-up and evaporation of the film without tracking the small particles. In the film simulation a longer time-step is used which enables the user to run the simulations for a factor of 20 longer than using the standard method.
the film simulation. Therefore, it is normally needed to re-run the spray simulation. How frequently you need to couple the simulations depends on the temperature regime. If the wall temperature is entirely below the lower transition temperature of the Bai-Gosman wall impingement model, found in Section 2.1, then when a droplet impinges a wall it will build a liquid film. In this case the source terms will not change and the spray simulation doesn’t have to be re-run. The same thing occurs if the wall temperature remains above the lower transition temperature, then the droplets will impinge a dry surface and rebound or splash in the same way over time. However, if the wall temperature passes the lower transition temper-ature there is a need to update the spray simulation in order to accurately capture how the droplets interaction with the wall. The more times the spray simulation is re-run, the more the computational time will increase. In general three couplings are needed. The coupling is based on the wall area below the lower transition temperature, and is implemented using field functions with thresholds.
2.3 Evaluation metrics
The metrics used to evaluate the fluid film formation and the potential reduction of NOX are the fluid film thickness, the temperature distribution and the ammonia to NOX ratio (ANR). The build-up of fluid film is of interest to predict the risk of solid deposits. The fluid film thickness is evaluated at the end of the injection cycle, before the next cycle starts. The thickness of the film will vary with time because of evaporation and a periodic injection cycle. The total fluid film mass describes the build-up of film over time and is also used as a tool to predict the deposit risk. Additionally, the footprint of the spray, average fluid film thickness, maximum thickness and temperature are evaluated when predicting the risk of deposits. Area standard deviation An area standard deviation of the di↵erence in wall film thick-ness found with di↵erent settings is used to compare results, see Equation 2.4. This will reflect the rms of the di↵erence between each case and a reference case.
= v u u tPf f 2 Af P fAf (2.4)
f denotes the wall film thickness di↵erence between the reference case ( f,ref) and case i
( f,i) for face f , Af is the area of face f and is the mean wall film thickness di↵erence.
ratio (ANR) is often described using the ANR uniformity at the inlet of the SCR. This value represent how well ammonia has been mixed with the exhaust gases.
ANR = xNH3
xNOX
(2.5) In the simulations the mass fraction of NH3 is used to observe the mixing of ammonia, since it is proportional to the ANR.
2.3.1 Deposit risk assessment
To model the formation of solid deposits the simulation has to include inter-phase reactions. This functionality is available in Star-CCM+ and is usually used for simulating de-iceing. However, to model the inter-phase reactions in the film the reactions must be specified. The spectrum of reactions that form deposits is a very complex process. A flow-chart of the dif-ferent reactions involved in the formation of solid deposits can be found in Figure 2.5. The developed model cannot capture the inter-phase reactions within the fluid film. Instead, to predict the formation of solid deposits a risk assessment can be done.
Figure 2.5: Reaction network for urea decomposition with byproduct formation and decom-position. Reprinted from A. M Bernhard [8], Copyright (2018) with permission from Elsevier.
Wall film temperature [ C] Urea crystallization risk Melting point of Urea No deposit risk Slow Urea decomposition Low deposit risk Intermediate deposit risk
Low temperature deposits: Urea, biuret & cyanuric
acid Increased Urea decomposition High deposit risk High temperature deposits: ammelide & ammeline Low temp. deposits start to decompose High temp. deposits start to decompose 100 133 160 210 250 350 700
Figure 2.6: Temperature regimes for decomposition and crystallisation of urea. Hydrodynamic risk factors consist of the wall film pathway, initial footprint, wall film dynamics, wall film thickness and wall film velocity. These factors become important if the temperature is favourable for the build-up of solid deposits such as biuret, cyanuric acid, ammline or ammelide. The pathways of the fluid film are potential areas where deposits can build-up. The initial footprint is the primary impingement area during the injection. This area is not a risk area because the formation of solid deposits are impeded by the intense dynamics, and continuous dilution and mixing of the fluid film. The wall film dynamics (WFD) describe the mass fluctuation of the fluid film according to Equation 2.6, where m0
is the liquid film mass at the beginning of a studied period. WFD = mmax
m0
= mmax mmin m0
(2.6) The coefficient is based on the maximum and minimum mass on each cell face exposed to wetting during the investigated period. A period can consist of several injection cycles. If W F D W F Dc the area is not considered to be at risk for deposits. The critical values
for wall film dynamics, velocity and thickness are not fixed values. The criterion is only a correlation and measurements are needed to calibrate the critical values for the prediction of solid deposits. In general, a thick film with high velocity is considered to be an area of transport, which is locally not a risk area. In contrast a thick film with low dynamics or low velocity is considered to be a deposit risk area.
temperature is not high enough for biuret to be formed. Thus, this region has no deposit risk. The third temperature regime is where secondary reactions take place. In this regime the formation of biuret (C2H5N3O2) can become sufficient if the concentration of gaseous isocyanic acid HNCO is above a threshold value. The combination of high temperatures and a high concentration of HNCO can result in the formation of ammelide (C3H4N4O2) and ammeline (C3H5N5O). The high temperature deposits decompose at temperatures T > 700 C which is above the operating conditions of Diesel engines.
3
Models and methods
This section presents the flow modelling and numerical methods used in the simulations. The first part describes the turbulence modelling, the boundary layer resolution and the boundary conditions. The second part presents the precision, solver and discretization used in the simulations.
log RANS
Largest scale Inertial scale Dissipative scale
LES DNS
log E()
Figure 3.1: Energy spectrum for a turbulent flow and which scales di↵erent turbulence meth-ods resolve.
3.1 Flow modelling
model is used in this work.
Sources of errors in the flow modelling includes not knowing the important physical mech-anism, complex physics of the flow system and shortcomings of the closure models for the unresolved turbulence. In the case of after-treatment simulations the time and length scales are problematic because the spray includes small particles that need a small time-step and the formation of wall film is a transient phenomena with much longer time scales. The turbulence is also important in order to predict the mixing of the spray. There is a trade-o↵ between accuracy and turn-around time. For the URANS simulations, all turbulent fluctuations are modelled. Thus, reducing the accuracy of the results.
The URANS equations are an approximation of turbulent flows and are derived using en-semble averaging of the Navier Stokes. The flow equations are separated into an enen-semble- ensemble-averaged part 0(t) and a fluctuating part 0(t), which is modelled. This separation is called
Reynolds decomposition, see Equation 3.1 where T represents a time interval for the changes in the flow variable (t).
(t) = 0(t) + 0(t), 0(⌧ ) = 1 N N X i=0 (⌧ + iT ) (3.1)
Turbulent boundary layer Close to a wall there is a boundary layer. The turbulent boundary layer can be described by three sub-layers: the viscous sub-layer, the bu↵er region and the log region. The near wall region called the viscous sub-layer is characterised by wall shear stress, turbulence production and dissipation of the smallest scales and the velocity increases linearly with the distance to the wall. In the log region there is a known logarithmic relation between the velocity and the wall distance for the flow over a flat plate. The bu↵er region is difficult to model and therefore it is desirable to have the first cell either in the viscous sub-layer or in the log region. The y+ of the mesh defines the position of the first
cell layer, and thus defines which treatment that should be used. y+ is defined in Equation 3.2 using the kinematic viscosity ⌫, the friction velocity u⌧ =
p
⌧w/⇢ and the viscous length
scale l⇤= ⌫/u⌧. Derivations can be found in [6].
y+⌘ y l⇤ =
yu⌧
⌫ (3.2)
A low y+ mesh (y+ 5) is recommended to ensure that the heat transfer at the wall is well resolved. However, the Fluid Film model requires that the first cell height is twice the film thickness, and to ensure this a high y+ mesh (y+ 30) is used, putting the first cell in the log-layer.
3.2 Numerical method
3.2.1 Discretization
For the spatial discretization a Hybrid Gauss - LSQ model is used to solve the gradients. The convective terms are discretised with a second-order upwind scheme. The gradient limiter MINMOD is used to prevent spurious oscillations. A first order temporal discretization is used for the simulations.
4
Investigated cases and set-up
This chapter describes the investigated cases and the computational set-up. Two di↵erent geometries were studied: a test rig and a real after-treatment system used at Scania. Further, the chapter presents the studied operating conditions for which some experimental measure-ments exist. A series of verification studies are carried out, including a grid convergence study, investigation of flow conditions, parameter sensitivity and injection strategies.
The simulation procedure is the following, the CAD-model of the geometry is imported to the software ANSA where modifications are made. The model is then imported into Star-CCM+ where meshing, solving and post-processing are performed. For a transient simulation it is beneficial to establish a steady-state flow field before starting to inject Lagrangian particles into the system. A steady-state solution is used as an initial condition for the transient sim-ulations. The initial conditions are mapped using the function Data Mappers.
The boundary conditions at the inlet are specified using the total temperature TT ot and
the total mass flow rate ˙mT ot. The specific mass flow rate at each face of the inlet boundary
is calculated using the face area| ~Af|, the inlet area | ~AT ot| and the total mass flow rate ˙mT ot,
see Equation 4.1. ˙ mf = ˙mT ot· | ~ Af| | ~AT ot| (4.1) The boundary condition at the outlet is specified using a static pressure.
4.1 Test rig
The first geometry used in the simulations is a simple test rig consisting of an inlet, a test section and an outlet. A schematic figure illustrating the test rig is shown in Figure 4.1. Inside the test section there is a 3 mm solid evaporation plate of stainless steel. The urea injection is placed at the roof of the test rig and directed at the evaporation plate. The nozzle is angled 30 degrees to the vertical axis in the flow direction. There are windows at the sides to capture the build-up of fluid film at the evaporation plate.
4.1.1 Measurement data
Figure 4.1: Schematic figure of the test rig. The urea injection is placed at the roof and directed to the evaporation plate.
presented in the figure: no dosing, no wall film, pool, boarder case and deposit. Observe that the operating conditions and dosing strategies are di↵erent for the di↵erent pictures, because the di↵erent behaviours are not present at every case.
4.1.2 Computational set-up
The computational domain replicates the experimental set-up. The standard method pro-posed by Siemens is used for simulations of the test rig. The initial condition used is a steady-state simulation based on the operating conditions. A uniform mass flow and con-stant total temperature is imposed at the inlet of the test rig. A static pressure is used for the outlet boundary. The dosing unit operates at constant pressure and injects 195 g/min during the first part of the 1 Hz injection cycle. The dosing is adjusted by the time the nozzle valve is open. The nozzle is a hollow cone injector. A particle size diameter distribution is implemented as a table over the cumulative density function. The initial particle temperature is 60 C. The species mass fractions of UWS used in the simulations are xH2O = 0.675 and
Figure 4.2: Pictures of the wall film behaviour at di↵erent operating conditions and dosing. The pictures are taken from the side of the test rig along the spray direction [9].
4.1.3 Operating conditions
Two di↵erent operating conditions are studied. A low temperature and low mass flow, and a high temperature and high mass flow, see Table 4.1.
Description Temperature [ C] Mass flow [kg/h] Dosing [g/min] Deposit
Low temperature - low flow 200 400 5 x
High temperature - high flow 300 1000 10
-Table 4.1: Operating conditions investigated in the simulations of the test rig.
4.1.4 Validation and verification tests
Several di↵erent settings were investigated and the e↵ect of the di↵erent setting are presented below.
Grid convergence study The computational domain is discretized with polyhedral cells. Using the polyhedral cells, an automatic mesh generation is available in Star-CCM+. Polyhe-dral cells include more connections to neighbouring cells compared to hexagonal cells. Thus, a better mesh quality can in general be achieved for complex geometries.
Parameter Value
Global cell size 5 mm 4 mm 3 mm Refinement cone 4 mm 2.5 mm 1.5 mm Local cell size on plate 4 mm 2.5 mm 1.5 mm
Table 4.2: Cell sizes used for the grid convergence study. The convergence study includes meshes with combinations of the parameters presented.
Figure 4.3: The computational domain used for the test rig. The mesh depicted uses a base cell size/ refinement cone/ local cell size on plate of 4 mm/1.5 mm/1.5 mm respectively.
There is a trade-o↵ between accurately integrating the PDE across the computational do-main and computational time. Therefore a grid convergence study was performed to find a grid which does not change the results significantly between grid sizes. The fine grid can be observed in Figure 4.3. The grid convergence study was performed by alternating three di↵erent parameters: the global cell size, the local cell size in a refinement cone along the spray direction, and the local cell size on the evaporation plate. The values used can be found in Table 4.2. In the results section the used grid is given as base cell size/ refinement cone cell size/ local cell size on plate in mm.
Smooth transition of the grid size from the near wall region to the flow is obtained by using a stretching factor of 1.2. The mesh settings were chosen to ensure that any fluid film that forms will not be thicker than 12 of the first cell thickness. The solid component, i.e the evaporation plate, was meshed by using the thin mesher with three layers. This was recommended by Siemens in order to resolve the temperature gradients within the solid.
Case Time step spray [ms] Time step flow [ms]
0.1 ms/0.5 ms 0.1 0.5
0.5 ms/1 ms 0.5 1
1 ms/5 ms 1 5
Time-step In order to accurately integrate the partial di↵erential equations a small time step must be used. As a general rule: if a smaller time-step is used, fewer inner iterations are needed because the solution changes less between two time-steps. However, a small time-step increase the computational time, which is not desirable. A periodic time-step is implemented where the time-step is lowered when the spray is active and 0.2 s after the spray is not injecting. When the the Lagrangian particles have left the domain an increased time-step is used to save computational time. The optimal convergence balance between time-step, number of inner iterations and the under relaxation factor must be calibrated for the given problem. The di↵erent time steps tested in the simulations can be found in Table 4.3. The number of inner-iterations was altered between 15 and 25.
Number of parcel streams 50
100 200
Table 4.4: The di↵erent number of parcel streams used in the simulations.
Parcel stream The number of parcel streams injected at each time step is important when analysing the behaviour of the wall film. A statistically independent number of parcels should be introduced to remove isolated events and to have an accurate representation of the droplet size distribution. If the number of parcels increase, the statistical accuracy of the simulation is improved. The parcel streams per time step implies that if the time-step is smaller, the number of parcel streams per time-step can be decrease and vice verse. This was investigated in the simulations using di↵erent time-steps and parcel stream per time-step. The number of parcels tested can be found in Table 4.4.
The Lagrangian source terms for the continuous phase are directly proportional to the par-ticle count per parcel. The numerical stability in terms of parcels can be described as: when the number of parcels is increased the particle count per parcel is decreased, by decreasing the particle count the Lagrangian source term in any volume cell of the continuous phase is decreased, and by decreasing the source terms the interactions between the dispersed phase and the continuous phase is smoothed. Therefore, to improve solver stability three parame-ters can be tuned. First, a small time-step can be used during the injection period and film evaporation. Secondly, a smaller parcel mass will be obtained by increasing the number of parcels. Thirdly, larger cell volumes can increase the stability by avoiding small prism layer cells, but this may cause accuracy issues.
Case Scale factor for the specific heat
No scaling 1
SF = 75 75
SF = 150 150
Figure 4.4: A schematic figure of the Medium silencer.
Scaling the specific heat The conjugate heat transfer between solid and fluid regions can be scaled in order to reach steady state faster. There are two places where the scaling factor (SF) is applied. The specific heat of the solid is scaled by dividing the it with the scaling factor. The impingement heat transfer which accounts for the heat transfer between particle and wall, is scaled by setting the heat penetration coefficient (HPC multiplier) equal to the valuepSF . The analysed scale factors for the conjugate heat transfer can be found in Table 4.5. If nothing else is stated, a scale factor of 75 is used.
Cone and time randomization In the Lagrangian injector the condition Time Ran-domization and Cone RanRan-domization were studied to analyse if these parameters a↵ect the footprint of the wall film. The Time- and Cone randomizations are two numerical models which describe how the particles are released from the nozzle.
4.2 Medium silencer
Figure 4.5: Cut of the evaporation system of the computational domain. The flow direction is illustrated with arrows. The di↵erent colours indicate di↵erent surfaces. The nozzle position is illustrated with a white arrow and a blue box called AdBlue injector.
4.2.1 Performed tests
The focus of these investigations were to apply the method evaluated on the test rig and use it on the Medium silencer to observe if it is possible to capture any di↵erences in the wall film formation between the di↵erent operating conditions.
4.2.2 Measurement data
Experimental testing for solid deposits have been performed on the Medium silencer. The testing was made during 3-12 h of operating time. The build-up of wall film and temperature distribution are not available for this geometry.
4.2.3 Computational set-up
The computational domain replicates the real geometry without DOC at the inlet section. An illustration of the mesh used in the simulations is presented in Figure 4.6. The substrates are removed from this figure. The injector is the same that was used on the test rig, see Section 4.1. The set-up used for this geometry has a base cell size of 4 mm and a local cell size of 1.5 mm. The number of parcel stream used was 200 and the scaling factor of the specific heat was 75.
4.2.4 Operating conditions
Figure 4.6: A cut of the evaporation system of the computational domain. The mesh used has a base size of 4 mm and a local refinement on surfaces with a cell size of 1.5 mm. The substrates are removed from this figure.
Case Temperature [ C] Mass flow [kg/h] Relative dosing to case A
A 350 500 1
B 350 500 0.81
C 260 1200 0.49
D 290 1200 0.81
Table 4.6: Operating conditions investigated in the simulations of the Medium silencer.
5
Results
This section will present the results of the methods described in Section 4. The results obtained from the simulations are analysed in terms of the area standard deviation of the di↵erence in wall film thickness, the footprint of the wall film and the mass fraction of ammonia (NH3). This section is divided into three parts: e↵ect of parameters using the test rig, e↵ect of operating conditions on the Medium silencer and e↵ect of the speed-up method using the Medium silencer.
5.1 E↵ect of parameters on the test rig
To validate the method several parameters were investigated such as mesh independence, time-step, inner-iterations, parcel streams, time/cone randomisation, scaling of the specific heat and di↵erent operating conditions.
mm/1.5 mm, time-step 0.1 ms/0.5 ms, inner-iterations 15, parcel stream 200 and SF = 75. The picture in Figure 5.1 is obtained from an observation point directly above the evapora-tion plate and facing straight down on the plate. The flow is coming from the left to right and only the top of the evaporation plate is shown. The footprint of the spray forms a round shape. In the middle of the footprint there is a region without wall film. In front of this region there is a strong gradient in the wall film thickness. To the sides of the region without wall film there are streaks with thicker wall film. The maximum thickness is 50 µm.
Figure 5.1: The wall film thickness for the reference case used to measure the dependence of parameters.
Mesh independence The resolution of the mesh is of interest when predicting the risk of deposits. Three di↵erent base cell sizes were analysed with values of 3, 4 and 5 mm. The area standard deviation between the reference case and the altered base cell sizes can be found in Table 5.1. The time-step used was 0.5ms/1ms and the number of parcel stream was set to 200.
Case Mesh Area standard deviation Number of cells
1 3 mm/1.5 mm/1.5 mm 1.87 µm 3376791
2 4 mm/1.5 mm/1.5 mm 1.88 µm 2312368
3 5 mm/5 mm/1.5 mm 2.23 µm 1608888
Table 5.1: The mesh independence analysis using three di↵erent mesh set-ups. The area standard deviation is used to find a convergent mesh.
standard deviation is low considering the film thickness and the grainy pattern.
Figure 5.2: The wall film thickness using a base cell size of 3, 4 and 5 mm.
If the local cell size on the evaporation plate is altered from 4 mm to 1.5 mm this a↵ects the resolution of the wall film, see Figure 5.3. When a local cell size of 4 mm is used the footprint is smoother compare to the granular pattern observed using 2.5 mm or 1.5 mm. In case of analysing the risk of deposits all three cases will indicate approximately the same areas. Thus, for the industrial application, a base size of 5 mm could also be used. For the local cell size on the evaporation plate a cell size of 2.5 mm could be used.
Figure 5.3: The wall film thickness using di↵erent local cell sizes at the plate.
The mass fraction of ammonia (NH3) is used to capture the mixing of NH3 in the domain. The results using the di↵erent base cell sizes can be found in Figure 5.4. The figure is taken at the outlet of the test rig. The mixing of NH3 is a↵ected by altering the base cell size. Slightly better mixing is observed for the case using a base cell size of 3 mm compared to 5 mm.
Figure 5.4: The mass fraction of NH3 at the outlet of the test rig using di↵erent base cell sizes.
Figure 5.5: The wall film thickness using three di↵erent time-steps. The time-step is smaller when the spray is active.
the three di↵erent time-steps. Figure 5.5 shows the wall film thickness for these di↵erent periodic time-steps. For all cases in this parameter study the number of parcel stream per step was 200 and the scaling factor of the specific heat was S = 75. The smallest time-step 0.1ms/0.5ms gives a rather smooth footprint. Increasing the time-time-step to 0.5ms/1ms a grainy pattern is observed. Further, increasing the time-step to 1ms/5ms and coarsening the mesh in the spray direction the footprint is even more grainy and the position of the maximum thickness has moved. However, the area standard deviation of the film thickness compared to the smallest time-step is relatively small, see Table 5.2, indicating that there is no large di↵erence in the wall film thickness between the cases.
Time-step Mesh Area standard deviation Computational time
0.1ms/0.5ms 4 mm/1.5 mm/1.5 mm 0 (reference) 110 h
0.5ms/1ms 4 mm/1.5 mm/1.5 mm 1.88 µm 21 h
1ms/5ms 4 mm/4 mm/1.5 mm 2.34 µm 10 h
Table 5.2: The table present the time-step analysis using three di↵erent time-steps. The area standard deviation is used to find a convergent time-step with reasonable computational time.
risk of deposits. Figure 5.6 illustrates the wall film for three di↵erent states in time 3, 12, 20 s. The maximum thickness decreases with time as a result of the extra spray in the first injection cycle. The region in the middle of the footprint is filled with wall film after 12 s. The time development of the film is important to capture, because the risk of solid deposits is greater along the pathway of the film and not at the initial footprint.
Figure 5.6: The wall film thickness at 3, 12 and 20 s. The maximum thickness decreases as a result of the extra spray in the first injection cycle.
Figure 5.7: The figure illustrate the wall film thickness using 50, 100 and 200 number of parcels injected per time-step.
an e↵ect of the injection strategy. A parcel stream of 200 is recommended for this application to avoid the circular pattern.
Figure 5.8: The total number of parcels injected per second is the same in both cases. This is done by altering the time-step and parcel stream. To the left the time-step 1 ms/5 ms using 200 parcel streams and to the right the time-step 0.5 ms/1 ms using 100 parcels streams. Using the same time-step and di↵erent parcel stream the total number of parcels per injection cycle will not be the same for all cases. This may a↵ect the behaviour of the wall film and was analysed using two di↵erent time-steps and parcel streams obtaining the same total number of parcels per second. The results can be found in Figure 5.8. The fan-like footprint is still present for 100 parcels even if the smaller time-step is used. To conclude, in order to get rid of the numerical artefact a higher number of parcels should be used. It does not help reducing the time-step.
less wall film evaporation.
Figure 5.9: The results using no scaling of the specific heat, using a scale factor of 75 and 150. The figure shows 3 s of physical time without scaling the time.
Operating conditions Two di↵erent operating conditions were studied low temperature - low mass flow and high temp - high mass flow, see Section 4.1.3. The results are shown in Figure 5.10. In the low time - low flow case a wall film is forming with a thickness of 30-50 µm. In the high temp - high flow case less wall film is formed and there is a stronger evaporation rate. This is in agreement with measurement where a wall film built-up in the low temperature case and no distinct film was forming in the high temperature case. However, the thickness of the wall film is maximum 50 µm which is a very thin film. In the figure a hole in the middle of the footprint is present. This shape is not seen in measurements, where an evenly distributed film is observed. However, this figure is taken after 3 s of simulation time and the film smears out with time, as seen in Figure 5.6.
Figure 5.10: The figure illustrate the wall film thickness using the operating condition low temp - low flow to the left and high temp - high flow to the right.
Base size 4 mm Surface refinement 1.5 mm
Time-step 0.5 ms/1 ms Parcel stream 200
SF 75
Table 5.3: Recommended parameter settings based on the results using the test rig.
5.2 E↵ect of operating conditions on the Medium silencer
In this section the e↵ect of operating conditions on the wall film in the Medium silencer are presented. Four di↵erent operating conditions were investigated to observe the di↵erences in wall film thickness to predict the risk of deposits, see Table 4.6. The results of the wall film thickness can be found in Figure 5.11. The Cases A and B, have the same temperature and
Figure 5.11: The wall film thickness of four di↵erent operating conditions. The specific details of the mass flow rate and temperatures can be found in Table 4.6.
mass flow but di↵erent dosing rates, and it is seen that they have similar wall film thicknesses. Similar results are obtained for Cases C and D. However, the position of the maximum wall film thickness is altered between the tested Cases A, B and C, D. The wall film thickness is extremely thin⇡ 1 ⇥ 10 10m which means that almost all wall film has evaporated. Based
on this result there would be no risk of deposits.
Comparison with AVL Fire Investigation of the wall film formation on the Medium silencer was performed in the software AVL Fire [16]. The operating condition Case D, see Table 4.6, was used to predict the risk of solid deposits. The wall film thickness can be found in Figure 5.12. The maximum wall film thickness was 2⇥ 10 5m positioned in the bowl.
The minimum value for the wall film thickness was set to 1⇥ 10 6m. Known di↵erences between the Fire and Star-CCM+ methods:
Figure 5.12: The figure illustrate the wall film thickness for Case D using AVL Fire. • The Fire chemistry model includes reactions and formation of solid deposits • The scaling factor of the specific heat
5.3 Evaluation of the speed-up method
In this section the standard method is compared to the speed-up method. A comparison of the wall film thickness using the standard method and the speed-up method on the Medium silencer is shown in Figure 5.13. The figure illustrate the wall film thickness at the inner evaporation pipe and at the bowl after 5 s of physical time using the standard method and after 100 s of physical time using the speed-up method. This would correspond to the end of the injection cycle using the standard method. The wall film footprint is similar for both methods. Using the standard method the maximum wall film thickness can be found in the evaporation pipe and the value is 9.35⇥ 10 10m, compared to the speed-up method where
the maximum value is 4.06⇥ 10 6m, but this is related to the periodic behaviour of the wall
film thicknesses with the standard method.
The periodic behaviour of the maximum wall film thickness using the standard method can be found in Figure 5.14. In this figure the e↵ect of the extra spray in the first injection cycle is seen for the first 3 s. After 3 s of physical time the maximum film thickness oscil-lated between 1⇥ 10 5m and close to zero. In comparison the maximum thickness using the speed-up method was 4⇥ 10 6m, which is the middle of this range.
The scaling factor of the specific heat was analysed using the speed-up method. The simu-lation was run for 100 s of physical time using a scaling of the specific heat of SF = 1 and SF = 75. The scaling of the specific heat was analysed using the fluid film average thickness which can be observed in Figure 5.15. The red curve for the scaling factor of the specific heat SF = 1 increase rapidly to a value of 2.39⇥ 10 7m after which it still increases but with a
lower rate. It reached a maximum value of 2.5⇥ 10 7m after 30 s. The curve then decreases
and converges to a lower value of 2.44⇥ 10 7m. The green curve, corresponding to a scaling factor of the specific heat SF = 75, had a similar behaviour in the first seconds with a rapid increase up to 2.42⇥ 10 7m after 3 s. After this the curve remains stable for the rest of the
simulation in contrast to SF = 1.
Figure 5.14: The maximum fluid film thickness using the standard method.
6
Guide for set-up of simulations
This section will present a guideline for a proper computational set-up based on the outcome of this project. This section is mainly included to be used at Scania and will therefore not be included in the report for KTH.
7
Summary and conclusions
This Master thesis summarizes an e↵ort to improve the simulation method for understanding of urea deposit formation in an after-treatment system. The developed method is one step closer to be used by engineers at Scania. The investigated method has shown promising results in terms of computational time, stability and user-friendly set-up.
A sensitivity analysis of a test rig showed that a convergent solution for the wall film thickness could be found using a base cell size of 4 mm, a cone refinement of 4 mm, a local refinement on a plate of 1.5 mm. A periodic time-step of 0.5 ms when the spray is active and 1 ms else could be used. A convergent number of parcel stream was found to be 200. This number of parcel stream is needed to get a statistical representation of the droplet size distribution and to reduce numerical artefacts. The time development of the wall film thickness is important to capture for the prediction of solid deposits. The scaling factor for the specific heat was altered from 1 to 75 and 150. The scaling factor can be used to speed-up the cooling of the wall and thus the initial build-up of wall film. However, there was no additional cooling when using a scaling factor of 150 compared to using a scaling factor of 75. Thus, a scaling factor up to 75 is sufficient.
The method was then applied to the so called Medium silencer to investigate if the method can capture the deposit risk. The results showed strong evaporation and very thin wall film formation for all four cases studied. There was no distinct di↵erence between the cases which were indicated in previous measurements. The wall film thickness was very thin compared to simulations made in AVL Fire. The reason for this could be that the heat transfer within the solid walls of the geometry are treated di↵erently. The heat convection is solved in Star-CCM+ by meshing the solids. In AVL Fire the heat convection is modelled by coupling the sides of the solid and defining a heat flux. A mesh sensitivity analysis should be performed in the Medium silencer for both software.
the standard method or 2-5 days using the speed-up method. However, the speed-up method is shown to not capture the periodic behaviour of the build-up and evaporation of the wall film. This may a↵ect the mixing of ammonia in the system and thus give inaccurate ANR values. The di↵erence in wall film thickness between the standard method and the speed-up method will depend on when the comparison is made. If the comparison is made at the end of the injection cycle the speed-up method will indicate a thicker film compared to the standard method. The speed-up method is a promising method in terms of computational time, and since the simulation time can be reduced significantly the time development of the film can be better captured.
Overall, Star-CCM+ is a user-friendly software in comparison with AVL Fire. The tree structure used in Star-CCM+ is intuitive to navigate. It has a flexible structure using field functions which enables the user to design and control the monitoring of interesting pa-rameters. A base-line for the field functions and monitors should be established based on evaluation parameters to improve the post-processing.
The following main conclusions can be drawn from this Master thesis:
• A new method using the software Star-CCM+ has been implemented and tested for a number of parameter settings. The method is stable, robust and user-friendly.
• The wall film formation in the after-treatment system has been simulated for a num-ber of operating conditions. Strong evaporation was found for all the studied cases indicating low deposit risk, which needs to be investigated further.
• A speed-up method has been implemented and evaluated. The speed-up method can reduce the computational time by a factor of 20.
• The risk factors for solid deposits have been identified based on literature. However, the critical thresholds for the deposit risk have not been calibrated and this was not applied to the studied cases.
7.1 Recommendations to future work
This project has brought new insight to the simulation methods for deposit predictions in Star-CCM+ at Scania. There are still improvements to be done and it will be interesting to apply the method to other geometries and compare with measurements and other software. The following areas would be interesting to develop further:
• Compare simulation results to other software such as AVL Fire using the same set-up. • Compare wall temperature results to measurement data.
• Study ammonia mixing, compare to measurements and investigate the e↵ect of mesh, time-step and speed-up method.
• Investigate the droplet size distribution. This factor could influence the evaporation rate. If the injected droplets are small then it can result in a faster evaporation or if the droplets are large the droplets will rebound and not adhere.
• A mesh sensitivity analysis should be performed on the Medium silencer. • Turbulence intensity could a↵ect the mixing and should be studied.
• The speed-up method could be applied to other geometries and cases and tested further to calibrate criterias for updating the spray simulation.
• For verification measurements of more parameters connected to spray, film and conver-sion is necessary. An initiative is on-going to further develop a measurement set-up. • Develop post-processing tools for the risk assessment of solid deposits.
• Study how the solid properties and heat transfer modelling a↵ect the evaporation of the fluid film.
References
[1] EC, 2009. Regulation (EC) No 595/2009 of the European Parliament and of The Council of 18 June 2009 on type-approval of motor vehicles and engines with respect to emissions from heavy duty vehicles (Euro VI) and on access to vehicle repair and maintenance in-formation and amending Regulation (EC) No 715/2007 and Directive 2007/46/EC and repealing Directives 80/1269/EEC, 2005/55/EC and 2005/78/EC. O↵. J. Eur. Commu-nities (L 188/1).
[2] L. Guzzella and C. Onder, Introduction to Modeling and Control of Internal Combustion Engine Systems, ISBN: 9783540222743, 2004.
[3] S. Fisher, Simulation of the urea-water-solution preparation and ammonia homogeniza-tion with a validated SCR-systems Fisher, Vienna Technical University, Diss., 2012. [4] H. Smith, T. Lauer, M. Mayer and S. Pierson, Optical and Numerical Investigations on
the Mechanisms of Deposit Formation in SCR Systems, SAE International, vol. 7, no. 2, pp. 525-542, 2014.
[5] Siemens Training, Selective Catalytic Reduction (SCR) Training, S¨odert¨alje, Scania, 2018-01-15 - 2018-01-16.
[6] A.V Johansson and S. Wallin, An introduction to turbulence, Stockholm, KTH, 2015. [7] J Mutyal et al, Development and Validation of a Simulation Model for
Urea-Water-Solution Decomposition for Automotive SCR Systems, SAE Technical Paper 2015-01-2795, 2015, https://doi.org/10.4271/2015-01-2795.
[8] A M. Bernhard, et al, Hydrolysis and thermolysis of urea and its decomposition byprod-ucts biuret, cyanuric acid and melamine over anatase TiO2, Applied Catalysis B: Envi-ronmental. s 115–116. 129–137. 10.1016/j.apcatb.2011.12.013., 2012.
[9] C. Nottbeck, Prediction of Urea Deposit Using CFD-Simulation and Experimental Test-ing, Master Thesis, Stockholms University, 2017.
[10] A. Lundstr¨om, B. Waldheim, H. Str¨om and B. Westerberg, Modelling of urea gas phase Thermolysis and theoretical details on urea evaporation, Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering, 2011.
[11] H. Smith, T. Lauer, V. Schimik and K. Gabel, Evaluation and Prediction of Deposit Severity in SCR Systems, SAE International, vol. 9, no. 3, pp. 1735-1750, 2016.
[12] T.J Wang, S.W Baek, S.Y Lee, et al, Experimental Investigation on Evaporation of Urea-Water-Solution Droplet for SCR Applications Propulsion and Combustion Laboratory, School of Mechanical, Aerospace and Systems Engineering, Korea Advanced Institute of Science and Technology (KAIST), AIChE J., 55: 3267-3276. doi:10.1002/aic.11939, 2009.
[14] C. Brown, J. Bastiaans, C. Antes, V. McDonell and U. Mondragon, Spray Characteriza-tion: Phase Doppler Interferometry and Photos of S3.2, S4.2, S5.2 amd S5.3 Mozzles, 2012.
[15] C.E. Brennen, Fundamentals of Multiphase Flow, Califonia Institute of Technology, Cambridge University press, 2005.
