cfd simulation of 2010:144 civ hydrochloric acid regeneration …1024682/fulltext01.pdf ·...

MASTER’S THESIS

2010:144 CIV

MASTER’S THESIS

Universitetstryckeriet, Luleå

2010:144 CIV

Simon Johansson

CFD Simulation of Hydrochloric Acid Regeneration

with a Ruthner Process

MASTER OF SCIENCE PROGRAMME Engineering Physics

Luleå University of TechnologyDepartment of Applied Physics and Mechanical Engineering

Division of Fluide Mechanics

2010:144 CIV • ISSN: 1402 - 1617 • ISRN: LTU - EX - - 10/144 - - SE

MASTER’S THESIS


2010:144 CIV

Simon Johansson







MASTER’S THESIS


2010:144 CIV

Simon Johansson







Preface

This thesis work has been done at the Division of Fluid Mechanics atLulea University of Technology for the company SSAB Tunnplat AB. Ithas been a challenging task to characterize the process of hydrochlorideacid regeneration with the commercial CFD program CFX. The projecthas been a collaboration with Vadims Geza at University of Latvia.

First of all I would like to thank my supervisor Lars-Goran West-erberg for always believing in my work and for the help when I gotstuck. I also want to thank Gunnar Hellstrom, Per Burstrom and ReineGranstrom at Division of Fluid Mechanics for valuable support and ad-vices about the software and how to get forward with this project. Mysupervisor at SSAB Tunnplat AB Bengt Johansson has helped me alot with drawings, reports about the process and has always respondedquickly to my questions.

During the year I’ve been part of the Research Trainee group to-gether with ten colleagues. I would like to thank them all for sharingall amusing, and sometimes frustrating, days at the office. I will foreverhave the interesting and providing study tour to Tokyo, Japan, in mind.

I do also want to thank my housekeeper colleges and friends who hasbeen spending time at Vassijaure youth hostel for a pleasant spring atRiksgransen Ski Resort. Those wonderful blue bird and useless white-out days has been the incitement for me tough days at the office.

Thank you all!

Simon JohanssonRiksgransen, May 2010

i

Abstract

A process at SSAB regenerates waste hydrochloric acid (HCl), from thepickling process, with the spray roasting technique. Waste HCl acid issprayed into a hot chamber where the water in the droplets is evaporatedthen a chemical process when iron chlorides is oxidized to form hematiteand HCl gas. The regenerated HCl is sent to the process for steel picklingand the hematite powder is sold. With a good quality of the powderthe process is ran with profit. But today they have problem to controlquality.

Because of the aggressive environment inside this process the furnaceis closet which makes it hard to do measurements inside the process. Toget an impression of the velocity profile and temperature distribution itis therefor a good idea to simulate the process.

This project aims to, raise the knowledge about the process. ByComputational Fluid Dynamic (CFD) simulations in the commercialsoftware Ansys CFX the present study will explain temperature distri-bution, velocity profiles, particle distribution and how the evaporationprocess changes when nozzle position and angle is changed.

Results shows that the temperature is highest close to the wall anddecreases in the region 40-100cm from the wall and the temperature isalmost constant close to center. These results has also been comparedwith measurements.

Results further shows that the evaporation process is dependent ofposition and angle of the nozzle. Where position has a larger impact onthe rate of evaporation.

iii

Contents

I Summary 1

1 Introduction 1

1.1 The company and their products . . . . . . . . . . . . . . 11.1.1 History of SSAB . . . . . . . . . . . . . . . . . . . 1

1.2 Hot Rolling . . . . . . . . . . . . . . . . . . . . . . . . . . 21.3 Steel Pickling Process . . . . . . . . . . . . . . . . . . . . 31.4 Recycling of acid . . . . . . . . . . . . . . . . . . . . . . . 41.5 Fluid Dynamics . . . . . . . . . . . . . . . . . . . . . . . . 5

1.5.1 Computational Fluid Dynamics . . . . . . . . . . . 51.6 Aim of this project . . . . . . . . . . . . . . . . . . . . . . 6

2 Theory 7

2.1 Governing Equation . . . . . . . . . . . . . . . . . . . . . 72.1.1 Mass Conservation . . . . . . . . . . . . . . . . . . 72.1.2 Change Rate of Fluid Particle . . . . . . . . . . . . 82.1.3 Newtons second law . . . . . . . . . . . . . . . . . 82.1.4 The Navier-Stokes Equation . . . . . . . . . . . . . 10

2.2 Turbulence . . . . . . . . . . . . . . . . . . . . . . . . . . 112.2.1 The time averaged Navier-Stokes equation . . . . . 12

2.3 Finite Volume Method . . . . . . . . . . . . . . . . . . . . 132.4 Droplets . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152.5 Buoyancy . . . . . . . . . . . . . . . . . . . . . . . . . . . 152.6 Evaporation . . . . . . . . . . . . . . . . . . . . . . . . . . 162.7 Regeneration of waste HCl in a Spray Roaster . . . . . . . 16

3 Method 19

3.1 Properties of Reggen . . . . . . . . . . . . . . . . . . . . . 193.1.1 Geometry . . . . . . . . . . . . . . . . . . . . . . . 193.1.2 Burners . . . . . . . . . . . . . . . . . . . . . . . . 20

v

3.1.3 Spray Nozzles . . . . . . . . . . . . . . . . . . . . . 213.2 Mesh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213.3 Model setup . . . . . . . . . . . . . . . . . . . . . . . . . . 24

3.3.1 Boundary Conditions . . . . . . . . . . . . . . . . 243.3.2 Turbelence and numerics . . . . . . . . . . . . . . 263.3.3 Simulation method . . . . . . . . . . . . . . . . . . 27

II Papers 31

A Characteristics of Flow and Temperature Distribu-

tion in a Ruthner Process 33

vi

Part I

Summary

Chapter 1

Introduction

1.1 The company and their products

SSAB is developing high strength steel products and is one of the world’sleading companies in this area. By aking steel harder the amount of steelin vehicles, buildings and other products can be reduced, which reducesthe affect on the environment since less iron has to be used. Lighterproducts require less energy to transport and less steel is used from theearth (SSAB, 2009).

1.1.1 History of SSAB

When the new railway between Falun and Gothenburg was built in 1872a decision was made to build a new iron mill at the riverbed of Dom-narvsforsen in Borlange and Domnarvets Jernverk was built. A largepopulation moved to Borlange to work at the iron mill and the papermills. The owners of those mills built houses in the nearby region forthe people working at the mills. In the 1950s Domnarvets Jernverk wasmodernized and could produce 400,000 tonnes per year (SSAB, 2010).

The ore was exported from the port in Oxelosund and in 1914 a newiron mill was built. This mill was the first in Sweden to use coke fromstone coal as fuel in the iron production.

In 1940 Norbbottens Jarnverk AB, NJA, was established in Lulea.The connection with the railway to the mills in Kiruna and Gallivarewas a big benefit for NJA and in the 1960’s they were the second largeststeel producer in Sweden.

The energy crisis in 1970’s raised the production cost for those threecompanies and the competitor from new modern low production cost in

1

other countries made the loss even bigger.The Swedish parliament merged those three companies together in

1978 and SSAB was established with the Swedish state as the biggestowner. At this time SSAB had eight blast furnaces today they havethree. But the steel production is bigger. In 1989 SSAB was introducedat the Swedish stock market.

In 2007 SSAB streghtend it’s position as one of the world’s leadingsteel production companies by buying the North American companyIPSCO.

Today SSAB has 8700 employees in 45 different countries and haveproduction in Sweden and in the USA.

1.2 Hot Rolling

Hot rolling is a technique to make steel plates thinner. The thicker sheetis heated in an oven to around 900 ◦C. When temperature is homogenoustwo rolls press the sheet together and sheets are moved back and forthuntil the desired thickness is received.

Figure 1.1: Hot rolling of steel slabs

In addition to make steel soft and workable heat works as a catalyzerin the oxidization of iron (Fe) to different iron oxides. Kladnig (2008)describes the complex zone of those Iron Oxide. Wuesuite (FeO) andmagnetite (Fe3O4) is closest to the steel while hematite (Fe2O3) is onthe surface.

2

1.3 Steel Pickling Process

Before sheets are adapted to high strength steels, iron oxide complexeshas to be removed. The principle of steel pickling is pulling the sheetsthrough a container of acid the oxidized iron reacts with the acid andsolves into it.

Historically there are two main acids that has been used in steelpickling plants, hydrochloride acid (HCl) and sulfuric acid (H2SO4).Until the 1960’s H2SO4 was the acid used in most pickling plants butnowadays HCl is used in almost all pickling plants (Kladnig, 2008). Thebenefits of using HCl compared to H2SO4 are stated by Kladnig (2008);Beck et al. (2007)

• Surfaces get smoother.

• The dissolve of iron oxide is higher.

• Velocity for the reactions is 10 times higher.

• HCl can be recycled to 99%.

• It is less expensive than H2SO4.

• Lower pickling temperatures.

The container of acid is normally 100 m3 and has a temperatureof 30-40◦C. The concentration is about 180-200 g HCl/l. To keep theconcentration at a constant level fresh acid is continuously added to thecontainer and the used acid is removed.(Kladnig, 2008)

In the last years new techniques of steel pickling has been developedsuch as push-pull pickling and Venturi tank pickling (Kladnig, 2008).These techniques enforce the turbulence in the acid so that saturatedacid close to the sheet mixes with fresh acid and the reaction goes fasterand the chemical losses are reduced. The leading company in developingthe steel pickling process is the Andritz Group in Austria.

When the sheets are drawn through the container of acid the follow-

3

ing reactions dissolves the iron oxides:

FeO + 2HCl → FeCl2 +H2O (1.1a)

Fe3O4 + 8HCl → FeCl2 + 2FeCl3 + 4H2O (1.1b)

Fe2O3 + 6HCl → 2FeCl3 + 3H2O (1.1c)

Except from these three wanted reaction a side reaction attacks the steel

Fe + 2HCl → FeCl2 +H2. (1.1d)

Reaction (1.1d) is slower and then reaction (1.1a), (1.1b) and (1.1c) andis only to be considered in holdups of the process. The reaction can bereduced by adding inhibitors to the acid but since the H2 acceleratesthe pickling process some amount of H2 is wanted in the pickling acid.(Kladnig, 2008)

1.4 Recycling of acid

In this section a short introduction to different acid recycle processesare presented. The technique used at SSAB, the so called spray roaster,is explained in more details later in the report.

There are two mainly used techniques to regenerate the waste acidfrom the pickling process.

• Pyrohydrolysis

– Spray Roaster Pyrohydrolysis

– Fluidized Bed Pyrohydrolysis

• Hydrothermal Regeneration(Hydrolytic Distillation)

The two different Pyrohydrolysis techniques use the same chemicalreactions with water evaporating in the first step. Then oxidization ofthe FeCl2 and FeCl3 to Fe2O3. One can say that this process is aninversion of the chemical reactions taking place in the pickling process.The setup between the two different Pyrohydrolysis processes differs andso the product, Fe2O3, in a spray roaster a fine powder of hematite isproduced while pellets of hematite is the product of the fluidized bed.

Hydrolytic distillation, described by Demopoulos et al. (2008), ismore energy effective then the two Pyrohydrolysis techniques. First

4

reaction 1.2a oxidize the FeCl2, then FeCl3 reacts with water as in 1.2b

12FeCl2 + 3O2 → 8FeCl3 + 2Fe2O3 (1.2a)

2FeCl3 + 3H2O → HCl + Fe2O3 (1.2b)

1.5 Fluid Dynamics

Fluid dynamics is the science about gases and liquids, describing e.g.how water flow in a turbine, aerodynamics of a car, or hydrodynamicsin a dam. Fluid dynamics is also of importance in other areas e.g. inchemistry were reaction rates are dependent on the mixing of two ormore substances or in heat transfer problems.

Analytical modeling is only possible for simple cases such as e.g. flowthrough a channels and pipes.

1.5.1 Computational Fluid Dynamics

When looking at a real physical problem one realizes that they oftenbecome to complicated to solve analytically. Therefore numerical meth-ods has been developed in order to solve complicated problems that caninvolve heat transfer, chemical reactions and electro magnetics.

Nowadays CFD (Computational Fluid Dynamics) is used as a Com-puter Aided Engineering program (CAE). Instead of doing measure-ments of a product CFD softwares can predict how changes of a processaffect the drag force, heat transfer, reaction rate, or other variables ofinterest. By using the power of CFD companies can reduce the opti-mization cost of their products.

During the last 10 years the cost of doing CFD simulations has beenreduced and it is no longer only available for big industries like automo-bile, biochemistry, aerospace, power, chemical and mineral processing.Areas that do not generate such amounts of money can also afford CFDsimulations e.g. the drag force dependency of a swimmer’s position ordesigning an aerodynamic helmet for bikers.

It is hard to say who did the first CFD calculation but in 1910 LewisFry Richardson (1881-1953) developed a numerical weather forecast sys-tem. He divided the map into a grid and then used the finite differencemethod of Bjerknes’s equation. He tried to do an eight hour forecast ofthe pressure. It took him six weeks to get the results and the resultswere wrong. (Fluent, 2010).

5

In the 1960s the computers become powerful enough to calculatesimple cases. At this time most of the development was done at NASA.In the 1980s the first commercial CFD program was released.

1.6 Aim of this project

The previously described Ruthner process at SSAB generates (8m3)waste HCl acid every hour. If the process is run correct it does notonly regenerates acid, it also produces hematite powder of good qualitywhich can be sold to the electronic industry and SSAB can run this pro-cess and make a profit. The parameters that are critical for the qualityof the hematite is the Chloride content and the surface to volume ratioof the particles.

Since the gas inside Reggen -the combustion furnce- is acetic and hasa high temperature it is a closed system with only a few points whereit is possible to measure variables such as temperature, humidity andvelocities. By only some spot measurements it is hard to get the fullpicture of how the process works.

The aim of this project has been to simulate the process with thecommercial CFD software Ansys CFX. The project is divided in twoparts:

• Simulations that describes the flow and temperature distributioninside Reggen.

• Simulations that investigates how position and angle of spray noz-zles affect the distribution and evaporation of waste acid droplets.

Results from these simulations will hopefully increase the knowledgeof this process at SSAB.

Chapter 2

Theory

2.1 Governing Equation

CFD is based on the governing equations of fluid mechanics. This sectionintroduces the equations solved in CFD and the numerical method usedby CFX, the Finite Volume Method. Notice that all equations are onlyderived for the the x-component. With the same procedure equation forthe y and z-equations can be derived. Fore a more detailed descriptionthe reader is referred to the work by Versteeg & Malalasekera (1995).

2.1.1 Mass Conservation

If mass conservation is not fulfilled for the system it means that matteris created or destroyed inside the system. Which is not physical undernormal conditions. This gives the equation (Versteeg & Malalasekera,1995)

dm

dt=

�

in

m−�

out

m (2.1a)

this formula can also be written as:

∂ρ

∂t+

∂ρu

∂x+

∂ρv

∂y+

∂ρw

∂z= 0 (2.1b)

which can be written in vector form as

∂ρ

∂t+∇·(ρu) = 0 (2.1c)

where the first term is density change over time and the second termthe flow through the element boundaries. If the density is constant the

7

formula becomes∇·u = 0 (2.2)

2.1.2 Change Rate of Fluid Particle

Momentum and energy conservation laws are dependent of propertychanges of a fluid particle. There properties, denoted with the variableφ, are a function of position and time (x,y,z,t). The material derivativewith respect to time can be written as:

Dφ

Dt=

∂φ

∂t+

∂φ

∂x

dx

dt+

∂φ

∂y

dy

dt+

∂φ

∂z

dz

dt(2.3)

Fluid particles follows the flow in the domain, so that dx/dt=u,dy/dt=vand dz/dt=w (2.3) then reads as

Dφ

Dt=

∂φ

∂t+

∂φ

∂xu+

∂φ

∂yv +

∂φ

∂zw =

∂φ

∂t+ u·∇φ (2.4)

This expression defines the change of φ per unit mass. To get the changerate regarding to unit volume (2.4) has to be multiplied with ρ such that

ρDφ

Dt= ρ

�∂φ

∂t+ u·∇φ

�(2.5)

2.1.3 Newtons second law

Newton’s second law of motion states that rate of change of momentumis equal to the forces acting on the fluid particle. (2.5) defines the rateof change for a variable φ by substituting this variable with u, such thatthe change rate in x -momentum becomes

ρDu

Dt. (2.6)

In fluid dynamics forces can be divided into two types; surface forcesand body forces. Surface forces are pressure forces and viscous forces.These are common for all types of problems. Body forces such as grav-ity force, centrifugal force, electromagnetic force and Coriolis force areusually added to the equation as a source term SMx. Mx in this casestands for momentum in x-direction.

Consider figure 2.1 the parallel faces are to be considered simultane-ously since their forces are acting on the same lever arm. Forces having

8

Figure 2.1: Stresses in x-direction. Sides are denoted such that, East(E), West (W), North (N), South (S), Top (T) and Bottom (B).

the same direction as the coordinate axis are considered positive, andnegative if they are anti parallel to the coordinate axis.

Considering at the sides E and W the following expression yields

��p− ∂p

∂x

1

2δx

�−�τxx −

∂τxx∂x

1

2δx

��δyδz

+

�−�p+

∂p

∂x

1

2δx

�+

�τxx +

∂τxx∂x

1

2δx

��δyδz

=

�−∂p

∂x+

∂τxx∂x

�δxδyδz, (2.7)

and for N and S respectively

−�τyx −

∂τyx∂y

1

2δy

�δxδz +

�τyx +

∂τyx∂y

1

2δy

�δxδz =

∂τzx∂y

δxδyδz.

(2.8)For T and B the following yeilds

−�τzx −

∂τzx∂z

1

2δz

�δxδy +

�τzx +

∂τzx∂z

1

2δz

�δxδy =

∂τzx∂z

δxδyδz.

(2.9)

9

Summarizing (2.7), (2.8) and (2.9) and dividing with the volume,δxδyδz, the rate of change in x-momentum becomes

∂(−p+ τxx)

∂x+

∂τyx∂y

+∂τzx∂z

(2.10)

As described above the change rate of momentum is described ac-cording to (2.5). An expression for x-component of the momentum canthen be described from (2.10) and adding the body forces, SMx

ρDu

Dt=

∂(−p+ τxx)

∂x+

∂τyx∂y

+∂τzx∂z

+ SMx. (2.11)

2.1.4 The Navier-Stokes Equation

In (2.11) there still are some unknown variables. The viscous stresscomponent τij has to be further described. One can see the stresses asa function of deformation rate. Most fluids can be considered isotropic,meaning they have the same properties in all directions. Before τij isexplained further the ten different deformation rates have to be intro-duced. Three of them are linear elongating deformation componentssuch that

exx =∂u

∂xeyy =

∂v

∂yezz =

∂w

∂z, (2.12a)

and six of them are shearing linear deformation components accordingto

exy = eyx =1

2

�∂u

∂y+

∂v

∂x

�exz = ezx =

1

2

�∂u

∂z+

∂w

∂x

�

eyz = ezy =1

2

�∂v

∂z+

∂w

∂y

�.

(2.12b)

The volumetric deformation is written as

∂u

∂x+

∂v

∂y+

∂w

∂z= ∇·u (2.12c)

These deformations have to be coupled to the material properties inorder to obtain the stresses. When looking at Newtonian fluids theviscous stresses are proportional to the deformation. Two different typesof viscosity are to be considered: dynamic viscosity µ which is related to

10

the linear deformations (eij) and volumetric viscosity λ, which is relatedto the volumetric deformation. The viscous stresses can be written as

τxx = 2µ∂u

∂x+ λ∇·u τyy = 2µ

∂v

∂y+ λ∇·u τzz = 2µ

∂w

∂z+ λ∇·u

τxy=τyx=µ

�∂u

∂y+

∂v

∂x

�τxz=τzx=µ

�∂u

∂z+

∂w

∂x

�τyz=τzy=µ

�∂v

∂z+

∂w

∂y

�.

(2.13)

By substituting τij in (2.11) with the expressions in (2.13) we get theNavier-Stoke equation

ρDu

Dt= −∂p

∂x+

∂

∂x

�2µ

∂u

∂x+ λ∇·u

�

+∂

∂y

�µ

�∂u

∂y+

∂v

∂x

��+

∂

∂z

�µ

�∂u

∂z+

∂w

∂x

��+ SMx (2.14)

where the stresses can be written as

∂

∂x

�2µ

∂u

∂x+ λ∇·u

�

+∂

∂y

�µ

�∂u

∂y+

∂v

∂x

��+

∂

∂z

�µ

�∂u

∂z+

∂w

∂x

��+ SMx

= ∇·(µ∇u ) + SMx (2.15)

The Navier-Stokes equation can be simplified when the stresses in (2.14)are written as in (2.15)

ρDu

Dt= −∂p

∂x+∇·(µ∇u) + SMx (2.16)

2.2 Turbulence

The Navier-Stokes equation can be solved analytical for simple laminarflows e.g. pipes and chanels. For more complicated cases the turbulencehas to be calculated. The amount of turbulence is described by theReynolds number (Re)

Re =UL

ν(2.17)

11

where U and L are the characteristic velocity and length scale and ν theviscosity of the fluid. The higher Re-number the more turbulence in theflow. For pipes L is the diameter and U the mean velocity.

In order to solve a real world problem those swirls in the flow hasto be modeled. There are a few different types of numerical methods tomodel the turbulence such as Reynolds Average Navier-Stokes (RANS),Large Eddy Simulations (LES) and Direct Numerical Solutions (DNS)to mention a few. RANS simulates the turbulence as a kinetic energy,LES resolves the large swirls while small swirls are simulated as kineticenergy. DNS captures the small swirls. RANS takes less computationalpower and is the standard in industry for most applications. LES is usedmore and more especially for application with very high Re-Numberlike Jet engines and in areas where turbulence need to be predictedmore efficiently then what is possible with RANS. DNS is mostly usedin academic work due to the computational cost and is only practicalwhen looking at very small geometries.

2.2.1 The time averaged Navier-Stokes equation

As in section 2.1 the equation is only derived for the equation in x-direction.

First the mean of a variable ϕ is describes as

Φ = ϕ =1

∆t

� ∆T

0ϕ(t)dt (2.18)

where ∆t has to be larger then the timescale coupled to the property ϕin order to get a good mean value of Φ.

The property ϕ can be describes as the sum of mean and the fluctu-ations in time; hence ϕ(t) = Φ+ ϕ�(t). Where the time average of ϕ�(t)is zero

ϕ� =1

∆t

� ∆T

0ϕ�(t)dt = 0 (2.19)

Fluctuations of the variable Φ is given by the root mean square as

ϕrms =�(ϕ�)2 =

�1

∆t

� ∆T

0(ϕ�(t))2dt

�1/2

(2.20)

12

The intensity of turbulence is defined as

Ti =

�2312

�u�2 + v�2 + w�2

��1/2

Uref=

�23k

�1/2

Uref, (2.21)

where k is the kinetic energy connected to the turbulence and Uref areference mean velocity

To introduce the turbulence into Navier-Stokes equation (2.16) andthe continuity equation (2.1a), vector u = U+ u’ with velocity compo-nents in the x-, y- and z-direction as u, v and w is added as

∇·U = 0 (2.22)

∂u

∂t+∇·(uu) = −1

ρ

∂p

∂x+ ν∇·(∇u). (2.23)

Taking the time average of every component in (2.23) yields

∂u

∂t+∇·(uu) = −1

ρ

∂p

∂x+ ν∇·(∇u), (2.24)

which by some simplifications can be re-written as

∂U

∂t+∇·(UU) +∇·(u�u�) = −1

ρ

∂P

∂x+ ν∇·(∇U). (2.25)

This equation is very close to the instantaneous equation, (2.23), butthe time averaging has introduced the term ∇ ·

�u�u�

�which is the the

fluctuating velocities. The Reynolds equation then becomes

∂U

∂t+∇·(UU) = −1

ρ

∂P

∂x+ν∇·(∇U)+

−∂(u�2)

∂x− ∂(u�v�)

∂y− ∂(v�w�)

∂z� �� stresses

+SMx

(2.26)where SMx is the body forces and the stresses are expanded from the∇·

�u�u�

�term in (2.25).

2.3 Finite Volume Method

The Finite Volume Method (FVM) is the discretization technique usedby most commercial CFD softwares. In this section the method will be

13

explained briefly. Fore a more detailed description the reader is referredto the work by Versteeg & Malalasekera (1995).

The geometry of interest has to be divided into control volumes (CV).This is done by the meshing procedure described in section 3.2, whereeach element or cell is a CV. The variables of interest such as mass,momentum and energy are placed in the middle of the CV.

To focus on the FVM method the discretization of the diffusion equa-tion (2.27) is shown in the one dimensional case. Which makes theformulas shorter and the diffusion equation is also simpler than the mo-mentum equations. In (2.27) Γ is the diffusion coefficient and S thesource term.

d

dx

�Γdφ

dx

�+ S = 0. (2.27)

First step is to integrate the diffusion equation over the CV whichgives the following expression

�

CV

∇·(Γ∇φ)dV +

�

CV

SφdV =

�

A

n·(Γ∇φ)dA+

�

CV

SφdV = 0 (2.28)

Before any further work on the mathematics, the CV volume and itsnotations has to be explained. Fig. 2.2 shows five elements. Every dotis the centroid of each CV. The dotted line shows the element P whichsides, e and w, are shared with element W and E. The distances from Pare denoted δxWP and so on. The length of element P, distance betweenw and e, is denoted ∆x. For the one dimensional diffusion case (2.28)

Figure 2.2: The control volume (CV) and its notations.

can be written as (2.29) where e and w are the sides of the CV shownin Fig. 2.2, A is the area of the side, S is the mean value of the source

14

and ∆V the volume of CV�ΓA

dφ

dx

�

e

−�ΓA

dφ

dx

�

w

+ S∆V = 0. (2.29)

Since there are no nodes at position e and w these values are interpolatedby central difference

�ΓA

dφ

dx

�

e

= ΓeAe

�φE − φP

δxEP

�(2.30a)

�ΓA

dφ

dx

�

w

= ΓwAw

�φP − φW

δxWP

�. (2.30b)

In real world problem the source term S is often a function of φ. Theexpression for the source term can be written as

S∆V = Su + SPφP . (2.31)

By substituting (2.30a), (2.30b) and (2.31) into (2.28) and re-arrangingthe equation so that all φP is on the left side and the rest of the termson the right side, the equation becomes

�Γe

δxPE+

Γw

δxWP− SP

�φP =

�Γw

δxWPAw

�φW +

�Γe

δxPEAe

�φE .

(2.32)

2.4 Droplets

Two way coupled Lagrangian approach is used for the particle track-ing. The two way coupling means that forces and mass transfers arecalculated both ways between continuous phase and particles. If oneway coupling is used the gas phase acts on particles but not the otherway around which would result in that evaporated water would not betransferred into the continuous phase (ANSYS Europe, 2010a).

2.5 Buoyancy

Buoyancy is introduced to the momentum equation as a source term asmentioned in section 2.1.3

15

SM,buoy = (ρ− ρref ) g (2.33)

where g is the gravity and the ρref the reference density. There aretwo ways of calculate the density difference, the full model, or Boussi-nesq model. During this project full model is used where difference iscalculated as stated in (2.33) where ρ is the density of the CV.

When buoyancy is active the pressure term in the momentum equa-tion is corrected so that the pressure is not calculated twice (ANSYS Eu-rope, 2010a).

2.6 Evaporation

In CFX it is important to specify the continuous phase as a mixture ofαg and βg. The mass fraction of βg in the mixture can be set to 0 atthe inlet. When droplets evaporates, βl → βg, the mass fraction of βgin the continuous phase will rise. (ANSYS Europe, 2010b)

During this project the liquid evaporation model has been used tocalculate the evaporation. The liquid evaporation model uses Antoine’sequation to determine the boiling temperature of the droplets which isdependent of the pressure (ANSYS Europe, 2010a).

Pvap = Pref × eA− BT+C , (2.34)

where A, B and C are user defined constants. If Pvap is higher then Pref

the mass transfer is calculated by

dm

dt=

−Qc

V, (2.35)

where Qc is the convective heat transfer and V the latent heat of thedroplet.

2.7 Regeneration of waste HCl in a Spray Roaster

Droplets from the nozzles are heated to 100 ◦Cand water starts to evap-orate while the droplet shrinks. When the mass fraction of H2O is 0.36at the surface H2O and FeCl2 forms a crystalline structure of Thetrahy-drate (FeCl2 • 4H2O). At this point the droplet stops to shrink and allwater is evaporated until there is only a hollow sphere with a shell of

16

FeCl2•4H2O. At this point the temperature rises until all 4H2O in theTetrahydrate is evaporated at 538 K (Beck et al. , 2007).

From now the chemical reactions start. What the reaction will beis dependent on the composition of the surrounding gas. For an idealreaction a molar fraction of H20 to O2 has to be greater than 4 and thetemperature should not exceed 900 K. Under these circumstances thereaction becomes

4FeCl2 + 4H2O → 4FeO + 8HCl− 508kJ (2.36a)

4FeO + O2 → 2Fe2O3 + 563kJ, (2.36b)

which can be summarized as

4FeCl2 + 4H2O+O2 → 2Fe2O3 + 8HCl + 55kJ. (2.37)

17

Chapter 3

Method

3.1 Properties of Reggen

The combustion furnace where the regeneration of HCl is performedis called Reggen. It can be described as a hot chamber were used acidfrom the steel pickling process is introduced at the top of the furnace as aspray. The water in the drops evaporates and then the chemical reactionswhen chloride gas (Cl2) or HCl gas is formed. The iron chlorides reactwith O2 and formes Fe2O3. This type of acid regeneration is called Sprayroaster and is one of the two pyrohydrolysis processes. In this sectiondetails of the geometry and the mesh are described and how the physicaland chemical processes are implemented into the simulation model.

3.1.1 Geometry

A sketch of the side view of Reggen can be seen in Fig. 3.1(a). Theheight of the conical part is 6.28 m, the cylindrical middle part is 8.8 mhigh, the transitional part, where radius changes from 4.3 m to 1.21 m,is 1.05 m high. Four burners are placed symmetrically in the peripheryas in Fig. 3.1(b) of the cylindrical wall 1.06m above the conical part.Diameter of the outlet is 1.05 m and the burner radius is 0.174 m.

In a previous work by Lars-Goran Westerberg, Lulea University ofTechnology, and Vadims Geza, University of Riga, during spring 2009Westerberg and Geza was involved in a project with SSAB Tunnplat ABwith an aim to reduce the amount of sintered iron in the burning cham-bers. A kick-out that reduces the back flow into the burning chamberswas developed during this project and is now applied in Reggen and itsshape and position can be seen in Figs. 3.3(a) and 3.3(b).

19

(a) Reggen from side (b) Reggen from top

Figure 3.1: Drawings of Reggen from side view and top view

3.1.2 Burners

The burners are low-NOx natural-gas burners with an air-staging tech-nique which means that the air is added to the combustion at two differ-ent places. According to Richard (1997) 50−70 % of needed air shouldbe added as primal so that the combustion is complete when secondaryair is added. They are designed in this way to reduce the NOx gases bykeeping the flame temperature low. The extra O2 added in the secondaryair is necessary for the chemical reactions in

The burners are driven by two different types of gases, LPG andnatural gas. Each of them are used for half a year then changed to theother.

In this project only natural-gas is used. The composition of the twodifferent fuels is shown in Tab. 3.1.

20

Table 3.1: Volume fraction of LPGFuels used in Reggen

FuelMethane Ethane Propane Butane iso-Butane Pentane

CH4 C2H6 C3H8 C4H10 C4H10 C5H12LPG - 0.01 0.98 - 0.01 -

Naturagas 0.002 0.044 0.24 0.704 - 0.01

The chemical reactions during combustion of Natural Gas is:

CH4 +O2 → CO2 + 2 H2O (3.1a)

2 C2H6 + 7 O2 → 4 CO2 + 6 H2O (3.1b)

C3H8 + 5 O2 → 3CO2 + 4 H2O (3.1c)

2 C4H10 + 9 O2 → 8 CO2 + 10 H2O (3.1d)

C5H12 + 8 O2 → 5 CO2 + 6 H2O (3.1e)

The burners are feed with 71 m3/s fuel and 2239 m3/h air. Theratio O2 to fuel is 1.4, which means that more O2 than needed for acomplete combustion is added. The extra O2 is used for oxidization ofiron chlorides to iron oxides.

3.1.3 Spray Nozzles

SSAB have tried a few different spray nozzles over the years. The onesused now are manufactured by Lechler. The four nozzles are placedsymmetrically at radius 1.5 m from center as in Fig. 3.2(b) and 13.82m above bottom of Reggen and are pointed 5◦ towards the center as inFig. 3.2(a).

3.2 Mesh

During the project three different mesh techniques have been used: un-structured, structured and a hybrid of these two. At the first stage ofthe project unstructured meshes were used since these are fast and sim-ple to generate. Unstructured meshes do also work well within complexgeometries. Such as the region around the kick out.

In order to get a better quality of the mesh structured mesh wastried but since the geometry is complex around the kick out this attemptfailed.

21

(a) Spray angle is pointing 5◦ to-wards center as the thin line

(b) Nozzles are placed symmetrically

Figure 3.2: Nozzle position and spray angle

The final simulations have been done with a hybrid mesh with 780,000nodes created with a script. The volumes around the kick out is meshedwith a tetrahedral mesh and the rest of the domain with a hexahedralmesh. Pyramids are created at the interface between those two mesheswhere they are merged. An important property of the mesh is that thenodes on the perpendicular sides maps 1:1, has the nodes on the samedistance from center. With the 1:1 property the mesh can be rotatedand the nodes of the rotated part can be merged with the nodes of theoriginal mesh. In Fig. 3.3(a) the cross section in burner level is shownand in Fig. 3.3(b) the region is shown from the side.

The hybrid meshing can be done with Ansys ICEM with the followingprocedure:

1. Import the geometry and name all the surfaces.

2. Create interface surfaces that will be used to merge the unstruc-

22

tured and structured mesh.

3. Create blocking for the structured mesh and convert the pre meshto a unstructured mesh and save it.

4. Create the unstructured tetra mesh on the other part of the ge-ometry. The spacing on the interface surface should be almost thesame as for the structured mesh. Do not create any inflations theyare created at the last step. Save the mesh with a new name.

5. Import the hexahedral mesh saved in point 3, choose merge in thedialog box.

6. Go to the edit mesh tab and chose merge nodes and merge nodes atthe Interface. This step creates pyramid elements at the interface.

7. Delete the surface mesh at the interface.

8. Create inflations in the create mesh feature. It is enough to createone or two inflation layers. These can later be split into morelayers by the feature split mesh in the edit mesh tab.

9. Export the mesh.

(a) Mesh at cross section at burnerlevel

(b) Mesh at the wall around theburner

Figure 3.3: Mesh from two different views

23

3.3 Model setup

In this section the boundary conditions in the model is described andsimplifications are motivated. It also contents a section of how simula-tion procedure is done.

3.3.1 Boundary Conditions

Burners

The four burners are modeled as inlet with normal air with a 10 %mass fraction of steam and a temperature of 1040 ◦C. In reality thetemperature and mass flow vary a little between burners but in orderto make the flow more stable and ease the convergence all burners aremodeled as the mean of all burner.

Wall

At zero distance from the wall the velocity of the gas is zero. Close tothe wall a boundary layer where the velocity changes rapidly from zeroto the bulk velocity. This boundary layer is dependent of the roughnessof the wall. Since the inner walls in Reggen is made of bricks the wallis not smooth. The walls in the model are however approximated assmooth walls due to the complexity of doing a mesh that captures theboundary layer in a correct way.

Droplets that hits the wall are simulated to bounce of the wall whilein reality some of them might get stuck.

A heat transfer occurs through the wall. SSAB approximates theheat transfer to 250 W/m2 which corresponds to a heat transfer coeffi-cient h = 0.6 W/m2K.

Buoyancy

An important effect to model in the process is the buoyancy effect inthe flow. According to Archimedes principle the lift force of an object inwater is the same as the volume of displaced water times gravity, workson gases as well. According to the ideal gas law

pV = nRT (3.2)

hot gas requires more space than the cooler one. The hot gas, which haslower density than the surrounding gas, will rise compared to the cooler

24

one.This effect is introduced to the model since air is imported from the

material library in CFX as ideal gas and gravity (g) is set to -9.81 m/s2.

Outlet

There are two ways to model an outlet in CFX. Either with an openingor with an outlet. The difference between the models is that openingallows air to both enter and exit the domain through this surface andoutlets only allows fluid to exit through the surface. In most cases outletseems to be more physical at the first glance but it can cause errors whensolving the problem. If a fluid wants to enter the domain through theoutlet a wall is placed on the surface. In this case it might be better touse an opening which tolerates both in- and outflow.

The present simulations have been done with an opening. To mini-mize the amount of gas entering the domain through the outlet has beenextruded perpendicular to the gravity as in the real case. Fig. 3.4 showthe importance of correct geometry in this case. In the right picture flowis entering the domain through the opening while in the left picture allvelocity vectors are pointing out through the opening.

Nozzles and spray

SSAB Tunnplat AB has tried many different spray nozzles over the years.The last years nozzles from Lechler, which provide a good spray forma-tion and have a high durability, have been used. The four nozzles aremodeled as conical point sources with a spray angle of 60 ◦. The dropletsare normal distributed with a mean diameter of 370 µm correlating withthe Sauter diameter of the droplets the standard deviation is 100 µm.Initial velocity is 13 m/s. Every hour around 8 m3 waste acid is regen-erated and the temperature of the acid is set to 60 ◦C. Three differentpositions 1 m, 1.5 m and 2.5 m and three angles, straight own and30 ◦ towards center/wall has been simulated. Every position has beensimulated with the three different angles of the nozzles resulting in 9simulations.

Evaporation The waste acid has a mass fraction of 0.49 H20 and 0.51Iron Chlorides. H2O has a density of 1000 kg/m3 and FeCl2 1600 kg/m3

which corresponds to a density of 1300 kg/m3 of the waste acid. Themass flow of the waste acid through each nozzle is 0.5048 kg/s.

25

Figure 3.4: Comparing two different outlets

Particle study In simulations 1200 droplets are introduced every sec-ond, 300 through each nozzle, to represent the total amount of droplets.Fig. 3.5 shows that this amount of droplets gives a good approximationof distribution of droplets. The droplets are not modelled to brake-upor stick to each other.

3.3.2 Turbelence and numerics

SST turbulence model is used which uses k−ε for the free stream flowand k−ω for the region close to wall. These are standard models inindustry.

26

Figure 3.5: Particle study. y-axis is percentage of all particles, x-axis isthe height over bottom of Reggen.

Second order schemes are used. Simulations are converged at 5×10−6.First order scheme is used to model the turbulence

Simulations are ran as steady state and the pseudo timestep is setto 0.16s.

3.3.3 Simulation method

Simulations are run first with an upwind scheme and then with thesecond order scheme with the upwind result as initial guess. For everynew position of the nozzles a new upwind simulation is done. The sameupwind solution is used for the three different nozzle angles.

27

Bibliography

ANSYS Europe, Ltd. 2010a. ANSYS CFX solver theory guide. AN-SYS.

ANSYS Europe, Ltd. 2010b. ANSYS CFX user guide. ANSYS.

Beck, M., Wirtz, S., & Scherer, V. 2007. Experimental and nu-merical studies of Fe2O3 particle formation processes in a flat flameburner. Chemical Engineering and Technology, 30(6), 790–796.

Demopoulos, G. P., Li, Z., Becze, L., Moldoveanu, G., Cheng,

T. C., & Harris, B. 2008. New technologies for HCl regenerationin chloride hydrometallurgy. World of Metallurgy - ERZMETALL,61(2), 89–98.

Fluent. 2010 (July). History of Fluent.http://www.fluent.com/about/history.htm.

Kladnig, W. 2008. New Development of Acid Regeneration in SteelPickling Plants. Journal of Iron and Steel Research International,15(4), 1–6.

Richard, T (ed). 1997. Low NOx Burner Applications in ProcessHeaters for Refineries and Chemical Plants. XXXVIII ConvencionNacional IMIQ Coatzacoalcos,.

SSAB. 2009. The steel book. 1st edn. SSAB.

SSAB. 2010 (June). History of SSAB.http://www.ssab.com/en/About-SSAB1/History-DOLD.

Versteeg, H.K, & Malalasekera, W. 1995. An introduction tocomputational fluid dynamics the finite volume method. 1st edn. Long-man.

29

Part II

Papers

Paper A

Characteristics of Flow and

Temperature Distribution in a

Ruthner Process

Characteristics of Flow and

Temperature Distribution in a

Ruthner Process

By Simon Johansson1, Vadims Geza2,

Lars G. Westerberg1 and Andris Jakovic2

1 Division of Fluid Mechanics, Lulea University of Technology,

SE-971 87 Lulea, Sweden

2 Faculty of Physics and Mathematics, University of Latvia,

LV-1002 Riga, Latvia

Published in the Proceeding of the International Scientific Colloquium

Modelling for Material Processing, Riga, Latvia, 2010

Abstract

This study is devoted to CFD modeling of the gas flow

and particle dynamics inside the reactor of a furnace used for

regeneration of hydrochloric acid from iron chloride - a rest

product from the pickling process in the steel industry. The

understanding of the dynamics inside the reactor has shown

to be of great importance in order to optimize the process.

So far the process has been a black box, where only the

inflow conditions are known together with the quality of the

final product. In this work the gas flow is resolved together

with the thermal distribution and the particle trajectory for

the injected acid molecules.

35

1 Introduction

During hot rolling of thick metal slabs to thinner sheets in the

steel industry, the surface gets oxidized by the surrounding air.

Before the sheets are further refined in the production line these

surface oxide films are removed in a process where the metal sheets

are dragged through a pickling bath containing hydrochloride acid

(HCl). Due to the formation of iron chloride (FeCl2), HCl loses its

ability to remove iron(II) oxide (FeO) from the surface. Hence fresh

acid is continuously admixed to the container and the used acid is

collected at the other end of the container (Kladnig, 2008). Acid is

hazardous to the environment and has to be regenerated which can

be done by different techniques, such as spray roasting and different

membrane techniques (Regel-Rosocka, 2010). In a spray roasting

reactor FeCl2 is oxidized to iron(III) oxide (Fe2O3) and HCl is re-

generated in a hot furnace. The hourly production/regeneration

rate for such a process typically lies between five and ten cubic

meters of acid.

Previous work about the process by Beck et al. (2007a,b) has

been focusing on the chemistry when drops of used HCl, H2O and

FeCl2 forms Fe2O3 and HCl gas by the reaction

4FeCl2 + 4H2O+O2 → 2Fe2O3 + 8HCl + 55kJ. (1)

According to Beck et al. (2007a,b) water evaporates from drops

at 373 K until the mass fraction of FeCl2 is 0.64 which corre-

sponds to the mass fraction in Thetrahydrate (FeCl2•4H2O). The

temperature of the drops increases and the water evaporates from

FeCl2•4H2O and is fully evaporated at 538K. The ideal reaction

in (1) requires a ratio of H2O to O2 of 4:1 in the surrounding air.

Otherwise several side reactions will appear which can be written

as

36

12FeCl2 + 9O2 → 6Fe2O3 + 12Cl2 + 851kJ (2)

when reduced to simplest form. This process has been modeled nu-

merically by Beck et al. (2007a,b), where the work in (Beck et al. ,

2007a) also is validated with experimental data. The present study

focus on the flow characteristics in the process which is of impor-

tance to understand when optimizing of the process. Optimization

parameters to be investigated involves changes of nozzle position,

spray direction, size of droplets and power of burners. Optimiza-

tion can be done in two ways: either by minimizing fuel costs or by

achieve better quality of the particles.

This work is focusing on the initial two seconds after spray is

introduced, which is the timescale of the vaporisation. After va-

porization particles density will rise due to reaction 1 and 2. These

changes in density is not modelled, therefore will particles only be-

have physical until vaporisation.

The spray roasting reactor used in the present study has a height

of 18.5 m and a middle section radius of 4.3 m. The bottom part of

the reactor is conical shaped with a height of 6.82m and a bottom

radius of 0.23 m. As viewed in Fig. 2 showing the cross section of

the geometry, the main diameter of the reactor decrease fast close to

the top of the reactor. Four natural gas burners placed tangentially

to the reactor introduces heat for evaporation of the droplets, while

the four spray nozzles feeding the system with used HCl are placed

symmetrically at distance of 14.6 m from the bottom and at a radius

of 1.5 m from the centre of the chamber. The nozzles are pointed

5 ◦ towards centre of the reactor. The droplets of acid from the

nozzles have a mean diameter of 370 µm and has a mass fraction of

49 % H2O and 51 % other substances, which are presented in Tab.

1.

37

2 Theory

2.1 Governing equations and turbulence model

The Reynolds Averaged Navier-Stokes equation is considered to

govern the gas motion inside the reactor. As turbulence model, the

Shear Stress Turbulence model (SST) which combines the k−ε and

the k−ω model using a blending function, is considered. Far from

the wall the k−ε model is used while the k−ω is used closer to the

wall. The fluid is treated as non compressible.

2.2 Liquid evaporation model

To calculate mass transfer due to evaporation, Antoine’s equation

is considered

Pvap = Pref × eA− BT+C (3)

where Pvap is the vapour pressure and Pref the gas pressure, A,

B and C are constants and T is the temperature of the droplets.

If Pvap is higher than Pref the mass transfer of water to steam is

calculated using

dm

dt=

−Qc

V(4)

where QC is convective heat transfer and V the latent heat.

3 Numerical modeling

3.1 Geometry and Mesh

Some simplifications of the geometry are considered in order to

simplify the mesh generation and the calculations: i) The lances

which holds spray nozzles are neglected. ii) wall - which in reality

38

is made of bricks and hence imply a non-zero value of the surface

roughness - is treated as smooth (i.e. the surface roughness is not

considered).

In order to obtain reliable results a mesh of good quality and

which have sufficient number of nodes is crucial. Here the region

close to the burner is meshed with tetrahedrals due to sharp angle

in geometry while the rest of the domain is meshed with hexahedral

elements. To verify that enough number of nodes were used a mesh

analysis was made, see Fig. 1.

3.2 Boundary conditions

Each of the four burners is fed with natural gas and air: 71 nm3/h

and 2240 nm3/h respectively, where nm3/h is normal cubic meters

per hour. This corresponds to a mass flow of 0.83 kg/s. The water

mass fraction is set to 0.1. The temperature of gas is 1040 ◦C, which

is measured continuously in the process. The flow is simulated as

non compressible.

The four nozzles has a spray angle of 60 ◦ and are simulated

as point sources with the given spray angle. The droplets are con-

sidered normal distributed with a mean diameter of 370 µm and a

standard deviation of 100 µm. The initial velocity is 13 m/s and the

temperature is set to 60 ◦C. Through the nozzles 1200 particles is

introduced and tracked during the simulations. Eulerian-Langerian

approach is used for particles tracking the coupling between those

is two way so evaporated water from particles will be transferred

into the continuous phase. Forces acting on particles in simula-

tions are buoyancy and dragforce. The heat transfer coefficient

for the wall is set to 0.6 W/m2K. The evaporation is calculated

from 1 using values of the constants such that A = 5.11564 ∗ ln(10),B = 1689.54 ∗ ln(10) and C = −65.15 (ANSYS Europe, 2010). The

39

density of the other products than water is set to 1500 kg/m3 to ful-

fil the initial condition of acid density which is 1240 kg/m3. When

the particles starts to react with O2 and H2O the density will rise

in reality which is not implemented in this model. The two stage

evaporation process described in Regel-Rosocka (2010) is not con-

sidered in the present study but is a highly interesting development

to be treated in the future. Tab. 1 shows the exact composition of

the used acid.

Table 1: Mass fraction of substances in acid.Substance Fe2+ Fe2+ HCl Cl− H2O

g/l 204 28 39 362 607

4 Results and analysis

4.1 Mesh analysis

Mesh analysis of three different grid spacings are compared: 280,000,

780,000 and 1,700,000 nodes respectively. Simulations are made

with the final model, with heat loss and evaporation included. Re-

spectively results are compared in Fig. 1, showing no significant

difference in temperature or mass fraction. The final simulations

have been done using the 780,000 nodes mesh.

4.2 Droplet evaporation

In Fig. 2 the temperature distribution and velocity profile is shown.

The velocity is directed upwards close to the wall, while it makes

a rapid change with a dominantly downward direction at distance

of 0.3 m from the wall, with another change of direction in the

middle of reactor. This profile can be explained by the temperature

40

Figure 1: Mesh analysis of temperature and mass fraction of air

along a line perpendicular to wall.

distribution with hot air from burners rising along the wall, while

in the region of high temperature gradient, the flow changes in

direction and moves towards the bottom of the reactor.

4.3 Validation of numerical model with mea-

surements

The temperature profile has the same characteristics in both cases,

with highest temperature 0.1-0.2 m from the wall and temperature

change in the same region. The difference in temperature in the

region close to the wall can be explained by a low value of the

specific heat for the acid. Furthermore, measurements capture the

convection from the colder walls which could affect the accuracy of

the measurements. Also, the temperature of the exhaust gas could

have been set too high due to a high temperature on the burner

walls which results in convection heat to the thermoelements placed

in the burner chamber.

41

Figure 2: Left: temperature distribution. Right: velocity profile in

vertical direction. Arrows with equal length shows direction of flow

in the plane

4.4 Droplet evaporation

In order to optimize the Ruthner process, one parameter to vary

is the nozzle position. The results in Fig. 4 is taken two seconds

after injection, showing that particle evaporation and distribution

is changing with nozzle position. With nozzles placed close to the

centre, droplets tends to stay in upper part longer with a slower

evaporation as result. Having nozzle position close to the walls

makes the droplets travel faster in vertical direction and evapora-

tion is slightly faster compared with the position 1.5 m from the

centre.

42

Figure 3: Comparison between simulations and measurements.

Temperature is taken along a line, 1 m above burner, perpendicular

to wall.

5 Summary and conclusions

In this study the flow characteristics of a spray roaster furnace has

been investigated numerically. The physical models that has been

introduced to the simulations are evaporation, heat losses thorough

walls, drag force on particles and the dynamics of the gas flow. It

is shown a dramatical change in direction for the vertical velocity

when going from the boundary wall towards the centre of the re-

actor. This is caused by sharp temperature gradients, leading to

significant density variations and dominant buoyancy effects within

the gas flow. These fluctuations play a key role in the overall dy-

namics of the gas flow inside the reactor, and hence also for the

whole regeneration process. A comparison between modeled tem-

43

Figure 4: Particle evaporation depending on placement of nozzle at

radius 1.0, 1.5 and 2.5 m from centre. Dotted line: is particles con-

taining less then 20 % water. Solid line: particles containing more

than 20 % water. Particle distribution two seconds after injection.

perature and data from measurements reveal a good correlation in

the inner region. Also the characteristic of the the two are good,

with the highest temperature at distance 0.2-0.3 m from wall and

the highest temperature gradient in the same region. It is also been

shown how the time of evaporation is effected by the nozzle posi-

tion. The numerical model presented in this paper gives a good

idea of the flow and distribution of temperature in the process rep-

resenting the droplets until evaporation. To give the full picture

of particle distribution however, density change of particles has to

be introduced. This can either be done by implementing the full

chemistry model or by setting a mass exchange between gas and

solid, as a function of time and temperature. An uncertainty in

these simulations is the heat capacity of the solid part of the parti-

cles, here set to 460 J/kgK which corresponds to the value of iron

oxide. When monitoring the solver, fluctuation in velocity and tem-

perature is seen. This can be explained by transient effects in the

process. These effects can be captured by running transient simu-

44

lations, which are very time consuming. Steady state simulations

are therefore to prefer and is a good trade-off between quality and

time.

References

ANSYS Europe, Ltd. 2010. ANSYS CFX user guide. ANSYS.

Beck, M., Wirtz, S., & Scherer, V. 2007a. Experimental

and numerical studies of Fe2O3 particle formation processes in a

flat flame burner. Chemical Engineering and Technology, 30(6),

790–796.

Beck, M., Wirtz, S., Scherer, V., & Barhold, F. 2007b.

Numerical calculations of spray roasting reactors of the steel

industry with special emphasis on Fe2O3-particle formation.

Chemical Engineering and Technology, 30(10), 1347–1354.

Kladnig, W. 2008. New Development of Acid Regeneration in

Steel Pickling Plants. Journal of Iron and Steel Research Inter-

national, 15(4), 1–6.

Regel-Rosocka, M. 2010. A review on methods of regenera-

tion of spent pickling solutions from steel processing. Journal of

hazardous materials, 177(1-3), 57–69.

45

cfd simulation of 2010:144 civ hydrochloric acid regeneration …1024682/fulltext01.pdf ·...

Documents