Numerical Evaluation of Brick - DiVA...

LICENTIATE T H E S I S

Department of Engineering Sciences and MathematicsDivision of Mechanics of Solid Materials

Numerical Evaluation of Brick Lining Status in Rotary Kilns

ISSN 1402-1757ISBN 978-91-7583-478-8 (print)ISBN 978-91-7583-479-5 (pdf)

Luleå University of Technology 2015

Dm

itrij Ram

anenka Num

erical Evaluation of B

rick Lining Status in Rotary K

ilns Dmitrij Ramanenka

Numerical Evaluation of BrickLining Status in Rotary Kilns

Dmitrij Ramanenka

Division of Mechanics of Solid MaterialsDepartment of Engineering Sciences and Mathematics

Lulea University of Technology

SE-971 87 Lulea, Sweden

Licentiate Thesis in Solid Mechanics

Printed by Luleå University of Technology, Graphic Production 2015

ISSN 1402-1757ISBN 978-91-7583-478-8 (print)ISBN 978-91-7583-479-5 (pdf)

Luleå 2015

www.ltu.se

”The cat will mew, and dog will have his day.” Shakespeare

iii

Abstract

Rotary kilns are important in a variety of different manufacturing areas for e.g. calcina-tion and sintering of materials. In fact, two of the most produced materials in the world,cement and iron, are likely to start their journey in a rotary kiln.

A rotary kiln is a large cylinder-formed furnace which rotates about its axis andwhere certain chemical and physical reactions take place by the influence of heat. Theslope and the rotation make the material inside to move through the kiln from feedto discharge end. Due to high process temperatures the casing of the kiln is insulatedby a refractory lining. Service conditions inside the kiln are rough and the lining iscontinuously degrading, especially pronounced in the hot zone of the kiln. If the liningis significantly deteriorated and can no longer protect the casing from the heat − theproduction is shut-down − leading to very high production losses.

Despite many improvements of rotary kilns in the past decades there is still a gap inthe knowledge regarding refractory linings during usage. Many assumptions are based onpractical knowledge. One explanation to this could be the difficulty to study and observethe lining due to kiln size and high operating temperatures. Today, computer programsare of a great help for studying various issues without causing production delays or riskingfailures. However, the field of rotary kilns has stagnated on this matter and very littledocumentation can be found regarding numerical simulations of the lining, especially ofthe thermomechanical character.

Purpose of this licentiate work is to study the mechanical behaviour of the liningby means of the finite element method (FEM). For this, a simplified model of a kilnwas created and various fundamental cases were studied. The commercial FE-softwareLS-DYNA is used for the FE-calculations. The main work is based on cases of the kilnin cold condition. However, an initial study in warm condition is presented as well.The studied lining was a brick lining used in a kiln of dimensions typical for iron-orepelletizing. Additionally, this licentiate thesis gives an overview of some of the mostfundamental issues encountered in a refractory brick lining of a rotary kiln in general.Some material tests are presented as well.

Model’s geometry was based on a section at the position of the support wheels, havinga thickness of one brick. Some simplifications, such as choice of the material model anda rigid riding tyre, were done and a three-fold faster computational time was achieved.Response of the created model was partly verified analytically, by available in-house dataand data from literature. It was confirmed that the model gives a good response.

One of the important findings is that despite variation of conditions in cold state, e.g.rotational speed and relative ovality of the kiln, the induced stresses in the lining remained

v

harmless. This challenges traditional believes which imply that ovality is of considerableimportance for stress generation in the lining. On the other hand, by continuouslytracing gaps between the bricks and the casing, it was found that integrity of the liningwas significantly affected by rotational speed and ovality. Gaps as large as 80 mm couldbe observed between bricks and casing in a worst case scenario.

An initial study on the kiln in hot state was made. Thermal expansion of a perfectlylined and an disordered brick lining were performed. The results indicate that stressesdue to thermal expansion are rising slightly but are harmless in both cases. Additionally,expansion of the kiln stabilizes the lining and the effect of rotation compared to rotationin cold state is small.

Analytical and numerical calculations were compared, indicating that analytical as-sumptions are often coarse and misleading from the reality.

vi

ContentsPart I 1

Chapter 1 – Thesis Introduction 31.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.1.1 Rotary kilns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4Development of rotary kilns . . . . . . . . . . . . . . . . . . . . . 5

1.2 Research question . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71.3 Objective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71.4 Scope of work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

Chapter 2 – Principles of brick lining 92.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92.2 Refractory brick lining and analytical calculations . . . . . . . . . . . . . 10

2.2.1 Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102.2.2 Mechanical effects . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

Ovality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112.2.3 Thermal effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

Temperature profile calculations . . . . . . . . . . . . . . . . . . . 14Thermal load calculations . . . . . . . . . . . . . . . . . . . . . . 15

2.2.4 Chemical effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

Chapter 3 – Modelling 193.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193.2 Model approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193.3 FEM vs analytical calculations . . . . . . . . . . . . . . . . . . . . . . . . 21

3.3.1 Stresses due to ovality . . . . . . . . . . . . . . . . . . . . . . . . 213.3.2 Stresses due to thermal expansion . . . . . . . . . . . . . . . . . . 23

Chapter 4 – Conclusions and future work 274.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 274.2 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

Chapter 5 – Summary of Appended Papers 295.1 Paper A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295.2 Paper B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295.3 Paper C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

References 31

vii

Appendices 35Appendix A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35Appendix B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36Appendix C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38Appendix D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

Part II 1

Paper A 3

Paper B 15

Paper C 35

viii

Acknowledgments

The work presented in this licentiate thesis has been carried out in the group of SolidMechanics at the division of Mechanics of Solid Materials, Department of EngineeringSciences and Mathematics, Lulea University of Technology (LTU), Lulea, Sweden.

I would like to gratefully acknowledge LKAB (Luossavaara-Kiirunavaara Aktiebolag)for the financial support of the work carried out in this thesis.

I would like to thank my head supervisor, Par Jonsen, for his positive support andspirit during our discussions. Also, I am grateful to my co-supervisors Gustaf Gustafsson,Marta-Lena Antti and Lars-Olof Nordin for their participation and contribution in theproject. Furthermore, I would like to thank my previous co-supervisor, Jesper Stjernberg.He has been a great help in various matters and is very much appreciated for his friendlyway.

Warm thanks to my colleagues at LTU. Especially, Hans Ahlin, for his infinite supportin questions regarding modelling but also for the duels on the tennis court and ”pumping”in the gym. Also, many thanks to Stefan Golling, my office-mate and MATLAB-guru,for helping out in various ways and for organizing activities in our division. At last butnot least I would like to thank my friend Florian who has been a great company manytimes during the last years.

Lulea, November 2015

Dmitrij Ramanenka

ix

Part I

1

Chapter 1

Thesis Introduction

“This report, by its very length, defends itselfagainst the risk of being read.”

Winston Churchill

1.1 Background

For hundreds of years iron has been one of the most important materials for mankind. Asthe main constituent in the alloy steel it has explicitly revolutionized the world throughindustrialization started in the 19th century. Today, steel is by far the most producedmetal in the world − over 1600 million metric tons of crude steel were produced in 2013.

Figure 1.1: Iron-ore pellets.

Approximately 70 % of that was manufac-tured from extracted iron-ore, while the restwas made from recycled steel [1, 2].

The most common way of producing newiron is by reduction of raw materials in ablast furnace. In order to make the reduc-tion effective the raw materials are often de-livered in form of iron-ore pellets [3]. Pel-lets are centimetre large porous balls rich inhematite and containing some 70 wt% iron[4], see Figure 1.1. The form and size of pel-lets create pockets in the blast furnace, bythat exposing large surface area to reductionatmosphere. Also porosity and additives in pellets enhance the reduction. These proper-ties affect the final quality of steel and reduce coke consumption, and is the reason whypellets is desired. Manufacturing of iron-ore pellets is mainly done in two, similar, ways:Straight-Grate and Grate-Kiln processes. In the Straight-Grate process, the raw pellets(green pellets) is dried, sintered and cooled on a single line of travelling grates subjectedto different temperatures while it travels. In the Grate-Kiln technology, see Figure 1.2,

3

4 Thesis Introduction

the process is split in three main sections. The green pellets is firstly dried on travellinggrates, continues for sintering in a rotary kiln and lastly cooled on a rotating annularcooler [5].

Waste heat fan

Recoup stream 2

Green balls

Updraftdrying

Recoup stream 1

Down-draftdrying

Temperedpreheat

Preheat

Kiln 1300 oC200oC 60oC 160oC 300oC

FanFan

Fan

Supply fan

Cooling fan 1Cooling fan 2

PelletsRecoup stream 3

Figure 1.2: Schematics of the Grate-Kiln process [5].

The choice of the process partly depends on e.g. type of available ore (magnetiteor hematite), type of fuel to be used (gas, oil or coal) and the end user (blast furnaceor direct reduction). Both systems have their advantages. The Straight-Grate technol-ogy is somewhat cheaper to set up, is easier to maintain and is more flexible since ithas several burners. The Grate-Kiln technology uses much less electrical power, has ahigher quality of fired pellets due to tumbling in the kiln, can combust solid fuels such ascoal and residence time of pellets in the different sections can be controlled separately [5].

As the title of this thesis suggests the interest of this work will further on concernrotary kilns. The design, dimensions and material properties, when used, will referto a specific rotary kiln found in the Grate-Kiln process. However, the principles areapplicable to any rotary kiln.

1.1.1 Rotary kilns

As the green pellets have been dried, eventually oxidized and partially fused on thetravelling grates they continue directly into rotary kiln where final firing is accomplished.At this stage pellets become hard and durable for transportation. It is important thatpellets are of good quality and only a limited amount of pellets is worn down to fineson the arrival to the customer [6]. Figure 1.3 illustrates a typical rotary kiln used in theGrate-Kiln process.

1.1. Background 5

Figure 1.3: Illustration of a short dry-kiln used for sintering of iron-ore (true proportions).

Rotary kiln is a large refractory lined steel container which is slightly inclined (2-4◦) androtates about its axis (2-6 rpm). The slope and rotation make pellets move through thekiln from feed to discharge end. The heat is commonly generated by a single burner inthe discharge end, typically by the combustion of coal, oil or natural gas. The size ofrotary kilns used in the Grate-Kiln process is approximately between 30 to 45 m in lengthand 5-8 m in diameter. The thickness of the casing is typically 50-100 mm dependingon the diameter and the axial position (e.g. the casing is thicker at the position of thetyres). Sintering temperature is around 1300 ◦C but can locally be higher . The kiln iscommonly resting on two pairs of support rollers. It is equipped with solid steel tyresfor stiffening purpose that are riding on the support rollers. Between the riding tyresand the main body (the casing) filler pads are placed as sacrificing abrasion material.The casing is most often freely placed in the tyre without any links (welded or bolted)in-between. The inner part of rotary kiln is commonly lined with single layer of refractorybricks. This is required for heat protection of the steel casing of the kiln, surroundings(such as sensible equipment and personnel), reduction of heat losses (lower drift costs)and maintenance of a specific temperature. [7, 8, 9, 10] See Figure 1.4 for a close up ofthe cross section of a kiln.

Development of rotary kilns

History of rotary kilns begins with the production of Portland cement clinker in the end of19th century. Through out the years of use, rotary kilns has been dramatically improvedin many areas, e.g. fuel, raw materials, refractory lining, quality control, processing etc.

Rotary kilns can be subdivided into wet process kilns, semi-dry process kilns, dryprocess kilns, pre-heater kilns and pre-calciner kilns. The differences lie primarily in themoisture content of feed material and the way of heat exposer. Wet process kilns (40-45% water content) are the longest (> 180 m) and least energy effective kilns where drying,pre-heating and calcination all occur in the kiln. However, the feed material tend tobe more uniform due to wet mixing. Semi-dry and dry process kilns are shorter than


Figure 1.4: Cross section profile of a rotary kiln lined with bricks (true proportions).

wet process kilns due to a shorter drying zone, but are still rather energy inefficient.Today, nearly all newly installed kilns are short, dry kilns accompanied by a pre-heateror pre-calciner with an efficient heat reuse. The specific thermal energy consumptionof this type of kiln plants is about 50 and 75 % compared to that of old wet and dryprocess kilns respectively. This was a great milestone achieved for the kiln plants in the70’s. Along with the energy efficiency also production capacity of a single kiln has beenimproved dramatically, from earliest 300 metric ton/day up to 20 000 metric ton/daytoday. The use of many kilns with low capacity has been replaced by the use of few kilnswith high capacity thanks to a much higher reliability. To some of the improvements canbe counted use of two support kilns with self aligned support rollers. The low numberof supports lowers the risk of misalignment of the kiln which is often responsible for thewear problems of the rollers, tyres, gears etc. Furthermore, the kiln casing in modern,newly installed kilns has different thickness in different zones by this lowering ovalityof the kiln in sensible parts and allowing higher rotational speed. Flexible girth gearfastening system was developed replacing bolted or welded systems. Better design ofchair pads lowers their wear and facilitates replacement. Design of burners has beenimproved allowing better control of the length, width and direction of the flame. Therehas been large development of the lining materials, especially for the hot zone of the kiln.The lifetime of the earliest linings in hot zones did not exceed more than 10-15 days. Notuntil in the 50’s/60’s the effective lifetime of the lining in hot zones reached 200 days insome plants. Today, lifetime above 300 days is common in some processes. Of course,this is not due to the material development only but also thanks to an overall bettermanagement of a kiln. There has been substantial improvements in the maintenancetechniques. Surface temperature of the casing, rollers and tyres, the lining and the

1.2. Research question 7

feed material is continuously monitored. The air gap between the tyre and the padsis commonly monitored in order to avoid risk of seizure due to mismatched thermalexpansion. Vibration measurements are often performed in order to detect possibleissues in advance. Axial pressure exerted on the roller bearings is measured and eventualmisalignment can be adjusted. These are just some of the technological improvementsmade through the years of use. [11, 7, 8, 12]

1.2 Research question

Rotary kiln is a central part of the Grate-Kiln process. Despite the many improvementsthe kiln is always subjected to the risk of failure. Some factors, such as corrosion andwear of components do usually not require immediate attention. Others are much morecritical. One of the most critical factors controlling availability of the kiln is the status ofthe refractory lining. Failure of the refractory lining very often requires immediate shut-down of the production. Unplanned shut-downs cause (very) high production losses andput company in a demanding situation. Due to the need of slow cooling and repairingfollowed by slow heating of the kiln, the process of maintenance is time consuming (5-14 days). Since a single kiln might be responsible for maybe 50 % of the company’sproduction capacity it is very important to minimize the risk of unplanned shut-downsdue to brick lining failure. Therefore, the main research question in this project is: ”Isit possible, by means of numerical simulations of the kiln behaviour, to detect criticalsituations affecting the brick lining and by that minimize the risk of unplanned shut-downs due to brick lining failure?”

1.3 Objective

Objective of this work is to create an efficient numerical model of a rotary kiln thatcan be used for studying brick lining for a better management of the rotary kiln and assupport in decision making concerning rotary kilns in LKAB.

1.4 Scope of work

This work is focused on mechanical and thermomechanical aspects of the refractory bricklining used in rotary kilns for iron-ore pelletizing. Fundamental issues encountered bythe refractory lining are discussed. Some analytical calculations are performed in orderto better understand the state of refractory lining.

Furthermore, simplified numerical models of rotary kiln have been created. Thepurpose is to study the effect of various factors on the load state and behaviour of thebrick lining by means of finite element method (FEM). The main work is performedin the cold state of the kiln. However, an initial study in hot conditions is presentedas well. Commercial software, LS-Dyna [13], is used for the numerical calculations.


The numerical part of the work is mainly presented in the attached articles. Also, thetraditional analytical and the numerical calculations are compared with each other.

Chapter 2

Principles of brick lining

2.1 Introduction

Severe damage of the brick lining is usually presented by the fall outs of the bricks orsignificant thickness reduction of the bricks. This leads to the formation of so calledwarm and/or hot spots on the surface of the steel casing, commonly detected by an IR-camera, see Figure 2.1. Warm spots generally indicate spalling (thickness reduction) of

Figure 2.1: Left) Brick fall outs in the lining with spalling around area. Right) Brick fall outand spallation detected by IR-camera.

the brick lining. This stage requires surveillance but not necessarily shot-down of thekiln, depending on the severity and e.g. the remaining time to the planned shut-down.Hot spots on the other hand indicate fall out of the lining and leads in general to animmediate shut-down.

Casing of the kiln is a very expensive investment that is planned to last for decades

9

10 Principles of brick lining

and therefore is important to be handled with care. Warm and hot spots risk permanentdeformation of the casing and making it less perfect. Damages to the casing worsenintegrity of the brick lining, which additionally augments the risk of future fall outs.Sometimes an unplanned shut down can temporarily be delayed by promotion of slagformation over the damaged section. This can be done e.g. by cooling the area by watersprays. However, the casing is subjected to the risk of crack formation and is a badlong-term alternative.

2.2 Refractory brick lining and analytical calcula-

tions

Refractory products are often used in harsh service environments and therefore are proneto degradation. Tolerance to high temperatures, mechanical loads, thermal cycling, wearand chemical resistance are some of the common requests [14, 15, 16]. In practice, degra-dation of refractory lining of the kiln is inevitable. Common procedure is that the liningof the kiln is controlled and/or replaced at regular basis during planned maintenanceshut-downs. The best case scenario is when refractory lining is degrading in a controlledmanner without causing shut-downs in-between maintenance stops.

In general, life of the lining is influenced by mechanisms of thermal, mechanical andchemical character, and the coexistence of them. Dependent on the manufacturing pro-cess the severity of these effects may vary. To some degree these effects can be controlledby the choice of the refractory material. However, total immunity cannot be achieved bythe material choice. Therefore, it is important that technical limitation of the lining ma-terial is not violated and that the relevant steps in the process, e.g. heating and cooling,are adjusted to that.

2.2.1 Materials

High-alumina bricks [17] are commonly used as insulators in rotary kilns for iron-ore pel-letizing. The base materials are usually derived from bauxite or andalusite ores. Aftermixing and preparation the bricks are fired in approximately 1350 ◦C [18, 19] for creationof ceramic bonds. After sintering the bricks are usually dominated by two combinationsof phases: 3/2-mullite (3Al2O3 ·2SiO2) together with silica (SiO2) or α-alumina (Al2O3)together with silica. The alumina (the constituent, not the phase) content in the bricksis some 50-80 wt% and silica content some 15-40 wt%. The rest are small amounts ofhematite (Fe2O3), rutile (T iO2), burnt lime (CaO) and alkalis. Some part of the oxides ispresented in amorphous phase, forming up to 20 wt% of glass. An interesting behaviourthat has been documented by the author for some types of aluminasilicate bricks is thatthey increase their compression strength with rising temperature. In some cases morethan twice its original strength, without relation to phase transformations. For furtherreading about material properties and material data used in this work advice Paper C.

2.2. Refractory brick lining and analytical calculations 11

Some of the fundamental ideas regarding the brick lining will be discussed in thefollowing sections.

2.2.2 Mechanical effects

Brick lining is tightly fitted to the steel casing of the kiln. During the usage of the kiln(starts, stops, rotation) the steel casing and therefore the lining, are subjected to radialand longitudinal bending, vibrations and torsion. Additional stresses can typically arisefrom misalignment of the kiln or other abnormalities [20]. This results into differentstress-controlled loads in the lining. Here, stress-controlled loads define external loadssuch as gravity load, pressure load or any type of mechanical load.

Ovality

Radial bending of the kiln, known as the ovality of the kiln, traditionally belongs to oneof the most important load generators in the refractory lining and directly affects thelifetime of it. Ovality is an elastic distortion of the kiln casing that arise due to the gravityforce. The weight of the casing, the lining, the material charge and kiln’s hollow shapemake the casing somewhat oval rather than circular. This is illustrated in Figure 2.2.

g

⇓

Figure 2.2: Arbitrary representation of un-strained (solid line) and strained (dashed line)kiln shell profile.

The deviation from the circular line (δv andδh) may reach some 10-15 mm in a large-sized kiln. Due to the ovality of the steelcasing the lining will experience load oscilla-tions during rotation of the kiln. In a periodof 24 h a kiln for iron-ore pellet productiongoes typically through 2000-4000 revolutionsand the double amount of load oscillations(due to vertical symmetry line). This maylead to the formation of cracks and even-tually to the spalling of the refractory lin-ing. Additionally, bricks in the lining areforced to shift their relative position to eachother due to the ovality. By that opening upthe joints, leading to worsen integrity of thebrick lining which may cause unhealthy stress concentrations. [20]

Measurement of the ovality of a kiln is usually performed with a device called shell-tester [7] at a position close to a riding tyre during operation of the kiln in order toinclude the effect of the temperature. It is commonly presented in percent, as relativedeformation to the nominal diameter. If the deformation is known then relative ovality(ωr) is found by Equation 2.1.

ωr =δv + δhD0

· 100% (2.1)

where δh and δv are the horizontal and vertical deviations from the circular line and D0


is the nominal inner diameter of the casing. A rule of thumb based on the experience hasbeen established suggesting that the ovality of the steel casing should not exceed 10 %of the nominal inner diameter. E.g. a kiln with a nominal inner diameter of 7 m shouldnot have relative ovality of more than 0.7 %. [20, 7]

The magnitude of ovality is primely dependent on the thickness of the steel casing,the gap between the tyre and the pads and the operating temperature. The ovality ishighest near the tyres and statistically most of the repair jobs of the lining are doneclose to the downhill-tyre (closest to the flame). Additionally, the ovality of kiln is notat permanent state but is changing with operating conditions. When the lining is newlyinstalled the ovality tends to be at it’s lowest point and increases after some time. Wearof chair pads gradually increases ovality. If the lining is covered with protective slag, theeffect of temperature is lowered and therefore the ovality is lowered as well. [7]

Magnitude of compressive stresses in the lining induced by the ovality of the kiln canbe traditionally approximated analytically by Equation 2.2. Deduction of the equationcan be found in Appendix C.

σL = 3δh + δvD2

0

ELtL (2.2)

Where σL is the compressive stress induced on the inner wall of the lining, EL Young’smodulus of the lining material and tL is the thickness of the lining. From the equation canbe seen that stresses increase with increased ovality, Young’s modulus and thickness ofthe lining. Figure 2.3 graphically represents relationship between stress, ovality, diameterand thickness of the lining with typical dimensions for large-sized kilns. Young’s modulusis set to 10 GPa. A healthy kiln typically has ovality of around 0.5% − implying that the

0.2 0.4 0.6 0.8 1 1.20

2

4

6

8

10

12

14

16

Relative ovality, ωr (%)

Com

pres

sive

str

ess,

σL (

MP

a)

6 (200)

7 (200)

6 (250)

7 (250)

5 5.5 6 6.5 71

2

3

4

5

6

7

8

Inner diameter, D0 (m)

0.3 (200)

0.3 (250)

0.5 (200)

0.5 (250)

Figure 2.3: Compressive stress dependency in the lining with respect to ovality, initial innerdiameter and thickness of the lining. Notations inside figure: 6 and 7 - initial inner diameter(m), 200 and 250 - lining thickness (mm), 0.3 and 0.5 - relative ovality (%).


lining experiences maximum compressive stress of some 4-6 MPa in a large-sized kiln.This will later be compared to numerical calculations in Chapter 3.

2.2.3 Thermal effects

Thermal expansion of the lining, the casing and the tyre are fundamental issues relatedto the temperature. The gap between the tyre and the pads has to be sufficient for thethermal expansion of the casing. Too tight riding tyre can inhibit thermal expansion ofthe casing leading to the failure of the lining or even the tyre. On the other hand, thegap cannot be too large as it affects ovality negatively. Too fast heating of the kiln canlead to brick spallation or seizure of the casing in the tyre. The internal loads inducedby thermal expansion are defined as strain-controlled loads.

The lining is subjected to a higher temperature than the casing and expands partlymore than the casing. Some part of the expansion of the lining is compensated by theexpansion of the casing. Additionally, some part of the expansion is swallowed by thenatural joints between the bricks (see Appendix B). However, a part of the expansion isleading to increased stresses experienced by the lining. The advantage of the expansionis the increased stability of the lining. However, it is important that the expansion iswithin the technical limitation of the material. Some additional flexibility to a bricklining can be given by installing the lining with mortar or steel plates between the radialjoints. The steel plates and mortar erode during heating up and give additional expansionspace to the lining - by that lowering stresses. However, in cold state, the lining becomeslooser than normally after erosion of steel plates or mortar, which might be problematicif cooling and reheating of the kiln has to be done.

The expansion rate of the kiln is also of importance since the lining is heated fasterthan the casing and the casing is heated faster than the riding tyre. The opposite isapplicable during cooling. The casing and the lining risks seizure due to mismatchedthermal expansion if the heating or cooling is made too fast. Typical heating and coolingprocedure requires 24-36 h. In addition to the thermal loads due to expansion, the hotface (face of the lining exposed to the flame) of the lining experiences also thermal loadoscillations during rotation. This can be due to e.g. relatively fixed flame direction andgas flow, and the cooling effect on the lining when it passes bellow the bed of material.The oscillations lead to local stresses which risk spalling of the lining. Other factorsaffecting thermal load are e.g. the operation of the burner and the protective slag forma-tion. Misaligned burner or badly controlled power output of the burner can cause criticaltemperature peaks in parts of the lining leading to mismatched thermal expansion. Theslag is a very effective insulation and protection for the lining. In some processes, suchas cement clinker production, the slag formation is vital for the manufacturing. In otherprocesses it is not as critical. Sudden slag fall out in an area may locally increase thetemperature several hundred degrees, causing mismatched thermal expansion with therest of the lining [21].

In the following sections some analytical calculations are performed for analysing


thermal effects on the lining. For this purpose, data for a specific rotary kiln is defined.Consider a rotary kiln with following input data. Brick lining having: thickness,

tL = 250 mm; outer radius, RLO = 3429 mm; average radius, RLM = 3304 mm; hotface temperature, TH = 1300◦ C; Young’s modulus, EL = 10 GPa; coefficient of thermalexpansion, αL = 6 · 10−6 and thermal conductivity, λL = 2 W/(m ·K). Furthermore,steel casing having: thickness, tS = 100 mm; inner radius, RS = 3429; outer surfacetemperature, TS = 300◦ C; Young’s modulus, ES = 205 GPa; coefficient of thermalexpansion, αS = 12 · 10−6 and thermal conductivity, λC = 45 W/(m ·K).

Temperature profile calculations

Temperature profile through the brick lining and the steel casing can be found analyti-cally. By knowing temperature of the hot face, TH , and the outer surface of the casing,TS, the temperature in the intersection of the lining and the casing, TC , is found by (seeAppendix C):

TC = TSλL

tL+ TH

λC

tS(2.3)

The intersection temperature or temperature of the cold face is: TC ≈ 317◦ C. Assuminglinear relation the temperature profile in the brick lining and the casing in a steady stateis according to Figure 2.4. Non-linear FEM calculation confirms that linear assumptionis satisfactory. Figure 2.5 visualizes temperature distribution in a brick and steel casingcalculated by finite element method.

0 50 150 250 350300

500

700

900

1100

1300

Distance from hot face (mm)

Tem

pera

ture

( o C

)

Brick lining − AnalyticalSteel casing − AnalyticalBrick lining − FEM

Figure 2.4: Temperature profile through the brick lining and the steel casing.

Due to temperature gradient in the brick lining the expansion of the lining is not eventhrough its thickness. For a typical brick the difference in expansion between cold faceand hot face is some 1-1.5 mm. Due to this difference the radial joints of the bricks are


not having full contact with each other, see Figure D1 in Appendix D for an illustration.In practise it is difficult to define the size of the contact zone, since the bricks are notperfectly ordered, have some joint absorption, elastic deformation etc. However, for thefurther discussion will be assumed that the contact zone is 2/3 of the brick’s thickness.It will be shown, in the following section, that a theoretical limit can be iterated byanalytical calculations based on the assumption that bricks fracture at low tensile stress.

Thermal load calculations

Figure 2.5: Fringe levels of temperature (K)distribution in a brick and steel casing, calcu-lated by FEM.

Induced stresses in the lining due to tem-perature increase can be divided into twocomponents. Firstly, stresses that arisefrom the expansion of the lining andresistance of the casing to that expan-sion. If the lining can expand freelyno such stresses can be induced. Sec-ondly, stresses due to temperature gradi-ent through the brick. If temperature ofthe brick is homogeneous no such stresseswill be induced. [22] These stress compo-nents are illustrated in Figure 2.6. Fur-ther on, thermal stresses in the liningwill be approximated analytically accord-ing to methodology suggested by Schacht[22].

Stresses due to expansion difference

As defined previously 2/3 of the brick’s thickness will be considered to have full con-tact with another brick. Therefore, tL∗ = 2/3 · tL = 167 mm and temperature at thatthickness TC∗ = 625 ◦C, is found from Figure 2.4. Average temperature increase of thebrick lining in contact zone is:

ΔTL = (TH + TC∗)/2− TR (2.4)

Where TR = 21◦ C is the room temperature.

Difference in increase of radius, ΔδT , due to different thermal expansion between thelining and the casing is found by:

ΔδT = αLRLOΔTL − αSRSΔTS = 7.4 mm (2.5)

where ΔTS is average temperature increase in the steel casing. A positive value ofthe expansion difference implies that the brick lining expands more than the casing andtherefore the casing will impose compressive load on the lining. The relationship between


0

0

Figure 2.6: left) Illustration of stress components in the lining that arise due to expansiondifference between the lining and the casing (left figure) and temperature gradient in the lining(right figure).

the radial pressure, P , imposed by the expanding lining on the casing and the expansiondifference can be defined as:

ΔδT = P

(R2

LM

tLMEL

+R2

S

tSES

)(2.6)

Solving for P with equations 2.5 and 2.6 gives P = 1 MPa. The tensile stress in thecasing, σS, that resists lining’s expansion is found by:

σS = PRS/tS = 35.6 MPa (2.7)

Analogously for the lining the average compressive stress in the contact zone, σLC , is:

σLC = PRLM/tL∗ = −20.6 MPa (2.8)

Stresses due to temperature gradient in the lining

The stress due to temperature gradient in the brick, σLG, at depth tL∗ is derived fromthe Hooke’s law:

σLG = (ELαL(TH − TC∗))/2 = 20.3 MPa (2.9)

The total average stress σTOT in the brick lining at depth tL∗ from the hot face is:

σTOT = σLG + σLC = −0.3 MPa (2.10)

The stress is a low compressive stress, therefore the approximation of contact zone to2/3 · tL was satisfactory. The contact zone could be iterated even more precisely sincethe bricks can resist some tensile stress, however this will not be performed here.

Analogously the stress at hot face is found by:

σTOT = −σLG + σLC = −41 MPa (2.11)

Which is a compressive stress. Figure 2.7 graphically represents resultant stress profilein the lining after thermal expansion calculated by the methodology above.


For pedagogical reason and for a later use, an analogues calculations have been per-formed for a brick who has a contact zone along its whole thickness. Bottom figure inthe Figure 2.7 graphically represents resultant stress profile for that case. Compressivestress on the hot side of the brick is only slightly changed compared to the previous case.However, dangerous tensile stress arise on the cold side of the brick. Additionally, inboth cases is noted how Young’s modulus of the lining is important for the generation ofthe thermal stresses.

300 400 500 600 700 800 900 1000 1100 1200 1300−20

−10

0

10

20

30

40

E=10 GPa

E=5 GPa

Δ T (°C)

Str

ess

(MP

a)Distance from cold face (mm)

50 100 150 200 250

300 400 500 600 700 800 900 1000 1100 1200 1300−20

−10

0

10

20

30

40

E=10 GPa

E=5 GPa

Δ T (°C)

Str

ess

(MP

a)

Distance from cold face (mm)

50 100 150 200 250

Figure 2.7: Stress profile in the lining due to thermal expansion for a brick which has a contactzone of 2/3 of its thickness (upper figure) and for a brick with a contact zone along its wholethickness (bottom figure). The filled areas above zero stress mark the possible range of com-pressive stress variation due to variation of Young’s modulus. The filled areas below zero stressmark the range of tensile stress: green − a hypothetical, non-existing stress due to the absenceof contact in that area (for visualization); red − an existing tensile stress due to full contactbetween bricks. The upper x-scale denotes at which thickness the temperature and stress levelsare found. Only true for the cases presented in the calculations above. (Notations in the figuredenote Young’s modulus of the lining).


2.2.4 Chemical effects

Chemical effects are highly dependent on the process environment. In general, chemicalattack on the lining in rotary kilns is often associated with presence of alkalis (Na, K),sulphur and chlorine [8]. The severity increases with temperature and therefore hot zoneof the kiln is affected mostly. A good choice of the material with respect to the processenvironment is the primary solution to the issues of chemical character. Rotary kilns foriron-ore refinement have relatively mild corrosive conditions. A number of papers on thisissue regarding aluminasilicate bricks where published by Stjernberg et al. [23, 24, 25].

Chapter 3

Modelling

3.1 Introduction

Finite element modelling [26] is a widely used numerical solution method which is typi-cally used for evaluation of stresses and strains of a system. However, despite the wideuse of FEM in calculations of mechanical problems its use for simulation of rotary kilnsis poorly reported in scientific journals, especially that of thermomechanical character.Some information can be collected from the researches working in the field of steel-making industry, where refractory materials are extensively used. Damhof et al. [27]present a FE-model of thermal shock damage in the refractory lining of steel-makinginstallations. Others [28, 29, 30, 31] have presented a number of articles on the subjectof FE-simulation of thermomechanical behaviour of refractory linings. del Coz Diaz et al.[32] make a Finite Element Analysis (FEA) of a cement rotary kiln evaluating ovalizationand stresses in the steel casing. Currently there is little attention from academic researchon FE-simulation of rotary kilns especially that includes brick lining and evaluates itsmechanical response in static or dynamic cases, in cold or hot state.

3.2 Model approach

A full scale finite element model of a whole kiln with satisfactory mesh size is today tooheavy to handle in a practical way. Therefore, in this work a 3D model of a cross sectionof the kiln at the position of the riding tyres was created, see Figure 3.1. At this positionthe refractory brick lining experience the largest amount of failures [8]. Also, boundaryconditions at this position are easily defined. For further details on the model advisePaper B.

Despite its simplicity the created model is still rather computationally heavy, e.g.it takes approximately 50 h to calculate one revolution of the kiln in cold state. It isprimarily contacts between the bricks that require long computational time. Therefore,

19

20 Modelling

Figure 3.1: Cross section profile of the kiln used in modelling, dimensions in mm.

one of the main challenges have been and still is to make calculations more efficient.Here are some of the simple things that were verified and proven to be important forsaving computational time and avoiding premature termination when modelling the kilnin LS-Dyna:

• Solving different cases implicitly is some 2-5 times faster than solving explicitly.

• MPP (Massively Parallel Processing) is in most cases faster than SMP (SharedMemory Parallel processing).

• Mesh size between 25 and 35 mm is satisfactory for the whole model.

• Making support wheels rigid does not affect results.

• Replacing elastic riding tyre with a rigid riding tyre affect results only negligiblybut saves computational time three fold.

• There is no justified need to use the computationally heavier material models inthe definition of bricks − at least not in cold state. Simple elastic material modelis satisfactory.

• Slow ramping of gravity − typically for 1 s in the initial simulation and 5 s ifstarting with saved stresses in the model.

• Slow ramping of rotation − typically for 20 s if ramping up to 2 rpm.

• Time step is critical. Using adopted time step works well. However, during startwith saved stresses in the model, small initial time step (1 · 10−8) was found to beimportant. Additionally, maximum time step should not be too large (< 1 ·10−3) in

3.3. FEM vs analytical calculations 21

order to avoid premature termination and convergence failures that require time-consuming recalculations.

• Thermal conductivity of brick material can be raised from 2 to 200 WK−1m−2 forfaster thermomechanical calculations when reasonable. Higher than 200WK−1m−2

tend to lead to error termination.

3.3 FEM vs analytical calculations

In this section the analytical results obtained in chapter 2 will be compared with nu-merical results from the simulations. The results are based on the same dimensions andmaterial properties of the kiln and the lining.

3.3.1 Stresses due to ovality

The previously found analytical results in Figure 2.3 imply that a lining with ovality of0.4 % would have maximum compressive stresses of approximately 4 MPa. Figure 3.2shows numerical calculation of a similar case with ovality of approximately 0.4 %. Theobtained numerical results show that maximum compressive stress is nearly 3 MPa onthe inner wall of the lining. However, the boundary conditions are not exactly the same.In the numerical model the casing is resting in a riding tyre and in the analytical casethe casing is resting on the ground (see Appendix A). Additionally, the numerical modelis in three dimensions, therefore a less stiff response can be expected. Nevertheless, thisexample can be regarded as a vague validation of the model’s response.

12 3 6 9 120

0.5

1

1.5

2

2.5

3

Clock position

Effe

ctiv

e vo

n−M

ises

str

ess

( M

Pa)

Figure 3.2: left) Distribution of effective von-Mises stress (Pa) in the lining without brick joints.right) Oscillation of effective von-Mises stress on the inner wall of the lining during rotation.

22 Modelling

On the other hand, non of the cases above is reflecting reality in a good way, pri-marily due to the absence of brick joints. Figure 3.3 shows effective von-Mises stressdistribution in a brick lined kiln obtained from numerical simulation of a more realisticcase, including brick joints and small initial hinge (see Paper B). Additionally, figure 3.4shows variation of effective von-Mises stress on the bricks’ cold face during rotation inkilns having different ovalities. The results indicate that maximum stress reaches 0.5MPa while average stress 0.2 MPa, which is one order of magnitude lower compared toanalytical results. The joints make the lining much more flexible and therefore lowerstress is obtained. Steinbiss [33] measured, in his study, maximum compressive stress inthe lining of cold kiln to 0.4 MPa. For further discussion on the lining behaviour due tovariation of different parameters advise Paper B.

Figure 3.3: Distribution of effective von-Mises stress (Pa) in the lining with brick joints, kilnwith ovality of 0.4 %.

0 20 40 60 800

0.1

0.2

0.3

0.4

0.525 mm 0.2 rpm E=10GPa μ=0.75

Brick ID

Effe

ctiv

e vo

n M

ises

str

ess

(MP

a)

Mean ± standard deviation

0 20 40 60 800

0.1

0.2

0.3

0.4

0.535 mm 0.2 rpm E=10GPa μ=0.75

Brick ID

Effe

ctiv

e vo

n M

ises

str

ess

(MP

a)


0 20 40 60 800

0.1

0.2

0.3

0.4

0.540 mm 0.2 rpm E=10GPa μ=0.75

Brick ID

Effe

ctiv

e vo

n M

ises

str

ess

(MP

a)


Figure 3.4: Effective von-Mises stress variation on bricks’ cold face during rotation. Ovalityvariation from left to right: 0.6, 0.4 and 0.27 %. (Collected from Paper B).


3.3.2 Stresses due to thermal expansion

The previously shown analytical results in Figure 2.7, for the case where bricks havefull contact with each other, imply that maximum compressive stresses due to thermalexpansion are some 42 MPa on the hot face of the bricks. A corresponding numericalcase indicate that maximum compressive stress is around 40 MPa, see Figure 3.5. Theresults correspond well with each other and the case can be seen as a validation of themodels response. The small difference can be related to the fact that numerical modelis in three dimensions − giving a less stiffer response. However, again, the studied case

Figure 3.5: Distribution of effective von-Mises stress (Pa) in the lining due to thermal expansionobtain by numerical modelling imitating the analytical case for bricks having full contact witheach other.

above is not very realistic. The analytical case consider a lining that is completely tightwith the casing. A perfectly tight lining does not exist in reality. In fact, there is alwaysa small hinge on top of the lining (position at 12 o’clock) which absorb some of theexpansion (see Paper B). Furthermore, the boundary conditions of the analytical case donot consider brick motion in relation to each other − basically a monolithic lining.

Figure 3.6 shows stress distribution in a case where the brick lining is tight (having noor very small hinge) and brick joints are considered i.e. brick motion is allowed. Perhaps,this case can be regarded as a more realistic case of the analytical results from Figure2.7 which shows stresses for the bricks having contact zone of 2/3 of its thickness. Themaximum compressive stress in the numerical case reaches some 19 MPa, approximatelyhalf of the analytically found results. This implies the importance of consideration ofbrick joints.

Now consider a case where the lining has a hinge of approximately 20 mm at twelve

24 Modelling

Figure 3.6: Distribution of effective von-Mises stress (Pa) in the lining due to thermal expansionobtain by numerical modelling a case with a tight lining and with consideration of brick joints.(Missing piece of the lining was intentionally removed during collection of the results due to anunessential numerical reason).

o’clock position prior the thermal expansion. Brick joints and gravity force are considered(see Paper C). Figure 3.7 shows effective von-Mises stress distribution in the lining of suchcase. Additionally, variation of effective von-Mises stress on the hot face of the bricksduring rotation of the expanded kiln can be found in the same figure. The numericallyobtained results indicate that average von-Mises stress is in the vicinities of 1 MPa andthe maximum stress is slightly higher (1.7 MPa).

The series of calculations clearly show the importance of correct and realistic bound-ary conditions in the model. Based on the results presented above, perhaps it can beexpected that compressive stresses on the hot face of the lining, in steady state temper-ature conditions, vary somewhere between 1 to 20 MPa. The stress is highly dependenton the initial boundary conditions e.g. size of the hinge. In any case, more investigationis needed.


0 20 40 60 800

0.5

1

1.5

2

Brick ID

Effe

ctiv

e vo

n M

ises

str

ess

(MP

a)Mean ± standard deviation

Figure 3.7: left) Distribution of effective von-Mises stress (Pa) in the lining with a hinge ofapproximately 20 mm prior the thermal heating and with consideration of brick joints. right)Variation of effective von-Mises stress on the bricks hot face during rotation of a lining afterthermal expansion. (Missing piece of the lining was intentionally removed during collection ofthe results due to an unessential numerical reason).

26 Modelling

Chapter 4

Conclusions and future work

4.1 Conclusions

This licentiate thesis has described some of the fundamental issues encountered by arefractory lining in a rotary kiln. Analytical and numerical calculations for differentcases were performed. Main work was summarized in the attached papers.

It was shown that the created numerical models were able to reproduce analyticalcases in a satisfactory way.

With help of created numerical models the more realistic cases influencing brick liningcould be studied. The numerical solutions of the more realistic cases showed that thestudied analytical assumptions were very coarse and over predicted stress levels.

It was shown that initial boundary conditions are highly important in the generationof stress levels. The analytical assumptions should be used with care since they can bemisleading from the actual causes of failure.

Results in Paper A indicated that stress levels of the lining in cold condition in staticcases and at slow rotation are harmless.

In Paper B was found, amongst the studied factors in cold condition, that it is primar-ily fast rotation, in combination with high ovality, that could provoke unhealthy statusof the lining. Furthermore, the results indicated that in most of the cases the ovality hasinsignificant influence on the stress levels in the lining. However, the gaps between thebricks and the casing were found to be considerably related to the ovality.

The study in Paper C indicated that stress levels in the lining due to thermal ex-pansion of the lining with an initial hinge of 20 mm, are harmless. Additionally, bedintegrity of the lining and fast rotation did not influence stress levels in a considerableway. Material tests showed that bricks increased their strength when tested in elevatedtemperature.

In conclusion, the future challenges can be split into two parts. Firstly, to find thecritical situations that affects the brick lining − a challenge related to the understanding

27

28 Conclusions and future work

and knowledge of the kiln. And secondly, being able to create realistic and computation-ally effective models − a challenge related to numerical modelling.

4.2 Future work

Future work will mainly focus on the hot state of the kiln, with LS-Dyna as primary tool.It is of primarily interest to investigate heating and cooling sequences of the kiln

during start-ups and shut-downs. E.g. what should the maximum heating-up rate be?Furthermore, it is of interest to study sudden temperature variations affecting the

lining, e.g. slag fall-out and unstable burner.Investigation of the relation between stresses and how tight the lining is at its initial

stage prior the expansion will be performed.Since previous material tests lack some statistical foundation, new improved material

tests of bricks are as well in the consideration.To include consideration of the neighbouring bricks in the axial direction in the model

would be advantageous. Also, a global coarse model of the whole kiln with, perhaps, somefine-meshed areas of interest could be a goal.

Creation of the more advanced material model of the brick material for modelling inhot conditions can be advantageous, especially when studying the brick lining on a morelocal level.

Two articles with following titles are reasonable to aim for in the near future:

• Paper D: ”Simulation of heating and cooling sequences of a rotary kiln.”

• Paper E: ”Compression tests of mullite refractory brick at elevated temperatures.”

Chapter 5

Summary of Appended Papers

5.1 Paper A

”Modelling of refractory brick furniture in rotary-kiln using finite elementapproach”: This paper is a contribution to the proceedings of 11th World Congresson Computational Mechanics. In this work a model of a cross section of a rotary kilnat the position of support rollers was created. The ovality of the kiln was approximatedand a brick laying method was proposed. Von-Mises stress of the brick lining in coldstate was evaluated in static (elastic and rigid riding tyre) and dynamic cases (slow andfast acceleration). It could be concluded that the brick lining did not experience criticalsituations in the evaluated cases.

5.2 Paper B

”FEM investigation of global mechanisms affecting brick lining stability in arotary kiln in cold state”: Paper B is a follow up on the work made in Paper A. Themodel is made simpler and more effective, the choices behind simplifications are justified.Various cases investigate behaviour of the brick lining in cold state. Influence of variedovality, brick’s Young’s modulus and friction coefficient on the induced stress and brickdisplacement are evaluated at two rotational speeds. Again, it was found that variedconditions in cold state do not affect stress levels experienced by the brick lining in aconsiderable way. Also, integrity of the lining is studied by continuously tracing the gapbetween the bricks and the casing. It was found that brick displacement is significantlyaffected by rotational speed and ovality. In the worst case scenario gaps as large as 80mm could be observed between the bricks and the casing.

29

30 Summary of Appended Papers

5.3 Paper C

”Hot Compression Strength of High-Alumina Refractory Bricks and Mod-elling of Brick Lining in Rotary Kiln”: In the work presented in Paper C threecommercial aluminasilicate bricks were tested in compression until failure in the rangeof 25-1300 ◦C. The purpose was to find Young’s modulus and compression strength forthe fired brick materials at elevated temperatures. The data was later used for modellingof brick lining in a rotary kiln in hot state. It was found that brick materials increasedtheir compression strength with temperature until approximately 1000 ◦C. For one bricktype the increase was over 150 % compared to its compression strength at room tempera-ture. Young’s modulus was found to measure 2-10 GPa in the tested temperature range.However, no clear tendency between Young’s modulus and temperature increase could beobserved. Furthermore, after 1000 ◦C the materials gradually become more visco-plastic.

The modelling of hot rotary kiln indicated that induced stresses due to expansion ofthe system were harmless under studied conditions.

References

[1] World Steel Association, World Steel in Figures, ser. ISBN: 978-2-930069-73-9, 2014.

[2] M. F. Ashby, Materials and the Environment : Eco-informed Material Choice, ser.ISBN: 9780080884486. Butterworth-Heinemann, 2009.

[3] K. Walker, “The basics of iron and steel making - part 1 terminology,” Steel TimesInternational, pp. 38–40, 2012.

[4] LKAB, “Products.” [Online]. Available: https://www.lkab.com/en/About-us/Overview/Products/

[5] Metso Corporation, Basics in Mineral Processing, 10th ed. Metso Corporation,2015.

[6] S. Forsmo, S.-E. Forsmo, and P.-O. Samskog, “Mechanisms in oxidation and sinteringof magnetite iron ore green pellets,” Powder Technology, vol. 183, no. 2, pp. 247 –259, 2008.

[7] J. Saxena, The Rotary Cement Kiln, ser. ISBN: 8188305952. Tech Books Interna-tional, 2009.

[8] J. Saxena, Refractory Engineering and Kiln Mainteneance in Cement Plants, ser.ISBN: 8188305006. Tech Books International, 2003.

[9] J. Stjernberg, “Degradation Mechanisms in Refractory Lining Materials of RotaryKilns for Iron Ore Pellet Production,” Ph.D. dissertation, Lulea University of Tech-nology, Sweden, 2012.

[10] Metso Corporation, Iron ore pelletizing. Grate-Kiln system. Metso Corporation,2012.

[11] V. Shubin, “Design and service conditions of the refractory lining for rotary kiln,”Refract. Ind. Ceram., vol. 42, pp. 130–136, 2001.

[12] R. Saidur, M. S. Hossain, M. R. Islam, H. Fayaz, and H. A. Mohammed, “A reviewon kiln system modeling,” Renewable and Sustainable Energy Reviews, vol. 15, pp.2487–2500, 2011.

31

32 References

[13] Livemore Software Technology Corporation, LS-DYNA Keyword User’s Manual,Version R7.0, ser. ISBN: 8188305952. Livemore, California, USA: Livemore Soft-ware Technology Corporation, 2013.

[14] S. Carniglia and G. Barna, Handbook of Industrial Refractories. Principles,Types,Properties and Applications, ser. ISBN: 0815513046, 1992.

[15] P. Bartha, “The cement rotary kiln and its refractory lining,” Refractories Manual,pp. 14–17, 2004.

[16] R. Sindut, “Silicaalumina refractory materials: The modern choice,” Glass Interna-tional, vol. 33, pp. 69–71, 2008.

[17] ASTM standard C27 - 98, Standard Classification of Fireclay and High-AluminaRefractory Brick. ASTM International, 1998, vol. 15.01.

[18] Deutsche Gesellschaft. Feauerfest- und Schornsteinbau e.V., Refractory Engineering.Materials - Design - Construction, ser. ISBN: 3-8027-3155-7. Vulan-Verlag Essen,2004.

[19] D. Ramanenka, “High-temperature compression strength of high-alumina refractorybricks used in rotary kilns of lkab,” Master’s thesis, Lulea University of Technology,Sweden, 2011.

[20] V. Shubin, “Mechanical effects on the lining of rotary cement kilns,” Refract. Ind.Ceram., vol. 42, pp. 245–250, 2001.

[21] V. Shubin, “The effect of temperature on the lining of rotary cement kilns,” Refract.Ind. Ceram., vol. 42, pp. 45–50, 2001.

[22] C. Schacht, Refractory Linings: Thermomechanical Design and Applications., ser.ISBN: 0-8247-9560-1. Marcel Dekker, 1995.

[23] J. Stjernberg, M.-L. Antti, L.-O. Nordin, and M. Oden, “Degradation of refractorybricks used as thermal insulation in rotary kilns for iron ore pellet production,” Int.J. Appl. Ceram. Technol., vol. 6, pp. 717–726, 2009.

[24] J. Stjernberg, B. Lindblom, J. Wikstrom, M.-L. Antti, and O. M, “Microstructuralcharacterization of alkali metal mediated high temperature reactions in mullite basedrefractories,” Ceramics International, vol. 36, pp. 733–740, 2010.

[25] J. Stjernberg, J. Ion, M.-L. Antti, L.-O. Nordin, B. Lindblom, and M. Oden, “Ex-tended studies of degradation mechanisms in refractory lining of a rotary kiln foriron ore production,” J. Eur. Ceram. Soc, 2012.

[26] O. Zienkiewicz and R. Taylor, Finite Element Method, ser. ISBN: 0-7506-5049-4.Butterworth-Heinemann, 2000, vol. 1.

References 33

[27] F. Damhof, W. Brekelmans, and M. Geers, “Predictive fem simulation of thermalshock damage in the refractory lining of steelmaking installations,” Journal of Ma-terials Processing Technology, vol. 211, pp. 2091–2105, 2011.

[28] T. Auer, D. Gruber, H. Harmuth, and A. Triessnig, “Numerical investigations ofmechanical behaviour of refractories,” in Proceddings of UNITECR 05, 2005, pp.985–989.

[29] H. Harmuth, C. Manhart, T. Auer, and D. Gruber, “Fracture mechanical character-isation of refractories and application for assessment and simulation of the thermalshock behaviour,” Ceramic Forum International, vol. 84, pp. 80–84, 2007.

[30] K. Andreev and H. Harmuth, “Fem simulation of the thermo-mechanical behaviourand failure of refractories- a case study,” Journal of Materials Processing Technology,vol. 143-144, pp. 72–77, 2003.

[31] D. Gruber, K. Andreev, and H. Harmuth, “Fem simulation of the thermomechanicalbehaviour of the refractory lining of a blast furnace,” Journal of Materials ProcessingTechnology, vol. 155-156, pp. 1539–1543, 2004.

[32] J. del Coz Dıaz, F. Rodrıguez Mazon, P. Garcıa Nieto, and F. Suarez Domınguez,“Design and finite element analysis of a wet cycle cement rotary kiln,” Finite Elementin Analysis and Design, vol. 39, pp. 17–42, 2002.

[33] E. Steinbiss, “Analysis of mechanical and thermal stresses in the loaded refractorylining of cement kilns,” ZKG, pp. 625–627, 1977.

[34] K. Andreev, S. Sinnema, A. Rekik, E. Blond, and A. Gasser, “Effects of dry jointson compressive behaviur of refractory linings,” Interceram, pp. 63–66, 2012.

34 References

Appendices

Appendix A

Maximum strain and stress induced in the lining related to oval-ity, lining thickness and its Young’s modulus.

Consider a circular shell resting on a ground that has been deformed to an ellipse dueto gravity force, according to Figure A1. The ellipse has major axis of length 2 · a and

a

b

δv

δh/2

Figure A1: Illustration of an ellipse with minor, b, and major, a, axis. Dashed line denotescircular shape prior deformation.

minor axis of length 2 · b, where a = r0 + δh/2, and b = r0 − δv/2 and r0 is the nominalradius before the deformation. The ellipse’s smallest radius of curvature R(a, 0) is at themajor axis and is found by:

R(a, 0) =b2

a=

(r0 − δv/2)2

(r0 + δh/2)= r0

(1− δv/2r0)2

(1 + δh/2r0)≈

≈ r0(1− δv/r0 + δ2v/4r20)(1− δh/2r0) =

= r0(1− δv/r0 +��δ2v/4r20 − δh/2r0 +��δhδv/2r

20 −��δhδ

2v/8r

30 ) ≈

≈ r0 − δv − δh/2 (A1)

The ellipse’s largest radius of curvature R(0, b) is at the minor axis and is deducedanalogically with the calculations made in Equation A1:

R(0, b) =a2

b= · · · ≈ r0 + δh + δv/2 (A2)

35

36 Appendices

By knowing the largest and the smallest radius of curvature the largest induced strain canbe found. Now consider that the ellipse has thickness tL corresponding to the thicknessof the lining. The deformation of circular shape to ellipse will generate positive andnegative deformation of all elements in the lining, i.e. compressive stresses on the innerwall and tensile stresses on the outer wall. Consider an element of width tL and hightx, as illustrated in Figure A2. The maximum deformation (dl) induced in the element is

- dl

+ dl

R(a,0)

tL/2 tL/2

x

R(0,b) r0

Figure A2: Illustration of largest R(0,b) and smallest R(a,0) radius of curvature relation to theinduced deformation (dl).

related to the difference between maximum and minimum radius of curvature:

dl =

(x

R(a, 0)− x

R(0, b)

)tL2

(A3)

Whereupon equation for the maximum strain (absolute value) can be deduced:

Δε =dl

x=

(1

R(a, 0)− 1

R(0, b)

)tL2

=(r0 + δh + δv/2)− (r0 − δv − δh/2)

(r0 + δh + δv/2) · (r0 − δv − δh/2)

tL2

=3

4

δh + δv(r20 − r0δv − r0δh/2) + (r0δh − δhδv − δ2h/2) + (δvr0/2− δ2v/2− δvδh/2)

tL

≈ 3

4

δh + δvr20

tL = 3δh + δvD2

0

tL (A4)

And the maximum stress is:

Δσ = σL = ELΔε = 3δh + δvD2

0

tLEL (A5)

Since the brick joints separate under tensile forces only the compressive stresses areof relevance in this case.

Appendices 37

Appendix B

Approximation of joint swallowing

Due to naturally uneven surface of the bricks the joints have certain capacity to absorbsome load before two surfaces have full contact. Figure B1 shows an example a non-destructive compression test of full-size bricks used in rotary kiln for iron-ore pelletproduction. A series of results can be found in Figure B1. The tests show that stress -

0.1 0.2 0.3

1

3

5

7

9

Displacement per brick joint (mm)

Str

ess

(MP

a)1 joint2 joints4 joints

Figure B1: left) Non-destructive compression test of five bricks. right) Example of results fromthe compression of two, three and five bricks.

displacement behaviour does not become linear until a certain point. The material itselfis brittle − the non-linear behaviour is therefore related to the compression of joints.How to define joint absorption is up to user. In this case the definition is made semi-qualitatively. 1 MPa is chosen as the stress limit below which brick joint absorption isdefined. With the support of the evaluation of the brick joint closure made by Andreevet.al. [34], the value of 0.1 mm was chosen as the brick joint absorption in this evaluation.

38 Appendices

Appendix C

Heat transfer relationship

Heat transfer through a layer of a material is calculated by:

W = λA

tΔT (C1)

Where λ is the heat transfer coefficient of the material, A is the area of heat transfer,t the thickness of the layer and ΔT the temperature difference between two surfaces.

Lining Casing

TH

TC TS

tL tS

Figure C1: Illustration of the steady statetemperature profile through the lining and thecasing.

In a steady state situation, as illustrated inFigure C1, the heat flow in the intersectionbetween the lining and the casing is equalfrom both sides, this gives:

W

A= λL

ΔTL

tL= λS

ΔTS

tS(C2)

Where ΔTL = TH − TC and ΔTS = TC −TS.

Solving equation C2 for TC gives:

TC = TSλL

tL+ TH

λC

tC(C3)

Appendix D

HOT FACE

COLD FACE

Figure D1: Illustration of brick expansion. Dashed line denotes brick after expansion.

Part II

1

Paper A

Modelling of Refractory BrickFurniture in Rotary-Kiln Using

Finite Element Approach

Authors:Dmitrij Ramanenka, Jesper Stjernberg, Kjell Eriksson and Par Jonsen

Reformatted version of paper originally published in:Proceedings of 11th World Congress on Computational Mechanics, July 2014

3

Modelling of refractory brick furniture in rotary-kiln using finite elementapproach

D. Ramanenkaa,∗, J. Stjernbergb, K. Erikssona, P. Jonsena

aDivision of Mechanics of Solid Materials, Lulea University of Technology, Lulea, SwedenbLoussavaara-Kiirunavaara Limited, Lulea, Sweden

Abstract

Refractory brick lining in rotary kilns is an essential part that governs availability of a kiln. A severelydamaged lining will eventually lead to an unplanned, long-lasting and costly production stop. A brick liningexperiences various degrading mechanisms during its service life. Knowing and avoiding critical situationsis of great importance to its life length. This work is focusing on a global mechanical behaviour of thebrick lining at room temperature. Finite element method by the commercial software LS-Dyna is used asa tool. Fundamental challenges for a brick lining are presented and studied. Approximation of true kilngeometry and a method of sequenced bricklaying of a large-diameter rotary kiln is presented. Stresses in thelining in static and dynamic loading are evaluated. Maximum effective von Mises stresses experienced bythe brick lining in static and dynamic loading were found to be approximately 1 and 6 MPa, respectively.Fast acceleration to working speed leads to extensive brick movement in the kiln. Careful ramping of therotational speed makes the brick lining to keep its integrity. A conclusion is that maximum stresses in thebrick lining is rather unaffected by the rotational acceleration, but the relative positions between individualbricks can be affected. These results confirm necessity of avoiding cold rotation. Additionally, a generalconclusion is that in terms of experienced stresses the brick lining is rather unaffected at room temperature.

Keywords: Finite Element Method, Rotary Kiln, Brick Lining

1. INTRODUCTION

A rotary kiln is a large cylinder-formed furnace used in certain hot-process manufacturing industries. Itis a slightly inclined, refractory lined steel container which rotates about its axis and where certain chemicaland physical reactions take place by the influence of heat. The slope and the rotation make the materialinside to move through the kiln from feed to discharge end. The heat is commonly generated by a flamein the discharge end from the combustion of coal, oil, natural gas or waste. The size of a rotary kiln canbe as large as 180 m in length and 7.5 m in diameter, while service temperature can be up to 1800 ◦C.The kiln is commonly resting on two to five pairs of support rollers, depending on its length. Additionally,it is equipped with thick, tightly fitted steel tyres that are riding on the support rollers. There are manyindustrial users of rotary kilns, however most of them are found in the field of cement, lime and mineralproduction.[1]

In order to be able to operate at high temperatures, the inner part of a rotary kiln consists of one orseveral layers of refractory materials. This is required for heat protection of the steel casing of the kiln,surroundings (such as sensible equipment and personnel) and reduction of heat losses (lower drift costs).These materials are usually in form of castables or bricks and have varied chemical composition dependenton service conditions. Refractories are essential for a wide range of hot processes exceeding 1000 ◦C and the

∗Corresponding authorEmail address: [email protected] (D. Ramanenka)URL: http://www.ltu.se/staff/d/dmiram-1.88050?l=en (D. Ramanenka)

Published in proceedings of WWCM XI July 2014

availability of a rotary kiln is highly dependent on the condition state of the refractory lining. Depletionof the refractory lining can lead to significant failures with fall outs of bricks or castables that may requireshut-down of the production. Unplanned shut-downs can cause very high production losses and put companyin a demanding situation. [1, 2]

Refractory products are often used in harsh service environment and therefore are prone to degradation.Tolerance to temperature, mechanical loads, thermal cycling, wear and chemical resistance are some of thecommon requests.[3, 4] As a matter of fact degradation of refractories in use is inevitable. Normal procedurein industry is that refractories in a kiln are controlled and/or replaced at a regular basis during plannedmaintenance shut-downs. The best case scenario is when refractory lining is degrading in a controlled mannerwithout causing shut-downs in-between maintenance stops.

This work is focused on the problems influencing stability and reliability of refractory lining in rotarykilns. Some general aspects and basic concept of rotary kilns will be discussed. In order to better understandthe different loads encountered by refractory lining, finite element analysis (FEA) is used. In this work, thecommercial FE-software LS-DYNA is used for FEA calculations. The parameters used here (geometries,materials) are typical for grate-kiln plants in iron-ore pellet production. Mechanical aspects in cold conditionare of main focus in this paper.

2. BACKGROUND AND DESCRIPTION OF THE PROBLEM

Producers of upgraded iron-ore products for the steel industry are common users of rotary kilns. Standardproducts such as iron-ore pellets undergo oxidization and sintering during heat treatment in order to be costeffective at future steps of the life chain, such as transport and steel production. [5] The heat treatment istypically performed in the grate-kiln systems. This process consists mainly of three parts: a pre-heater, ashort dry rotary-kiln and a cooling section. The pre-heater zone has moving grates transporting green pelletsto the kiln. The oxidation process starts on the grate-band, while sintering is completed in the rotary kiln.See Figure 1 for the illustration of the process. Common dimension of a rotary-kiln used for iron-ore pellet

Figure 1: Schematic illustration of a typical grate-kiln process in iron-ore pellet production.

production is 30-45 meters in length and 5.0-7.5 meters in diameter. The steel casing is usually lined witha single layer of refractory bricks and is resting on two pairs of support rollers. The thickness of the steelcasing is typically 50-100 mm depending on the diameter and the axial position (e.g. the casing is commonlythicker close to the tyres). Figure 2 illustrates a typical rotary kiln for iron-ore pellets production. The fillerpads, placed between the casing and the riding tyre, function as sacrificing abrasion material. Most of thegrate-kiln plants for iron-ore pelletizing in the world use traditional alumino-silicate based bricks, derivedfrom e.g. bauxite, andalusite, clay or chamotte. The base material controls the final alumina (Al2O3)and silica (SiO2) content in the brick, phase composition, and therefore many of its properties. [3, 8] Thelife-time of a refractory lining is influenced by mechanisms of different severity of thermal, mechanical andchemical character. If the bricks technical limits are violated during usage, degradation will be speed up.Therefore, it is very important that all the relevant steps of any process, either during production or amaintenance stop, are well adjusted to the technical limitations of the brick lining.

The state of the refractory lining is one of the most significant factors influencing the availability of arotary-kiln. Brick furniture of a rotary kiln is in direct or indirect symbiosis with the rest of the system. To

2

Figure 2: Illustration of a typical short dry-kiln used in iron-ore pellet industry (true proportions).

some important factors influencing brick lining can be included ovality of steel casing, burner conditions,fit of the tires and alignment of kiln. All of which can have negative influence on the life time of therefractories. Too large ovality of the steel casing can cause unhealthy load peaks in the brick lining whenshifting position during rotation. Misaligned burner or badly controlled power output of the burner can causecritical temperature peaks in parts of the lining leading to mismatched thermal expansion. Too tight ridingtyre can inhibit expansion of the brick lining leading to failure of the lining or even the tyre. Misalignmentof the kiln can cause unnecessary stresses to the rollers, tires and the brick lining. Severe damage of thelining is usually presented by the fall outs or dramatic thickness reduction of the bricks. This leads tothe formation of hot spots on the surface of the steel casing, risking a partial permanent deformation ofthe shell and making it less perfect. Bad perfection of the steel casing causes a less perfect brick lining,which additionally augments the risk of future fall outs. Thus, when hot spots are detected the productionis stopped for emergency maintenance. Due to the need of slow cooling and heating of the kiln, and therepairing the process is time-consuming (5-14 days). [2, 6]

3. METHOD AND MODEL APPROACH

A full scale model of a brick lined rotary kiln requires large computational power and effort. In thispaper a three dimensional model of a 100 mm thick section of the kiln made at the position of supportrollers will be brick lined and evaluated. The model dimensions according to the Figure 3.

3.1. Approximation of shell geometry

Emphasize that a rotary kiln is a massive construction. The total operational mass of a large rotary kilnused in iron-ore pellet production is at least 1 500 tons (including steel casing, refractory lining, pellets andslag). The effect of gravity on a hollow construction with such mass cannot be neglected. For simplicitya kiln shell is often regarded as circular with some given nominal diameter. However, a well-known andunderstandable fact is that cross-section of a rotary-kiln casing is not perfectly circular, but flattened dueto gravity force [7, 2]. The influence of gravity is schematically represented in Figure 4. Due to the ovalitythe steel casing and the refractory lining will undergo repeatable deformation during kiln rotation since theyare tightly fitted together. Knowing true geometry of steel casing is necessary in order to correctly reflectbehaviour of the refractory lining. In order to estimate this geometry FEA is used.

3

Figure 3: Dimensions (mm) of the unstrained kiln cross section used in the FE-model.

Figure 4: Arbitrary representation of unstrained (solid line) and strained (dashed line) kiln shell profile.

3.1.1. Validity of casing model

Consider a circular shell with some nominal inner diameter and wall thickness resting on a flat ground(see Figure 4). The vertical and horizontal displacements caused by gravity force can analytically be foundby:

δh = 0.42912 · ρgr4

Et2(1)

δv = 0.46712 · ρgr4

Et2(2)

Where δh and δv are horizontal and vertical displacements respectively, as can be seen in Figure 4. ρ isdensity, g is standard gravity, r is the nominal inner radius, E is Youngs modulus and t is the thickness ofthe casing (SI-units).

A finite element model was created in order to evaluate this basic case and validated towards analyticalsolutions. A comparison of deformation found by analytical and numerical solutions for different thickness

4

and diameters of steel casing is graphically represented in Figure 5. The numerical response of this case

20 40 60 80 1000

50

100

150

200

250

300Deformation vs Shell Thickness

Shell thickness, tc (mm)

Def

orm

atio

n (m

m)

a)

D0=5.5 m

δv analytical

δh analytical

δv FEM

δh FEM

3 4 5 6 7 80

50

100

150

200

250Deformation vs Shell Diameter

Inner shell diameter, D0 (m)

Def

orm

atio

n (m

m)

b)

tc=50 mm

δv analytical

δh analytical

δv FEM

δh FEM

Figure 5: Graphical representation of analytical versus numerical calculations of steel shell deformation when: a) shell thickness(tC) varies and b) inner shell diameter (D0) varies.

is very close to the analytical solution. This confirms validity of the numerical model. The assumption ismade that the casing model can be used for the more complex load cases.

3.2. Bricklaying

Refractory brick lining cannot be regarded as one homogeneous unit of the structure. The number ofbricks in circumference directly influences flexibility of the refractory lining and steel casing during rotationand thermal load. Therefore, every brick in a section has to be taken into account in a simulation.

The quality of the brick lining is directly influenced by the skill and quality control of the craftsmen.There is no definite way of quantifying the quality of a brick work. Two different occasions will generatetwo different brick linings. However, generally a good brick lining is fitted tightly to the shell, it is nottwisted, number of cut bricks is restricted and use of mortar and chims is limited. In this work the bricksare added to the model by following the edge of the shell and the neighbouring brick with a distance ofless than 0.1 mm. The used brick dimensions used in the model can be seen in Figure 6. The last brick

Figure 6: Brick dimensions (mm) used in the model.

in the circumference is slightly wider than the rest. This is according to a new trend in the bricklayingmethodology, assuming it is better, if needed, to cut a wide brick than a standard-sized brick. The modeldoes not include use of chims or mortar. Moreover, it does not consider compressibility of joints, which isprimarily important during thermal expansion and is therefore not treated in this work.[9, 10]

5

3.3. Assembling of FE-model

As previously shown the start geometry of an empty steel casing is oval. However, the ovality of an emptykiln casing will also gradually increase with addition of bricks. Consequently, a model of brick laying hasto include stress and geometry updates in order to correctly reflect the final position of the bricks. Ideallythe updates would be made after every new-coming brick. However, numerically this would be very timeconsuming. In the model presented in this paper the geometry and stresses were updated in four sequences.86 bricks were used in the circumference. See Figure 7 for the illustration of the bricklaying sequences ofthe chosen kiln cross-section. In order to simplify the model also a rigid riding tyre has been tested. With

Figure 7: Illustration of the bricklaying sequences. Gravity load applied, followed by sequenced brick laying with geometry andstress updates in-between.

this choice the support rollers can be removed and the thickness of the riding tyre reduced.After assembling the model was tested in static and dynamic loading. Rotation of the kiln was divided

into two cases, slow (30 s) and fast (7 s) ramping to 2 rpm. Rotation of a model with an elastic riding tyreand a model with a rigid riding tyre were also compared.

3.3.1. FEM and model specifications

All the parts in the model are build with fully integrated solid elements. The model’s depth correspondsto the thickness of the brick, 100 mm. Total number of elements for the model is 450 000 with elastic ridingtyre and 135 000 for the model with rigid riding tyre. Support rollers are modelled with rigid material. Thepads and the casing are modelled with elastic material, defined by typical low-alloy steel properties. Theriding tyre is either elastic or rigid. For the bricks a material with predefined concrete behaviour, is used.Density of the brick material was set to 2700 kg/m3, cold compression strength to 70 MPa and maximumaggregate size to 5 mm. The contact between parts is defined by mortar, penalty based segment-to-segment

6

contact. The dynamic and static friction coefficients of bricks-bricks, bricks-casing and casing-tyre contactsare set to 0.5. The friction coefficients in support roller-riding tyre contact is set to 1.5. The rotation ofthe kiln is onset by prescribed rigid body motion, either by the rotation of rigid support rollers or the rigidriding tyre. In the numerical calculations implicit integration was used.

4. RESULTS AND DISCUSSION

Figure 8 is a zoom in of the upper part of the model when using elastic riding tyre. After appliedgravity load a gap of approximately 35 mm is created at 12-o’clock-position between the filler pads and theriding tyre. Additionally a gap of around 25 mm is observed between the brick lining and the casing at the12-o’clock-position, which is also commonly observed in reality. Effective von Mises stresses of the static

Figure 8: Zoom in of the upper part of the model. A gap between the filler pads and the riding tyre is revealed

case in the brick lining are relatively small, reaching no more than 1 MPa, see Figure 9 a). That is far belowthe limitation of the bricks. In Figure 9 b) the effective von Mises stresses from the dynamic case with a fastramping are shown for a model with elastic riding tyre. The maximum effective von Mises stress increasesin average by a factor of six. Additionally, a substantial movement of the bricks is noticed in this case. Thisillustrates the importance of having a tight brick lining and avoiding rotation under cold conditions beforethe thermal expansion leads to proper self-locking. A slow ramping to 2 rpm does not change the stressdistribution substantially, however the integrity of the lining is kept better. The case with rigid riding tyreshow that the stresses are of the same order of magnitude as when using elastic riding tyre, see Figure 9 c).

For a case of a model with rigid riding tyre the solving time was approximately 20 hours at slow ramping,total simulation time of 60 seconds. The case was solved on 32 CPUs by massively parallel processing (MMP).When using shared memory parallel processing (SMP) the computational time increases by at least factortwo on the same conditions. In the case of a model with elastic riding tyre the computational time reachesabove 100 hours either by MPP or SMP on the same conditions. Explicit integration has been tested andwas found to require much longer solution times, > 200 h.

7

Figure 9: von Mises effective stress distribution in: a) Static case with elastic riding tyre b) Dynamic case with elastic ridingtyre, fast acceleration c) Dynamic case with a rigid riding tyre, slow acceleration, and theirs corresponding fringe levels. (MPa)

5. CONCLUSIONS

A rotary kiln is a fairly simple construction. Nevertheless, it is a complicated task to create a realisticmodel of it. It would be unrepresentative to scale down the task, thus the large size of the rotary kiln leadsto a demanding model in terms of computational power. Therefore it is important to create an effectivemodel. Additionally, the large number of bricks creates a heavy contact task.

We have created a model of a cross section of a rotary kiln at the position of rollers. We have approximatedthe ovality of the shell and proposed a bricklaying method. We have evaluated static and dynamic von Misesstresses and have seen the difference. It can be concluded that in terms of von Mises effective stresses thebrick lining does not experience critical situations at room temperature during static or dynamic loads.Nevertheless, in addition we can also conclude that a rotation in cold condition should be avoided due torisk of relative brick movement. Moreover, we have seen that we can assume the riding tyre as rigid, whichreduces the number of used elements drastically.

8

There are reasons to believe that thermo-mechanical aspects contributes greatly to the stresses experi-enced by the lining. It is highly important to evaluated this contribution and recognize critical situationswhere the loads are close or above the limits of the bricks. It is our intention to investigate that in futurework.

6. ACKNOWLEDGEMENTS

For financial support of the project Modelling and simulation of degradation mechanisms in thermaland mechanical loading of refractory lining in rotary kilns, LKAB and Lulea University of Technology aregratefully acknowledged.

References

[1] A.A. Boateng, Rotary Kilns. Transport Phenomena and Transport Process, Elsevier Inc., Oxford (2008) ISBN:9780750678773.

[2] J.P.Saxena, The Rotary Cement Kiln, Tech Books International, New Delhi (2009), ISBN: 8188305952.[3] S.C. Carniglia, G.L. Barna, Handbook of Industrial Refractories. Principles, Types,Properties and Applications, Park

Ridge (1992), ISBN: 0-8155-1304-6.[4] P.Bartha, The Cement Rotary Kiln and its Refractory Lining, Refractories Manual (2004); 14–17.[5] S.P.E. Forsmo , S.-E. Forsmo, P.-O. Samskog G.L., Mechanisms in Oxidation and Sintering of Magnetite Iron Ore Pellets,

Powder Technology (2008) 183: 247–259.[6] V.I.Shubin, Design and Service Conditions of the Refractory Lining for Rotary Kiln, Refractories and Industrial Ceramics

(2001) 42: 130–136[7] V.I.Shubin, Mechanical Effects on the Lining of Rotary Cement Kilns, Refractories and Industrial Ceramics (2001) 42:

245–250.[8] F.Cardarelli, Materials Handbook. A concise desktop reference, Springer Verlag, London (2008),ISBN: 978-1-84628-668-1.[9] K.Andreev, S.Sinnema, A.Rekik, S.Allaoui, E.Blond, A.Gasser, Compressive Behaviour of Dry Joints in Refractory Ce-

ramic Masonry, Construction and Building Materials 2012; 34: 402–408.[10] K.Andreev, S.Sinnema, A.Rekik, E.Blond, A.Gasser, Effects of Dry Joints on Compressive Behaviur of Refractory Linings,

Interceram (2012); 63–66.

9

Paper B

FEM investigation of globalmechanisms affecting brick liningstability in a rotary kiln in cold

state

Authors:Dmitrij Ramanenka, Jesper Stjernberg and Par Jonsen

Reformatted version of paper accepted for publication in:Engineering Failure Analysis, October 2015

15

FEM investigation of global mechanisms affecting brick lining stability in arotary kiln in cold state

D. Ramanenkaa,∗, J. Stjernbergb, P. Jonsena

aDivision of Mechanics of Solid Materials, Lulea University of Technology, Lulea, SwedenbLoussavaara-Kiirunavaara Limited, Lulea, Sweden

Abstract

Severe degradation of refractory lining in a rotary kiln often leads to very costly production delays. Use offinite element analysis for understanding the mechanisms behind the failure of the lining is poorly reportedin this field. To increase the knowledge and to update the field a simplified model of a kiln and a newmethodology for studying stability of the lining are suggested. Behaviour of the lining in cold state - instatic and dynamic cases - is studied. Influence of ovality, brick’s Young’s modulus and friction coefficient onstress and brick displacement are evaluated at two rotational speeds. It was found that the induced loads inthe lining are harmless regardless of the tested conditions - challenging the traditional beliefs. On the otherhand, recorded brick displacements were found to be significantly affected by rotational speed and ovality.Gaps as large as 80 mm could be observed between the bricks and the casing in a worst case scenario. Anintegrity coefficient was defined in order to quantify overall integrity of the lining.

Keywords: Rotary kiln, Brick lining, Lining failure, Finite element method

1. Introduction

Rotary kilns are important in a variety of different manufacturing areas for e.g. calcination and sinteringof materials. In fact, two of the most produced materials in the world [1, 2], cement and iron, are likelyto start their journey in a rotary kiln. The kiln simply consists of a cylindrical steel casing lined withrefractories. As the kiln is heated and rotates during production, it is subjected to a complex stress/straincondition. Availability of it is highly dependent on the state of the refractory lining. If the lining issignificantly deteriorated and can no longer protect the casing from the heat − the production is shut-down− leading to high production losses.

There has been many improvements of rotary kilns in the last century. Production capacity of a singlekiln has been increased from earliest 300 metric ton/day up to 20, 000 metric ton/day today. The lifetimeof the earliest linings in hot zones did not exceed more than 10-15 days, today 200-300 days is expected.Energy efficiency has been improved by 50-75 % compared to the early wet kilns [3]. However, despitethese improvements there is still little knowledge behind the mechanisms of the failure of the lining. Anexplanation to this is the difficulty to observe or study a kiln due to its size and harsh environment. Anotheris the slow adaptivity of the users who are, understandably, not willing to take the risk of practising newideas that might become expensive. Today, a part of the problem can be moved to the computer clusterswhere behaviour of the lining can be simulated and studied. In this matter there is a contribution to bemade.

Little attention is paid to the field of refractory lining and its failure, especially that of a non-chemicalcharacter [4, 5, 6, 7]. Schubin [8, 9, 10, 11] gives comprehensive and valuable series of scientific papers on the

∗Corresponding authorEmail address: [email protected] (D. Ramanenka)URL: http://www.ltu.se/staff/d/dmiram-1.88050?l=en (D. Ramanenka)

Accepted for publication in Engineering Failure Analysis October 30, 2015

subject of cement kilns. These papers discuss the development of refractory materials and service conditionsinfluencing the lining. The effect of temperature oscillations, the slag and the need of expansion joints arediscussed. Analytical solution for the temperature profile in the lining at different conditions is proposed.The mechanical stresses in the lining caused by the ovality of the steel casing are analysed analytically andsome data is presented. Other mechanical aspects affecting the lining are discussed. Saxena [3] gives a goodoverview on the problems encountered in the field of cement rotary kilns. In [12] authors perform a rigorouswork about refractory engineering covering description of materials, design theory, practical principles andmore. Como [13] presents a new comprehensive, unified theory of statics of masonry constructions, wheresimilarities with refractory linings can be find. Use of Finite element method (FEM) for the simulation ofrotary kilns is poorly reported in scientific journals. Some information can be gathered from the researchesworking in the field of steel-making industry, where refractory materials are extensively used. Damhof etal. [14] present a FE-model of thermal shock damage in the refractory lining of steel-making installations.Some authors [15, 16, 17, 18] have presented a number of articles on the subject of FE-simulation of thermo-mechanical behaviour of refractory linings. Andreev et al. [19, 20] discuss behaviour and importance of dryjoints in refractory linings of BOF (basic oxygen furnace) converters. Del Coz Diaz et al. [21] make a FiniteElement Analysis (FEA) of a cement rotary kiln evaluating ovalization and stresses in the steel casing. Tothe authors’ knowledge there is currently little attention from academic research on FE-simulation of rotarykilns especially that includes brick lining and evaluates its mechanical response in static or dynamic cases,in cold or hot state.

Producers of iron-ore pellets for iron making are common users of rotary kilns. The size vary between30 and 45 m in length and 5-8 m in diameter. The service temperature in the hottest zone is locallyapproximately 1300 ◦C. The kiln is resting on two pairs of support rollers. It is equipped with massive steeltyres for stiffening purpose that are riding on the support rollers. Between tyres and the main body (thecasing) filler pads are placed as sacrificing abrasion material. The thickness of the steel casing is typically50-100 mm depending on the diameter and the axial position (e.g. the casing is thicker close to the tyres).The inner part of a rotary kiln is typically lined with a single layer of refractory bricks. This is requiredfor heat protection of the steel casing and surroundings, reduction of heat losses and maintaining of desiredtemperature. Figure 1 illustrates a typical rotary kiln for iron-ore pellet production.

Figure 1: Illustration of a typical short dry-kiln used in iron-ore pellet industry (true proportions).

In this work some general aspects and basic concept of rotary kilns are discussed. Contribution of thispaper is to, by means of FEA, study the effect of ovality, rotational speed, Young’s modulus of the bricks andfriction of the bricks on the load state and behaviour of a single layer brick lining in cold condition in staticand dynamic cases. For this, a new methodology for studying the brick lining is proposed. The commercialFE-software LS-DYNA [22] is used for FE-calculations. The used parameters (geometries, material dataetc.) are typical for rotary kilns in iron-ore pellets production.

2

2. Theory

Brick lining of a rotary kiln is directly or indirectly in symbiosis with the rest of the system. To someimportant factors influencing brick lining’s life can be included ovality of steel casing, burner conditions, fitof the tyres and alignment of the kiln. Too high ovality of the steel casing can cause unhealthy load peaksin the brick lining during rotation. Misaligned burner or badly controlled power output of the burner cancause critical temperature peaks in parts of the lining leading to mismatched thermal expansion. Too tightriding tyre can inhibit thermal expansion of the brick lining leading to failure of the lining or even the tyre.Misalignment of the kiln can cause unnecessary stresses to the rollers, tyres and the brick lining [3, 8].

Severe damage of the lining is usually presented by the fall outs or essential thickness reduction ofthe bricks. This leads to the formation of ”hot spots” on the steel casing, indicating local temperatureincrease due to worsen heat insulation. The ”hot spots” risk permanently damage the casing and makingit less perfect (e.g. indents). Bad perfection of the steel casing worsen integrity of the brick lining, whichadditionally augments the risk of future lining failure. The damaged areas risk also crack formation followedby corrosion and other problems decreasing life of the casing. Thus, when ”hot spots” are detected theproduction is commonly stopped for emergency maintenance. Due to the need of slow cooling and heatingof the kiln, and repairing, the kiln can be out of service for 5-14 days [3, 8].

2.1. Mechanical effects on the lining

Lining of the kiln is tightly fitted to the steel casing. During usage of the kiln (starts, stops, rotation)the steel casing and therefore the lining, are subjected to radial and longitudinal bending, vibrations andtorsion. This results into varied stress-controlled loads in the lining. [10].

2.1.1. Ovality

Radial bending of the kiln, known as the ovality of the kiln, is traditionally regarded as one of themost important load generators in the refractory lining and directly affecting its life. Ovality is an elasticdistortion of the kiln casing that arise due to the gravity force. The weight of the casing, the lining, thecharge and possible slag inside the kiln make the cross-section of the casing somewhat oval rather thancircular. This is illustrated in Figure 2. The vertical and the horizontal deviation from the circular line (δv

g⇓

Figure 2: Arbitrary representation of unstrained (solid line) and strained (dashed line) kiln shell profile.

and δh ) may typically reach 10-15 mm in standard-sized kilns for iron-ore pellet production. Due to theovality of the steel casing the lining will experience load oscillations during rotation of the kiln. In a period of24 h a kiln for iron-ore pellet production passes typically through 2-4 thousands revolutions and the doubleamount of load oscillations (due to the vertical symmetry line). This may lead to the formation of cracksand eventually to the spalling of the refractory lining. Additionally, the bricks in the lining are forced toshift their relative position to each other during rotation due to the ovality. This results into opening of thebrick joints, potentially leading to worsen integrity of the brick lining and eventually to brick fall outs [10].

Relative ovality is typically presented in percent, as relative deformation to the nominal diameter. If thedeformation is known than relative ovality, ωr, is found by Equation 1 [3].

3

ωr =δv + δhD0

· 100% (1)

Where D0 is the nominal inner diameter of the casing. A rule of thumb based on the experience has beenestablished suggesting that the ovality of the steel casing should not exceed 10 % of the nominal innerdiameter. This implies that a kiln with a nominal inner diameter of 7 m should preferably not have relativeovality of more than 0.7 % [3, 10].

In the case presented in Figure 2 the magnitude of largest compressive stresses in the lining induced bythe ovality of the kiln can be approximated analytically by Equation 2.

σL = 3δh + δvD2

0

ELtL (2)

Where σL is the largest compressive stress induced in the inner wall of the lining, EL is the Young’s modulusof the lining material and tL is the thickness of the lining. From the equation can be seen that stressesincrease with increased ovality, Young’s modulus and thickness of the lining. Figure 3 graphically representsrelationship between stress, ovality, diameter and thickness of the lining with typical dimensions for large-sized rotary kilns used in iron-ore pellet production. Young’s modulus is set to 10 GPa as the upper boundbased on our in-house data.

0.2 0.4 0.6 0.8 1 1.20

2

4

6

8

10

12

14

16

Relative ovality, ωr (%)

Com

pres

sive

str

ess,

σL (

MP

a)

6 (200)

7 (200)

6 (250)

7 (250)

5 5.5 6 6.5 71

2

3

4

5

6

7

8

Inner diameter, D0 (m)

0.3 (200)

0.3 (250)

0.5 (200)

0.5 (250)

Figure 3: Compressive stress dependency in the lining with respect to ovality, initial inner diameter and thickness of the lining.Notations inside figure: 6 and 7 - initial inner diameter (m), 200 and 250 - lining thickness (mm), 0.3 and 0.5 - relative ovality(%).

The magnitude of ovality is primely dependent on the thickness of the steel casing, thickness of thepads, the gap between the tyre and the pads and the operating temperature. The ovality is highest nearthe tyres and statistically most of the repair jobs of the lining are done close to the downhill-tyre (closestto the flame). Additionally, the ovality of kiln is not at a permanent state but is changing with operatingconditions. When the lining is newly installed the ovality tends to be at it’s lowest point and increases aftersome time. Wear of filler pads gradually increases ovality. If the lining is covered with protective slag theeffect of temperature is lowered and therefore the ovality is lowered as well [3, 23].

3. Method and model approach

FEM is a well-known and widely used numerical solution method based on continuum mechanics mod-elling [24]. Typically, can stresses and strains of the system be evaluated and some conclusions regarding

4

e.g. the design can be made.A full scale FE-model of a brick-lined rotary kiln requires large computational resources. For this reason

only a cross section of the rotary kiln will be studied in this work. A model of the kiln at the position ofsupport rollers is created. It is lined with a single brick layer and is three dimensional. The depth of themodel is 100 mm, corresponding to the thickness of one brick. The model dimensions are shown in Figure4.


3.1. Analytical validation of casing model

Consider a circular shell with some nominal inner diameter and wall thickness resting on a flat groundas shown in Figure 2. When solving this statically indeterminate problem, the vertical and horizontaldisplacements caused by gravity force can analytically be found by:

δh = 0.42912 · ρgr4Et2C

(3)

δv = 0.46712 · ρgr4Et2C

(4)

Where δh and δv are horizontal and vertical displacements respectively. Here, ρ is density, g is standardgravity, r is the nominal inner radius, E and tC is Young’s modulus and the thickness of the casing,respectively (SI-units).

A FE-model was created in order to evaluate this case. The case is used as a simple validation towardsanalytical solution, equation (3) and (4). A comparison of deformation found by analytical and numericalsolutions for different thickness and diameters of steel casing is graphically represented in Figure 5. Thenumerical response of this case corresponds well with the analytical solution. This confirms usability of thenumerical model. The assumption is made that the model can be used for the more complex load cases,such as the case represented in Figure 4.

The mesh dependency was tested and the mesh size was decided to be approximately 25 mm for thewhole model presented in this work.

3.2. Bricklaying and assembling of the FE-model

Refractory brick lining should not be regarded as one homogeneous unit of the structure. The numberof bricks (joints) in the circumference directly influences flexibility of the refractory lining and steel casingduring rotation and thermal expansion. Therefore, it is advantageous to account every brick in a sectioninto the model.

5

20 30 40 50 60 70 80 90 1000

50

100

150

200

250

300Deformation vs Shell Thickness

Shell thickness, tc (mm)

Def

orm

atio

n (m

m)

a)

D0=5.5 m

δv analytical

δh analytical

δv FEM

δh FEM

3 4 5 6 7 80

50

100

150

200

250Deformation vs Shell Diameter

Inner shell diameter, D0 (m)

Def

orm

atio

n (m

m)

b)

tc=50 mm

δv analytical

δh analytical

δv FEM

δh FEM

Figure 5: Graphical representation of analytical versus numerical calculations of steel shell deformation when: a) shell thickness(tC) varies and b) inner shell diameter (D0) varies.

The quality of the brick lining is directly influenced by the skill and quality control of the craftsmen.There is no definite way of quantifying the quality of a brick work. Two different occasions will generatetwo different brick linings. In general, a good brick lining is fitted tightly to the shell, it is not twisted andnumber of cut bricks is restricted. Mortar and chims may or may not be used. In this work the bricks areadded to the model by following the edge of the shell and the neighbouring brick with a distance of lessthan 0.05 mm. The brick shape used in the model can be seen in Figure 6. The model does not include use

Figure 6: Brick shape (mm) used in the model. Thickness: 100 mm

of chims or mortar. Moreover, it does not consider compressibility of joints, which is primarily importantduring thermal expansion and is therefore not treated in this work [19, 20].As known the initial cross section of an empty steel casing is oval. Moreover, the ovality of an empty kilncasing will gradually increase with addition of bricks. Consequently, it is advantageous to include stress andgeometry updates in order to properly reflect position of the bricks and ovality of the kiln. Ideally the updateswould be made after every new-coming brick. In practise this would be very time consuming. In the modelpresented in this work the geometry and stresses were updated in four sequences. 86 bricks were used in thecircumference and an additional cut brick. See Figure 7 for the illustration of the bricklaying sequences ofthe chosen kiln cross-section. The model was constrained in z-translational and x- and y-rotational degreesof freedom.

3.3. Description of case studies

After assembling, the model was evaluated and tested in three studies, all in cold state.Study 1. A comparison between two material models in the definition of bricks is made - a purely elasticmaterial model and a predefined concrete material model. Also, a comparison between an elastic and a

6

a) b)

c) d)

gg

gg

. . . . .

⇓⇓

⇓⇓

�

�

�� x

y

z

Figure 7: Illustration of the bricklaying sequences. a) Gravity load applied to the shell, followed by sequenced brick laying (b),c) and d)) with geometry and stress updates in-between.

rigid riding tyre is made. The purpose is to proof that the model can be simplified by using purely elasticmaterial model for the bricks and a rigid riding tyre. A model using 35 mm thick filler pads is used for thispurpose. The evaluation is based on the results from static and dynamic (1 revolution at 0.2 rpm) loading.Study 2. Models with 35 mm, 25 mm and 40 mm thick filler pads are compared in static and dynamicloading (5 revolutions). Bricks are defined by elastic material model, the riding tyre is rigid. The rotationis made at two speeds: 0.2 rpm - a typical maintenance speed and 2 rpm - a typical production rate.Smoothly accelerated in both cases. The purpose of this evaluation is to simulate ovality variation and asevere thickness reduction of the pads due to abrasion but also to provoke brick movement in the model dueto increased distortion of the casing. Note: Rotation at 2 rpm in cold condition is not, generally, applied toa typical rotary kiln, however, the idea is to provoke a worse case scenario.Study 3. In this part Young’s modulus of the bricks and friction between the bricks is varied for the modelusing 35 mm thick filler pads. The cases are evaluated on the results from dynamic loading (1 revolutionat 0.2 and 2 rpm). The purpose of this evaluation is to study expected variation of results due to largeheterogeneity of the material properties. Furthermore, in the case of the variation of the friction coefficients,the authors cannot confirm that the extremes are fully realistic. However, the purpose is to validate theimportance of the factors prior eventual experimental testing.

7

The static loading was evaluated in terms of: vertical and horizontal distortion of the casing (δv and δh);relative ovality of the casing (ωr); the gap between the pads and the tyre at the twelve-o’clock-position (ΔR,see Figure 8); and maximum effective von Mises stress experienced by the brick lining (σeff

max). The dynamicloading was evaluated in terms of: effective von-Mises stress during rotation; by continuously tracing the gapbetween the bricks and the shell (Δi); and termination time (TT ) for Study 1. An integrity coefficient,K,is introduced in order to quantify the overall status of the brick lining based on the brick displacement.

3.4. Numerical model specifications

All parts in the model were build with selective reduced fully integrated solid elements - mesh sizeapproximately 25 mm. The model’s depth corresponds to the thickness of the brick, 100 mm. Total numberof elements in the model with elastic riding tyre is approximately 151,000 and 57,000 in the model withrigid riding tyre. Support rollers are modelled as rigid bodies. The pads and the casing are modelledwith purely elastic material, defined by typical low-alloy steel properties. The pads are attached to thecasing using nodal rigid bodies. The riding tyre is either elastic or rigid. The bricks use either elastic orpredefined concrete material model, depending on the study. Bulk density of the brick material was set to2700 kg/m3 [25], Young’s modulus to 2.5, 5 or 10 GPa (based on the in-house measurements [26]) for theelastic bricks and Poisson’s ratio to 0.15 [27]. Additionally, cold compression strength was set to 70 MPa [25]and maximum aggregate size to 5 mm [4] in the predefined concrete material model. The contact betweenparts was defined by mortar, penalty based segment-to-segment contact. The dynamic and static frictioncoefficients of brick-to-brick contact are both set to the same value of either 0.4, 0.75 [28] or 1.2 dependingon the study, and for brick-to-casing and casing-to-tyre contacts to 0.5. The friction coefficients in supportroller-riding tyre contact is set to 0.7. The rotation of the kiln is onset by prescribed rigid body motion,either by the rotation of rigid support rollers or the rigid riding tyre. Implicit integration is used for thenumerical calculation. All the studied cases presented in Section 4 were solved on 16 CPUs using massivelyparallel processing (MPP).

4. Results and Discussion

An example of a zoom-in of the upper part of a model after static loading can be seen in Figure 8. Allmodels have, after static loading, a gap between bricks and the casing (shell), the largest gap is at 12 o’clockposition which reduces radially until full contact at approximately 10 and 2 o’clock positions. The reasonbehind this choice is to provoke a worse case scenario - a loose brick lining. The gap between the bricks andthe casing (Δi is continuously measured by tracing one selected node on each brick to the closest point onthe casing, illustrated in Figure 8. The bricks are identified by counting counter-clockwise starting with thebrick placed at the six o’clock position in the initial stage.

ΔRΔi

�

�

�� x

y

z

Figure 8: Zoom-in of the upper part of a model after applied gravity load. A gap between the filler pads and the riding tyre(ΔR), and the bricks and the casing (Δi) is revealed. The black dots denotes traced nodes of each brick. Numbering denotesbricks’ identification number (ID).

8

4.1. Study 1. A predefined concrete material model vs. purely elastic material model. Elastic riding tyre vs.rigid riding tyre.

4.1.1. Static loading

Table 1 summarizes the results of static loading of the four studied cases. Young’s modulus is set to 10GPa and friction between the bricks to 0.75. The results indicate that the distortion (δh, δv, ωr,ΔR) of thecasing is affected by the choice of the riding tyre but not the choice of the material model. The distortionis < 10% lower when using rigid riding tyre, case 1 and 3. Furthermore, it is found that maximum effectivevon Mises stress in the brick lining is affected by the choice of the material model but not the choice of theriding tyre. The maximum effective von-Mises stress is 40 % higher when using predefined concrete model,case 3 and 4. The reason to this is a, per default, higher Young’s modulus used in the predefined concretematerial model. Furthermore, the ωr and ΔR, which are related to the δv and δh, vary inconsiderably.

Table 1: Results from static and dynamic (one revolution) loading of the model using 35 mm thick filler pads - brick materialmodel and riding tyre type varied.

Static Dynamic

δv δh ωr ΔR σeffmax TT

Brick model Tyre Case (mm) (mm) % (mm) MPa (h)

elasticrigid 1 13.3 14.5 0.4 36.7 0.5 17elastic 2 14.1 15.4 0.43 35.4 0.5 29.4

concreterigid 3 13.2 14.4 0.4 36.8 0.7 20.5elastic 4 14 15.3 0.43 35.5 0.7 56.7

4.1.2. Dynamic loading

Figure 9 shows results from dynamic loading in terms of bricks’ distance to the shell during one revolution.It is noticed that the choice of the riding tyre is not affecting bricks’ distance to the shell. However, thegap is affected by the choice of the material model of the bricks. The choice of predefined concrete materialmodel gives a stiffer response, case 3 and 4. The distance to the shell is approximately 30 % lower whenusing concrete material model. This is in correlation with the higher effective stresses found previously.

The CPU-time of the calculations (TT ) was affected by both the choice of the material model and thechoice of the riding tyre, see Table 1. Case 1 had 3.3 times shorter computational time (1 revolution at 0.2rpm) compared with case 4.

Further studied cases will use elastic material model in the definition of the bricks and a rigid riding tyre- case 1. The motivation behind this, after comparing the four different cases, is that: there is no reason tochoose a more advanced concrete material model; the difference in computational time is considerable whilethe stresses in the brick lining are not significant and the distortion of the casing is comparable. Figure10 represents comparison of fringe levels of effective von-Mises stress for cases 1 and 4 - the simplified andthe assumed most realistic case among the studied cases. It is noticed that the distribution of the stress issimilar. Moreover, maximum effective stress in the casing reaches approximately 24 MPa in both cases.

4.2. Study 2. Static and dynamic loading of models using 35 mm, 25 mm and 40 mm thick filler pads

In this study the Young’s modulus is set to 10 GPa and friction between the bricks to 0.75. The purposeof this evaluation is to study load state of the lining and to provoke position shifting of the bricks whenvarying ovality.

4.2.1. Static loading

Table 2 summarizes the results of static loading of the three cases. It is noticed that the distortion(δh, δv, ωr,ΔR) is affected rather significantly by the decreasing filler pad thickness. The relative ovalityis increased by approximately 50 % when decreasing pad thickness from 40 mm to 35 mm and 35 mm to25 mm respectively. The maximum effective stress in the brick lining was found to be around 0.5 MPa

9

0 20 40 60 800

5

10

15

20

25

Elastic brick − Rigid tyre (case 1)

Brick ID

Bric

k−to

−sh

ell g

ap (

mm

)

MaxMean ± standard deviation

a)

0 20 40 60 800

5

10

15

20

25

Elastic brick − Elastic tyre (case 2)

Brick ID

Bric

k−to

−sh

ell g

ap (

mm

)


b)

0 20 40 60 800

5

10

15

20

25

Concrete brick − Rigid tyre (case 3)

Brick ID

Bric

k−to

−sh

ell g

ap (

mm

)


c)

0 20 40 60 800

5

10

15

20

25

Concrete brick − Elastic tyre (case 4)

Brick ID

Bric

k−to

−sh

ell g

ap (

mm

)


d)

Figure 9: Brick-to-shell distance recorded during dynamic loading (one revolution at 0.2 rpm) of the model using 35 mm thickpads. Brick material model and riding tyre type varied.

and is insignificantly affected by the increased distortion of the casing. This is in coherence with work ofSteinbiss [29] who experimentally measured compressive stress in a lining to 0.4 MPa in a cement kiln incold conditions. It is noticed that the analytical calculations of the maximum stress in the brick lining,found by Equation 2 and graphically presented in Figure 3, are one degree of order larger and probablyoverestimate the maximum stress. This is partly because the analytical solution does not take into accountthe effect of brick joints that lower the stiffness of the lining. Additionally, in-house measurements by ashell-tester confirmed that the relative ovality (ωr) for the model using 35 mm thick filler pads found in thisstudy corresponds very well with the measurement performed on the real kiln of the same dimension.

Table 2: Results from static loading of the models using 35, 25 and 40 mm thick filler pads.

Pad thickness (mm) δv (mm) δh (mm) ωr (%) ΔR (mm) σeffmax (MPa)

40 8.9 9.7 0.27 22 0.4535 13.3 14.5 0.4 36.7 0.525 19.8 21.1 0.6 63.8 0.5

4.2.2. Dynamic loading

Brick displacement. Figure 12 shows evolvement of the gap between the bricks and the casing from dynamicloading at two different rotational speeds (2 and 0.2 rpm) during 5 revolutions. From the graphs can be seenthat at the lower rotational speed the gap is stable and comparable for the three cases. The varied ovalitydoes not have a significant effect at this speed. The bricks initially positioned at 12 o’clock (ID 40-45) tendto have slightly larger gap to the shell through the five revolutions. The maximum gap is around 20 mmlarge. At speed of 2 rpm the gap is irregular. The maximum gap values are increased with higher speed andthinner pads. Gap up to 80 mm is noticed for the case using 25 mm thick filler pads, while for the pads using

10

a) b)

c) d)

Figure 10: Fringe levels and distribution of effective von Mises stress in static case of: a) case 4. b) case 1. c) case 4 - bricklining only. d) case 1 - brick lining only.

35 and 40 mm thick pads the gap is approximately 40 mm at the most. An example of brick shifting seenin the models is presented in Figure 11. This illustrates the importance of having a tight brick lining andavoiding careless rotation under cold conditions before the thermal expansion leads to proper self-locking ofthe brick lining.At this stage we introduce an integrity coefficient, K, in order to quantify the general status of the bricklining based on brick-to-shell gaps:

K =1

3

[(RΔ

mean)∧(Rs

mean,i) + (RΔmean)

∧(Rs

mean) + (RΔmax)

∧(Rsmax)

](5)

11

Where RΔ coefficients describe the size of the gaps in relation to the thickness of the lining (t). Rs

coefficients describe the integrity of the brick lining based on the standard deviation of the results. K = 1means no gap is existing, K = 2 theoretically means that every brick has reached critical gap, here definedas t/2, and that average and maximum size of the gaps coincide. In practice values of 1.2 already indicatethat the lining is not stable. Notice that the integrity coefficient does not capture behaviour of an individualbrick but only the overall integrity of the lining. Definition of the coefficients follows below:

RΔmean is a relation between the average gap for all the bricks (Δmean) and half of the thickness off the

lining (t/2):

RΔmean = 1 +

2

t · n∑

Δmean,i = 1 +2

tΔmean (6)

n is number of the bricks and Δmean,i is the mean gap of an individual brick.RΔ

max is a relation between the average of maximum gap of all the bricks (Δmax) and half of the thicknessoff the lining:

RΔmax = 1 +

2

t · n∑

Δmax,i = 1 +2

tΔmax (7)

Rsmean,i is a relation between one standard deviation of the gap of every individual brick (smean,i) and

the brick’s individual mean gap (Δmean,i):

Rsmean,i =

1

n

∑ Δmean,i + smean,i

Δmean,i(8)

Rsmean is a relation between one standard deviation of average gap for all bricks (smean) and the average

gap for all the bricks (Δmean):

Rsmean =

∑ Δmean + smean

Δmean(9)

Rsmax is a relation between one standard deviation of average maximum gap for all bricks (smax) and

the average maximum gap for all the bricks (Δmax):

Rsmax =

∑ Δmax + smax

Δmax(10)

In Figure 12 can be found the calculated integrity coefficients of the presented cases. It explicitly revealsthe relative differences between the cases. In overall, cases a-e) are similar and could be interpreted asstable, while case f) as unstable. The reader should evaluate the results qualitatively since no criterion forunstable brick lining is defined.

Stress state. Figure 13 shows measured effective von-Mises stress in one element of each brick during 5revolutions. The element is located in the center of the brick at the face closest to the casing. The resultsare plotted in frequency of one plot per second for 0.2 rpm-cases and one plot every 10th second for 2rpm-cases. At 0.2 rpm the stresses in the brick lining are stable and comparable to the stress level in staticsituation (maximum stresses approximately 0.5 MPa) - subfigure a), c) and e). Rotation at 2 rpm results insomewhat increased stresses in the brick lining. In overall the stresses are low in all the models - subfigureb), d) and f). The maximum von-Mises stresses where chosen to be neglected since some contact problemswhere noticed at high rotational speed resulting in quick and high load peaks. It is concluded that stressesare of low interest to study and will therefore not be presented in the rest of the work.

4.3. Study 3. Influence of Young’s modulus and friction coefficient.

Figure 14 shows brick displacements while varying Young’s modulus (2.5, 5 and 10 GPa). It is noticedthat displacement of the bricks increases with decreased Young’s modulus. The same trend is visible atlower and higher rotational speed.

Figure 15 shows brick displacements while varying friction coefficient (0.4, 0.75 and 1.2) between thebricks. It can be concluded that variation of friction has insignificant influence on the displacement of thebricks, independent of the rotational speed.

12

Figure 11: Freeze image representing example of bricks’ position after 4 revolutions. Rotation speed 2 rpm, filler pad thickness35 mm.

0 20 40 60 800

5

10

15

20

25

40mm 0.2rpm E=10GPa μ=0.75

Brick ID

Gap

bric

k−sh

ell (

mm

)

K=1.09


a)

0 20 40 60 800

5

10

15

20

25

35mm 0.2rpm E=10GPa μ=0.75

Brick ID

Gap

bric

k−sh

ell (

mm

)

K=1.10


b)

0 20 40 60 800

5

10

15

20

25

25mm 0.2rpm E=10GPa μ=0.75

Brick ID

Gap

bric

k−sh

ell (

mm

)

K=1.11


c)

0 20 40 60 800

10

20

30

40

50

60

70

80

9040mm 2rpm E=10GPa μ=0.75

Brick ID

Gap

bric

k−sh

ell (

mm

)

K=1.12


d)

0 20 40 60 800

10

20

30

40

50

60

70

80

90 35mm 2rpm E=10GPa μ=0.75

Brick ID

Gap

bric

k−sh

ell (

mm

)

K=1.12


e)

0 20 40 60 800

10

20

30

40

50

60

70

80

9025mm 2rpm E=10GPa μ=0.75

Brick ID

Gap

bric

k−sh

ell (

mm

)

K=1.22


f)

Figure 12: Brick-to-shell distance recorded during dynamic loading (5 revolutions at 2 and 0.2 rpm) of the models using 40, 35and 25 mm thick filler pads, case 1.

13

0 20 40 60 800

0.1

0.2

0.3

0.4

0.540 mm 0.2 rpm E=10GPa μ=0.75

Brick ID

Effe

ctiv

e vo

n M

ises

str

ess

(MP

a)

Mean ± standard deviation a)

0 20 40 60 800

0.1

0.2

0.3

0.4

0.535 mm 0.2 rpm E=10GPa μ=0.75

Brick ID

Effe

ctiv

e vo

n M

ises

str

ess

(MP

a)

Mean ± standard deviation b)

0 20 40 60 800

0.1

0.2

0.3

0.4

0.525 mm 0.2 rpm E=10GPa μ=0.75

Brick ID

Effe

ctiv

e vo

n M

ises

str

ess

(MP

a)

Mean ± standard deviation c)

0 20 40 60 800

0.5

1

1.5

240mm 2rpm E=10GPa μ=0.75

Brick ID

Effe

ctiv

e vo

n M

ises

str

ess

(MP

a)

Mean ± standard deviation d)

0 20 40 60 800

0.5

1

1.5

2 35mm 2rpm E=10GPa μ=0.75

Brick ID

Effe

ctiv

e vo

n M

ises

str

ess

(MP

a)

Mean ± standard deviation e)

0 20 40 60 800

0.5

1

1.5


Brick ID

Effe

ctiv

e vo

n M

ises

str

ess

(MP

a)

Mean ± standard deviation f)

Figure 13: Effective von-Mises stress recorded during dynamic loading (5 revolutions at 2 and 0.2 rpm) of the models using 40,35 and 25 mm thick filler pads, case 1.

0 20 40 60 800

5

10

15

20

25

30

35

40

4535mm 0.2rpm E=10GPa μ=0.75

Brick ID

Bric

k−to

−sh

ell g

ap(m

m)

K=1.10


a)

0 20 40 60 800

5

10

15

20

25

30

35

40

45 35mm 0.2rpm E=5GPa μ=0.75

Brick ID

Bric

k−to

−sh

ell g

ap(m

m)

K=1.15


b)

0 20 40 60 800

5

10

15

20

25

30

35

40

4535mm 0.2rpm E=2.5GPa μ=0.75

Brick ID

Bric

k−to

−sh

ell g

ap(m

m)

K=1.15


c)

0 20 40 60 800

10

20

30

40

50

60

7035mm 2rpm E=10GPa μ=0.75

Brick ID

Bric

k−to

−sh

ell g

ap (

mm

)

K=1.12


d)

0 20 40 60 800

10

20

30

40

50

60

70 35mm 2rpm E=5GPa μ=0.75

Brick ID

Bric

k−to

−sh

ell g

ap (

mm

)

K=1.14


e)

0 20 40 60 800

10

20

30

40

50

60

7035mm 2rpm E=2.5GPa μ=0.75

Brick ID

Bric

k−to

−sh

ell g

ap (

mm

)

K=1.17


f)

Figure 14: Brick-to-shell distance recorded during dynamic loading (1 revolution at 0.2 and 2 rpm). Filler pads thickness: 35mm; E: 2.5, 5 and 10 GPa; μ: 0.75.

14

0 20 40 60 800

5

10

15

20

25

3035mm 0.2rpm E=10GPa μ=1.2

Brick ID

Bric

k−to

−sh

ell g

ap(m

m)

K=1.10


a)

0 20 40 60 800

5

10

15

20

25

3035mm 0.2rpm E=10GPa μ=0.75

Brick ID

Bric

k−to

−sh

ell g

ap(m

m)

K=1.10


b)

0 20 40 60 800

5

10

15

20

25

3035mm 0.2rpm E=10GPa μ=0.4

Brick ID

Bric

k−to

−sh

ell g

ap(m

m)

K=1.10


c)

0 20 40 60 800

10

20

30

40

50

60


Brick ID

Bric

k−to

−sh

ell g

ap (

mm

)

K=1.10


d)

0 20 40 60 800

10

20

30

40

50

60

7035mm 2rpm E=10GPa μ=0.75

Brick ID

Bric

k−to

−sh

ell g

ap (

mm

)

K=1.12


e)

0 20 40 60 800

10

20

30

40

50

60


Brick ID

Bric

k−to

−sh

ell g

ap (

mm

)

K=1.11


f)

Figure 15: Brick-to-shell distance recorded during dynamic loading (1 revolution at 0.2 and 2 rpm). Filler pads thickness: 35mm; E:10 GPa; μ: 0.4, 0.75 and 1.2.

15

5. CONCLUSIONS

Due to the size of the kiln and the working environment (e.g. high temperatures) it is often difficult toperform measurements or observe the kiln. Therefore, it is advantageous and desirable to create a model forunderstanding of a kiln. Rotary kiln is a fairly simple design. Nevertheless, it is a difficult task to create arealistic model of it. It would be unrepresentative to scale down the task, thus the large size of the rotarykiln leads to a demanding model in terms of needed computational power. Additionally, the large numberof bricks creates a heavy contact task. Thus, it is important to create an effective model.In this work a model of a cross section of a rotary kiln at the position of rollers was created. Ovality of thecasing was approximated and a bricklaying method was proposed. The efficiency of the model was studiedand a simplified model was justified. The simplified model was based on using purely elastic material modelin the definition of bricks and using a rigid riding tyre. This choice reduces the computational time threefoldfor the presented cases, without significantly affecting the output of the model.Distortion of the casing, brick displacement and effective von Mises stresses were evaluated in static anddynamic loading in cold state at two different rotational speeds for models using three different pad thicknessresulting in affected ovality of the casing. It can be concluded that in static load and at rotational speed of 0.2rpm the brick lining did not experience critical situations for all cases of relative ovality in terms of effectivevon-Mises stresses and it’s integrity. We can also remark that the traditional analytical approximation ofstress in the brick lining, as presented in Figure 2, gives considerably higher stress than the numerical results.Consideration of joints in the numerical model is probably the reason for this.At rotational speed of 2 rpm the integrity of the lining was less stable and severity increased with higherrelative ovality. The out-of-shell displacement of the bricks was up to 40 mm for casings of relative ovalityof 0.27 and 0.4 %, and up to 80 mm for casing of relative ovality of 0.6 %. The effective von-Mises stressesin the lining remained low also at high rotational speed for all the models. We could conclude that stressevaluation was of low interest in further presented cases.We could also conclude that friction between the bricks has insignificant influence on the integrity of thelining. However, lowered Young’s modulus of the bricks affected integrity of the lining negatively.Integration coefficient, K, was implemented for comparison between related cases. With its help the casescould be evaluated quantitatively. With more studies, a limit for an unstable lining could be defined.It should be emphasized that most of its life the rotary kiln is in hot state with different conditions comparedto cold state. The intention of this work was not to directly interpolate the findings from cold state intohot conditions, even though some results are useful. The method itself is of value in future work that willinclude the kiln in hot state.

6. Acknowledgments

LKAB (Luossavaara-Kiirunavaara Aktiebolag) is gratefully acknowledged for the financial support ofthis work.

References

References

[1] M. F. Ashby, Materials and the Environment : Eco-informed Material Choice, ISBN: 9780080884486, Butterworth-Heinemann, 2009.

[2] World Steel Association, World Steel in Figures, ISBN: 978-2-930069-73-9, 2014.[3] J. Saxena, The Rotary Cement Kiln, ISBN: 81-88305-95-2, Tech Books International, 2009.[4] J. Stjernberg, M.-L. Antti, L.-O. Nordin, M. Oden, Degradation of refractory bricks used as thermal insulation in rotary

kilns for iron ore pellet production, Int. J. Appl. Ceram. Tech. 6 (2009) 717–726.[5] J. Stjernberg, B. Lindblom, J. Wikstrom, M.-L. Antti, M. Oden, Microstructural characterization of alkali metal mediated

high temperature reactions in mullite based refractories, Ceram. Int. 36 (2010) 733–740.[6] J. Stjernberg, J. Ion, M.-L. Antti, L.-O. Nordin, B. Lindblom, M. Oden, Extended studies of degradation mechanisms in

refractory lining of a rotary kiln for iron ore production, J. Eur. Ceram. Soc. (2012).

16

[7] F. Herz, I. Mitov, E. Specht, R. Stanev, Influence of operational parameters and material properties on the contact heattransfer in rotary kilns, International Journal of Heat and Mass Transfer 55 (2012) 7941 – 7948.

[8] V. Shubin, Design and service conditions of the refractory lining for rotary kiln, Refract. Ind. Ceram. 42 (2001) 130–136.[9] V. Shubin, The effect of temperature on the lining of rotary cement kilns, Refract. Ind. Ceram. 42 (2001) 45–50.

[10] V. Shubin, Mechanical effects on the lining of rotary cement kilns, Refract. Ind. Ceram. 42 (2001) 245–250.[11] V. Shubin, Thermal effects on the lining of rotary cement kilns (in russian), Ogneupory i Tekhnicheskaya Keramika 4

(2001) 40–47.[12] Deutsche Gesellschaft. Feauerfest- und Schornsteinbau e.V., Refractory Engineering. Materials - Design - Construction,

ISBN: 3-8027-3155-7, Vulan-Verlag Essen, 2004.[13] M. Como, Statics of Historic Masonry Constructions, ISBN: 978-3-642-30132-2, Springer, 2013.[14] F. Damhof, W. Brekelmans, M. Geers, Predictive fem simulation of thermal shock damage in the refractory lining of

steelmaking installations, J. Mater. Process. Technol. 211 (2011) 2091–2105.[15] T. Auer, D. Gruber, H. Harmuth, A. Triessnig, Numerical investigations of mechanical behaviour of refractories, in:

Proceddings of UNITECR 05, 2005, pp. 985–989.[16] H. Harmuth, C. Manhart, T. Auer, D. Gruber, Fracture mechanical characterisation of refractories and application for

assessment and simulation of the thermal shock behaviour, Ceram. Forum Int. 84 (2007) 80–84.[17] K. Andreev, H. Harmuth, Fem simulation of the thermo-mechanical behaviour and failure of refractories- a case study, J.

Mater. Process. Technol 143-144 (2003) 72–77.[18] D. Gruber, K. Andreev, H. Harmuth, Fem simulation of the thermomechanical behaviour of the refractory lining of a

blast furnace, J. Mater. Process. Technol 155-156 (2004) 1539–1543.[19] K. Andreev, S. Sinnema, A. Rekik, S. Allaoui, E. Blond, Compressive behaviour of dry joints in refractory ceramic

masonry, Constr. Build Mater. 34 (2012) 402–408.[20] K. Andreev, S. Sinnema, A. Rekik, E. Blond, A. Gasser, Effects of dry joints on compressive behaviur of refractory linings,

Interceram (2012) 63–66.[21] J. del Coz Dıaz, F. Rodrıguez Mazon, P. Garcıa Nieto, F. Suarez Domınguez, Design and finite element analysis of a wet

cycle cement rotary kiln, Finite Elem. Anal. Des. 39 (2002) 17–42.[22] LS-Dyna, LS-DYNA Keyword User’s Manual, Version R7.0, Livemore Software Technology Corporation, Livemore, Cali-

fornia, USA, 2013.[23] J. Saxena, Refractory Engineering and Kiln Mainteneance in Cement Plants, ISBN: 81-88305-00-6, Tech Books Interna-

tional, 2003.[24] O. Zienkiewicz, R. Taylor, Finite Element Method, volume 1 of ISBN: 0-7506-5049-4, Butterworth-Heinemann, 2000.[25] H. Bjuf, http://brilliant.hoganasbjuf.com/en/Products/Bricks/High-alumina, 2010.[26] D. Ramanenka, High-Temperature Compression Strength of High-Alumina Refractory Bricks Used in Rotary Kilns of

LKAB, Master’s thesis, Lulea University of Technology, Sweden, 2011.[27] C. A. Schacht, Refractories Handbook, ISBN: 0-8247-5654-1, Marcel Dekker, Inc., 2004.[28] A. Gasser, K. Terny-Rebeyrotte, P. Boisse, Modelling of joint effects on refractory lining behaviour, in: Proc. Instn Mech.

Engrs, volume 218, 2003, pp. 19–28.[29] E. Steinbiss, Analysis of mechanical and thermal stresses in the loaded refractory lining of cement kilns, ZKG (1977)

625–627.

17

Paper C

Hot Compression Strength ofHigh-Alumina Refractory Bricks

and Modelling of Rotary Kiln

Authors:Dmitrij Ramanenka, Marta-Lena Antti and Par Jonsen

To be submitted to:Engineering Failure Analysis

35

Hot Compression Strength of High-Alumina Refractory Bricks andModelling of Hot Rotary Kiln

D.Ramanenkaa,∗, P. Jonsena, M-L. Anttib

aDivision of Mechanics of Solid Materials, Lulea University of Technology, Lulea, SwedenbDivision of Materials Science, Lulea University of Technology, Lulea, Sweden

Abstract

Refractory brick lining in rotary kiln for iron-ore production is of great importance for the process due to itsheat insulating properties. However, the high process temperatures make it difficult to study and observethe lining for improvements. With help of numerical simulations some of the problems encountered bybrick linings can be studied. Knowing material properties of the bricks for input in the models is thereforenecessary. Especially material properties at elevated temperatures are poorly documented for this type ofmaterials. In this work three commercial aluminasilicate bricks were tested in compression until failurefor a temperature range of 25-1300 ◦C. The purpose was to evaluate compression strength and Young’smodulus of the fully burned bricks at elevated temperatures. The found data was later used for modellingof hot rotary kiln lined with bricks, whereupon load state and behaviour of the lining were evaluated afterexpansion of the system. It was found that all the three types of bricks increased their compression strengthwith increasing temperature, with a peak in the vicinities of 1000 ◦C. The increase was 50 to 150 % for thedifferent brick types. After 1000 ◦C the compression strength rapidly dropped. Young’s modulus variedbetween 2 to 10 GPa without a clear relation to temperature. From modelling of hot rotary kiln was foundthat average effective von-Mises stresses in the brick lining increased from 0.4 MPa in cold static case to1 MPa in warm static case. Additionally, neither bad integrity of the lining nor rotation of the kiln atproduction speed substantially affected stress state of the lining after thermal expansion.

Keywords: High-temperature, Compression strength, Refractory bricks, Brick lining, Rotary kiln

1. Introduction

Ceramic refractory materials have been used as barriers between hot and relatively cold zones for thou-sands of years and are still of great importance for mankind, especially noticed in the production of cementand iron - two of the most produced materials on Earth [1, 2]. Without insulating properties of refractoriesthese and many other processes would not have been possible.

Manufacturing of iron-ore pellets for iron making is commonly made in the Grate-kiln process [3] wheresintering of pellets is performed in a refractory lined rotary kiln. The kiln is a large cylinder-formed furnacerotating about its axis with typical dimensions of 30-45 m in length and 5-8 m in diameter, and operating intemperatures of up to 1350 ◦C. High-alumina refractory bricks are commonly used in the lining. The serviceconditions inside the kiln are rough requiring chemical, mechanical and thermal resistance of the bricks.Damage of the bricks is inevitable and regularly planned maintenance is needed. However, unplannedemergency stops due to failure of the lining are frequent. The need of slow cooling and repairing followedby slow heating of the kiln make the process of maintenance time consuming (5-14 days) and expensive.

∗Corresponding authorEmail address: [email protected] (D.Ramanenka)URL: http://www.ltu.se/staff/d/dmiram-1.88050?l=en (D.Ramanenka)

To be submitted to Engineering Failure Analysis

Since a single kiln is probably responsible for a large part of the company’s production capacity, may beeven company’s total capacity, it is very important to minimize the risk of unplanned shut-downs.

There have been many improvements of rotary kilns in the last century regarding production capacity,reliability, energy efficiency and improved material properties of the lining [4]. Nevertheless, despite theimprovements there are some gaps in knowledge behind mechanisms of the failure of the lining. An explana-tion to this is the difficulty to observe or study a kiln due to its size and harsh environment. Stjernberg et.al. has contributed to the increase of the knowledge regarding chemical degradation of the aluminasilicatebricks [5, 6, 7]. Boateng [8], Saxena [4, 9] and Schacht [10] are noticeable in the literature regarding rotarykilns. Shubin [11, 12, 13, 14] gives valuable series of scientific papers on the subject of cement kilns withsome focus on analytical calculations. Authors in [15] perform a rigorous work about refractory engineeringcovering description of materials, design theory, practical principles and more.

Today, computers are of a great help for studying various issues without causing production delays orrisking failures, or being limited by the extreme conditions found in rotary kilns. However, academic researchin this field has stagnated and very little documentation can be found regarding finite element analysis ofthe lining, especially of the thermo-mechanical character.

To numerically reproduce lining behaviour at production conditions it is necessary to know materialproperties of the lining as input for a numerical model. In this work three commercial alumina silicate brickmaterials are tested in compression until failure in the range of 25-1300 ◦C. The purpose is to evaluate theunknown compression strength and Young’s modulus of the bricks at elevated temperatures. The obtaineddata is used in numerical modelling of rotary kiln at production temperature. The purpose of this part isto evaluate the load state and behaviour of the brick lining in a rotary kiln after thermal expansion of thesystem. The model is based on the previous study found in [16] and can be regarded as a link from coldstate analysis to hot state analysis. For details on the model consult previous work.

2. Materials

Three as-received brick qualities with trade names Victor HWM, Silox 60 and Alex were tested. Theseare high-alumina bricks [17] manufactured on base of different raw materials. Victor HWM is based onbauxite, Silox on andalusite and Alex on chamotte reinforced with bauxite. Table 1 summarizes someof the material properties and main chemical constituents found in the final burned products. Figure 1

Table 1: Summary of brick properties provided by manufacturer[18]. (AP - apparent porosity; CCS - cold compression strength).

Material properties Constituents (wt%)

ρ (g/cm3) AP (%) CCS (MPa) Al2O3 SiO2 T iO2 Fe2O3 CaO Alkalis

Victor HWM 2.7 (2.65-2.75) 18 (17-21) 80 (60-100) 79 17 2.2 1 0.2 0.5Silox 60 2.45 (2.40-2.50) 17 (15-19) 70 (50-90) 59 37 1.5 0.9 0.1 0.5Alex 2.33 (2.25-2.40) 19 (17-21) 50 (30-70) 54 40 2.1 1.4 0.3 1.3

represents optical light microscope and QEMSCAN (Quantitative Evaluation of Minerals by SCANningelectron microscopy) images of the brick samples. The later showing distribution of mineral phases in thesamples (previously reported by Stjernberg [19]). Victor HWM is dominated by corundum (α − Al2O3),Silox 60 by mullite (3Al2O3 · 2SiO2) and Alex is primarily a blend of the mentioned phases. Also unburnedandalusite (Al2SiO5) can be found in Silox 60 and Alex. According to the manufacturer Victor HWM, Silox60 and Alex contain 8-10, 14-16 and 19-20 wt% glassy phase respectively. Grain size is up to 5 mm, wherelargest grains are lumps of smaller grains and most of the glassy phase is in the matrix between the grains.Mullite grains however have a tendency to trap amorphous silica during transformation from andalusite [20].

2

1 mma) 1 mmb) 1 mmc)

1 mmd) 1 mme) 1 mmf)

AndalusiteMulliteMullite*Corundum

Figure 1: Light microscope (a-c) and QemScan (d-f) images (not correlated) of Victor HWM (a, d), Silox 60 (b, e)) and Alex(c, f). * denotes mullite with low aluminium content.

3. Method and Experimental procedure

3.1. High temperature compression tests

A test methodology was developed for performance of hot crushing tests in this work. In order to definethe methodology, standards for cold crushing strength (CCS) tests, hot modulus of rupture (MOR) testsand recommendations from other works were advised [21, 22, 23, 24].

The compression tests were performed on a universal screw testing machine (Dartec) equipped with a100 kN load cell (Instron) and a water jacketed furnace (Severn Science Limited, max temp. 1500 ◦C).Equipment is illustrated in Figure 2.

The specimens were chosen to be in form of cylinders with height and diameter of 35 mm. The testswere deduced at: room temperature, 700 ◦C, 900 ◦C, 1000 ◦C, 1100 ◦C, 1200 ◦C and 1300 ◦C. The heatingrate was 300 ◦C/h up to 1000 ◦C and 150 ◦C/h after 1000 ◦C. A pre-load of 1 kN was used during theheating procedure. A small amount of alumina powder was placed below every sample believing it wouldeven the load distribution e.g. when the faces of the samples are not perfectly parallel. When the testtemperature was reached the samples were soaked for 2 hours and thereafter compressed until collapsed ata rate of 1mm/min. The following cooling rate was 150 ◦C/h in all cases. Deformation of brick specimenswas approximated by measuring compression of the system at test-temperatures and then subtracting itfrom total deformation. For measurements of the system’s compression a high-density alumina dummy wasused instead of brick specimens in the same test conditions. At these temperatures and loads high-densityalumina dummy has negligible deformation compared to the brick specimens.

3

1

2 3

4

5 6 7

3

2

8

Figure 2: Experimental set-up: 1-Load cell; 2-Water cooled steel supports; 3-Alumina bars; 4-Furnace; 5-Thermocouples;6-Specimen; 7-Heating elements; 8-Chamber.

3.2. Modelling of hot rotary kiln

The concept used for modelling rotary kiln in this work is identical with the one in previously reportedstudy [16] evaluating kiln in cold state. In this study the kiln is evaluated in hot state. Commercialsoftware LS-Dyna [25] is used for numerical calculations. Only the final hot steady state is of interest −therefore heating rate was speeded up. Dimension of bricks are as defined in the previous study. Dimensionsof the kiln are given in Figure 3. The pads and the casing are modelled with purely elastic material,


defined by typical low-alloy carbon steel properties. Material properties of brick material are defined by a

4

thermal, piecewise linear elasto-plastic material model available in LS-Dyna (MAT 4). Summery of brick’smechanical properties with respect to temperature is found in Table 2. The selected properties are supportedby the results from the experiments in this study. Properties in-between temperature ranges are linearlyextrapolated by the model. The contact between parts was defined by mortar, penalty based segment-to-segment contact. The dynamic and static friction coefficients of brick-to-brick contact are both set to thesame value of 0.75 [26]; and for brick-to-casing and casing-to-tyre contacts to 0.5. The rotation of the kilnis onset by prescribed rigid body motion of rigid riding tyre. Implicit integration is used for the numericalcalculation. The heating of the kiln in the model is induced by defining temperature boundary conditions,

Table 2: Summery of mechanical material properties of the bricks used as input in the model.

Temperature (K) 273 973 1173 1273 1573

Bulk density (kg/m3)[18] 2700 2700 2700 2700 2700Young’s modulus (GPa) 3 7 4 2 1Poisson’s ratio [10] 0.2 0.2 0.2 0.2 0.2Coef. thermal expansion (×10−6K−1)[18] 6 6 6 6 6Yield stress (MPa) 50 60 80 80 30Tangential modulus (MPa) 100 100 100 100 50

as illustrated in Figure 4. Hot face temperature of the bricks (1300 ◦C), outer temperature of the casing(280 ◦C) and the tyre (140 ◦C) are provided by the user of the kiln. Temperature of the pads is assumedto be the same as outer temperature of the casing (280 ◦C). Inner temperature of the casing (297 ◦C) isobtained by applying linear heat transfer relationship for two surfaces that share the same boundary. Innertemperature of the tyre (250 ◦C) is an approximation. The heat transfer coefficient for brick material andsteel is set to 200 WK−1m−2 to save computational time. This is much higher than in reality. However,since only the final state of the simulation is of interest the heating rate is not essential in this study. Linearcoefficient of thermal expansion for steel is set to 12 · 10−6K−1.

Bricks

Casing

Pads

Tyre

1300

◦ C

297

◦ C

280

◦ C

250

◦ C

140

◦ C

Figure 4: Temperature boundary conditions used in the model

For simplicity, expansion of the elastic riding tyre is calculated separately. After the expansion is known,the elastic tyre is replaced with a rigid riding tyre having the new expanded inner diameter. It was previouslyshown in [16] that such simplification affects output of the model negligibly but decreases computationaltime threefold.

5

Three cases are evaluated. Firstly, expansion of the lining that has perfect integrity. Secondly, expansionof the lining lining whose integrity has been disturbed due to shifting position of the bricks after rotationin cold state (case of cold rotation can be found in [16]). Lastly, the kiln having bad integrity is rotated fiverevolutions at 2 rpm.

4. Results and Discussion

4.1. High Temperature Compression Tests

Figure 5 represents normalized and true compression strength of the tested materials. Normalization isrelative to the bricks’ compression strength at room temperature from the same batch. The lines denotethe average compression strength trend in the tested temperature range. The authors are aware thatthe conducted tests lack some statistical foundation therefore the results should be regarded only as anindication. However, we believe the results give a valuable insight into some important mechanical propertiesrarely evaluated. Additionally, since the bricks have naturally large scatter in properties the importance ofstatistically satisfying number of tests decreases.

25 700 900 1100 13000

0.5

1

1.5

2

2.5

Temperature (°C)

Rel

ativ

e ch

ange

SiloxSilox avg. trendVictorVictor avg. trendAlexAlex avg. trend

25 700 900 1100 1300

20

40

60

80

100

Temperature (°C)

σ (M

Pa)

SiloxSilox avg. trendVictorVictor avg. trendAlexAlex avg. trend

Figure 5: Normalized and actual compression strength at RT and 700-1300 ◦C interval.

The general trend for the three qualities of the bricks was increasing of compression strength withincreased temperature up to 900-1000 ◦C. Until this range the strength increased by factor 1.5, 2 and over2.5 for Victor HWM, Silox 60 and Alex respectively. Thereafter a sudden drop in strength was observed.The fastest decrease was between 1000-1100 ◦C. At 1300 ◦C the strength was decreased by some 80-95 %for all the bricks compared to their highest observed compression strength. Also, it is noticed that VictorHWM has the lowest variation in strength while Alex has the largest variation in strength.

Since the large strength increase was not expected prior to the experimental work the tests becamelimited by the capacity of the load cell (100 kN). Because of that, only the samples from the batches withlowest compression strength at room temperature were tested in high temperatures in order to avoid therisk of unsuccessful tests. However, the proportion in strength between batches seems to remain at highertemperatures. Based on this remark the highest compression strength in the region of 900-1000 ◦C for thesamples with highest room temperature strength can be estimated to be above 150, 190 and 180 MPa forVictor HWM, Silox 60 and Alex respectively, even though such test could not be conducted.

The origin to the brick’s strength increase cannot be related to phase transformations. Possible transfor-mations (e.g. andalusite to mullite) require higher temperatures and/or much longer soaking times [27]. Theincrease is believed to be due to interlocking effect between the phases due to mismatch of their coefficientsof thermal expansion. The amorphous phase has up to 5-10 times lower coefficient of thermal expansionthan mullite and corundum [28].

6

The decrease in strength is most likely related to the softening of the glassy phase when approaching itsglass transition temperature. Figure 6 represents typical stress-strain behaviour of bricks during compressionat different temperatures. The soft part in the beginning of the room temperature (RT) test is due tocompression of the powder below the sample. The behaviour of the samples is quasi-brittle with a clearelastic part up to vicinities of 900 ◦C. After 1000 ◦C the behaviour is of a more ductile character.

Table 3 summarizes statical measurements of secant Young’s modulus of the brick samples for all thetests up to 1000 ◦C. Measurement were made in the interval of 40 to 50 MPa. The result imply that VictorHWM is the stiffer material in the whole temperature range, while Alex is the softest which is coherentwith the alumina content. The tests indicate that Young’s modulus of bricks is in the range of 2-10 GPafor temperatures up to 1000 ◦C. No clear trend related to temperature can be observed from the conductedtests. However, perhaps the trend is that it increases at 700 ◦C and decreases at 1000 ◦C. After 1000 ◦Cthe definition of Young’s modulus becomes ambiguous.

0 0.01 0.02 0.03 0.04 0.05 0.06

20

40

60

80

100

RT

700 °C

900 °C

1000 °C

1100 °C

1200 °C

1300 °C

Strain (−)

Str

ess

(MP

a)

Figure 6: Typical stress-strain behaviour during compression tests at different temperatures. Here: compression curves ofVictor HWM samples from the same batch.

Table 3: Youngs modulus measurements

Temperature (◦C)

Brick type RT 700 900 1000

Young′s modulus (GPa)Victor HWM [4.9; 4.4; 8.6] [10.2] [5] [3.5; 2.1; 3.4]Silox 60 [3.5; 3.0; 2.3] [10.3; 3.4] [4.9; 2.9] [2.7; 2.0; 1.7]Alex [1.8] [2.2] - [1.5]

4.2. Modelling of hot rotary kiln

Figure 7 shows effective von-Mises stress distribution in the brick linings with ordered and disorderedintegrity after the expansion of the system. The maximum von-Mises stresses reach approximately 1.7 and2.5 MPa respectively, implying that bad integrity of the lining has small effect on the stress state in thelining.

After the expansion the kiln with disordered brick integrity is rotated five revolution. Figure 8 showsevolvement of effective von-Mises stress in one element of every brick during the revolutions. The elementis positioned in the centre of the brick’s hot face (facing inside of the kiln). A comparison to the stressevolvement in the cold state of the kilns found in the previous study [16] is made as well. The averageeffective von-Mises stress in the chosen elements during revolutions is close to 1 MPa. In the identical

7

case in cold conditions the average effective von-Mises stress during rotation of the kiln was measured toapproximately 0.3 MPa.

Figure 7: von-Mises stress (Pa) distribution after the expansion in the linings with ordered (left figure) and disordered (rightfigure) brick positions. (Missing piece of the lining was intentionally removed during collection of the results due to anunessential numerical reason)

0 20 40 60 800

0.5

1

1.5

235mm 2rpm E=10GPa μ=0.75 HOT

Brick ID

Effe

ctiv

e vo

n M

ises

str

ess

(MP

a)


0 20 40 60 800

0.5

1

1.5

2 35mm 2rpm E=10GPa μ=0.75 COLD

Brick ID

Effe

ctiv

e vo

n M

ises

str

ess

(MP

a)


Figure 8: Effective von-Mises stress evolvement on the inner face of the lining during 5 revolutions for kiln in hot (right figure)and cold (left figure) state.

8

5. Conclusions

In this work we have performed rarely seen hot compression tests of three similar refractory materialsused as insulators of rotary kilns in iron-ore pellets industry. The results clearly showed that compressionstrength of the tested materials increased considerably with temperature until approximately 1000 ◦C. Suchstrength increase was unexpected and is poorly documented in literature. We conclude that measuredincrease is likely to be related to the difference in the thermal expansion coefficient of the present phases inthe bricks. Measured Young’s modulus varied significantly but no clear trend related to temperature couldbe observed from the performed test.

The found mechanical properties of the bricks were used in finite element modelling of a warm large-sized rotary kiln. Temperature boundary conditions were defined and the kiln was allowed to expand untila steady state was reached. Maximum effective von-Mises stresses where found to be around 1.5 and 2.5MPa in the ordered and disordered cases of the lining respectively − compared to some 0.5 MPa in coldstate found in a previous study. It implies that worsened integrity of the lining have little effect on stressincrease during thermal expansion. Also, fast rotation after expansion of the kiln showed to have little effecton the stress state of the lining. We conclude that thermal expansion of the kiln was relatively harmless tothe lining under conditions studied in this work.

References

References

[1] World Steel Association, World Steel in Figures, ISBN: 978-2-930069-73-9, 2014.[2] M. F. Ashby, Materials and the Environment : Eco-informed Material Choice, ISBN: 9780080884486, Butterworth-

Heinemann, 2009.[3] Metso Corporation, Iron ore pelletizing. Grate-Kiln system, Metso Corporation, 2012.[4] J. Saxena, The Rotary Cement Kiln, ISBN: 81-88305-95-2, Tech Books International, 2009.[5] J. Stjernberg, M.-L. Antti, L.-O. Nordin, M. Oden, Degradation of refractory bricks used as thermal insulation in rotary

kilns for iron ore pellet production, Int. J. Appl. Ceram. Tech. 6 (2009) 717–726.[6] J. Stjernberg, B. Lindblom, J. Wikstrom, M.-L. Antti, O. M, Microstructural characterization of alkali metal mediated

high temperature reactions in mullite based refractories, Ceram. Int. 36 (2010) 733–740.[7] J. Stjernberg, J. Ion, M.-L. Antti, L.-O. Nordin, B. Lindblom, M. Oden, Extended studies of degradation mechanisms in

refractory lining of a rotary kiln for iron ore production, J. Eur. Ceram. Soc. (2012).[8] A. Boateng, Rotary Kilns. Transport Phenomena and Transport Process, ISBN: 9780750678773, Elsevier Inc., 2008.[9] J. Saxena, Refractory Engineering and Kiln Mainteneance in Cement Plants, ISBN: 81-88305-00-6, Tech Books Interna-

tional, 2003.[10] C. Schacht, Refractory Linings: Thermomechanical Design and Applications., ISBN: 0-8247-9560-1, Marcel Dekker, 1995.[11] V. Shubin, Design and service conditions of the refractory lining for rotary kiln, Refract. Ind. Ceram. 42 (2001) 130–136.[12] V. Shubin, Mechanical effects on the lining of rotary cement kilns, Refract. Ind. Ceram. 42 (2001) 245–250.[13] V. Shubin, The effect of temperature on the lining of rotary cement kilns, Refract. Ind. Ceram. 42 (2001) 45–50.[14] V. Shubin, Thermal effects on the lining of rotary cement kilns (in russian), Ogneupory i Tekhnicheskaya Keramika 4

(2001) 40–47.[15] Deutsche Gesellschaft. Feauerfest- und Schornsteinbau e.V., Refractory Engineering. Materials - Design - Construction,

ISBN: 3-8027-3155-7, Vulan-Verlag Essen, 2004.[16] D. Ramanenka, J. Stjernberg, P. Jonsen, Fem investigation of global mechanisms affecting brick lining stability in a rotary

kiln in cold state, Engineering Failure Analysis (2015). (in press) http://dx.doi.org/10.1016/j.engfailanal.2015.10.023.[17] ASTM standard C27 - 98, Standard Classification of Fireclay and High-Alumina Refractory Brick, volume 15.01, ASTM

International, 1998.[18] Hoganas Bjuf, (online) http://brilliant.hoganasbjuf.com/en/Products/Bricks/High-alumina, 2010.[19] J. Stjernberg, Degradation Mechanisms in Refractory Lining Materials of Rotary Kilns for Iron Ore Pellet Production,

Ph.D. thesis, Lulea University of Technology, Sweden, 2012.[20] J. Ildefonse, V. Gabis, F. Cesbron, Mullitization of andalusite in refractory bricks, Key Engineering Materials 132-136

(1997) 1798–1801.[21] ASTM C133 - 97, Standard Test Method for Cold Crushing Strength and Modulus of Rapture of Refractories, volume

15.01, ASTM International, 1998.[22] European standard EN 993 - 5, Methods of Test For Dense Shaped Refractory Products - Part 5: Determination of Cold

Crushing Strength, European Committee For Standardization, 1998.[23] ASTM C583 - 00, Modulus of Rupture of Refractory Materials at Elevated Temperatures, volume 14.03, ASTM Interna-

tional, 2000.

9

[24] N. Prompt, E. Ouedraogo, High temperature mechanical characterisation of an alumina refractory concrete for blastfurnace main trough part i. general context, J Eur. Ceram. Soc. 28 (2008) 2859–2865.

[25] LS-Dyna, LS-DYNA Keyword User’s Manual, Version R7.0, Livemore Software Technology Corporation, Livemore, Cali-fornia, USA, 2013.

[26] A. Gasser, K. Terny-Rebeyrotte, P. Boisse, Modelling of joint effects on refractory lining behaviour, in: Proc. Instn Mech.Engrs, volume 218, 2003, pp. 19–28.

[27] M.-L. Bouchetou, J.-P. Ildefonse, J. Poirier, Mullite grown from fired andalusite grains: the role of impurities and of thehigh temperature liquid phase on the kinetics of mullitization and consequences on thermal shocks resistance, Ceram. Int.31 (2005) 999–1005.

[28] G. Fiquet, P. Richet, G. Montagnac, High-temperature thermal expansion of lime, periclase, corundum and spinel, Phys.Chem. Minerals 27 (1999).

10

Numerical Evaluation of Brick - DiVA...

Documents

Transcript of Numerical Evaluation of Brick - DiVA...