Phase Chemistr\ in Process Models for Cement Clinker and Lime Production

3KDVHFKHPLVWU\LQSURFHVVPRGHOVIRU FHPHQWFOLQNHUDQGOLPHSURGXFWLRQ

%RGLO+|NIRUV

7KHUPDO(QHUJ\&RQYHUVLRQ/DERUDWRU\ 'HSDUWPHQWRI$SSOLHG3K\VLFVDQG(OHFWURQLFV 'RFWRUDOWKHVLV

Phase chemistry in process models for cement clinker and lime production Doctoral thesis

Bodil Hökfors

Thermal Energy Conversion Laboratory Department of Applied Physics and Electronics SE-901 87 Umeå Umeå 2014

Copyright©Bodil Hökfors ISBN:978-91-7459-801-8 Front: At the top left is a microscopy picture of a cement clinker nodule. The figure to the right is a schematic representation of the process modell for cement clinker production that is described in this thesis. Printed by: Print & Media, Umeå University, Umeå, Sweden 2014 Table of Contents

Abstract III Acknowledgements IV Populärvetenskaplig sammanfattning V 1. Introduction 1 1.1. Sustainability in the cement and lime industry 1

1.1.1. Energy and carbon dioxide (CO 2) 1 1.1.2. Carbon Capture and Storage (CCS) 3 1.2. Chemical calculations, thermodynamic data and process models 5 1.2.1. Chemical equilibrium calculations and thermo-dynamic data 6 1.2.2. Types of process models 6 1.3. The cement and lime production processes 8 1.3.1. The cement process 8 1.3.2. The lime process 10 1.3.3. Specific features of the rotary kiln processes 11 1.3.4. Types of fuel 12 1.3.5. Cement clinker and lime phase chemistry 14 1.3.5.1. Major elements (Ca, Si, Al, Fe and Mg) 14 1.3.5.2. Minor elements (K, Na, S, Cl, F, P, Zn and Ti) 18 1.3.5.3. Fuel elements (C, H, O and N) 18 1.4. Scope of the thesis 20 2. Methods used in developed process models 21 2.1. Aspen Plus™ 21 2.2. ChemApp™ 21 2.3. Thermodynamic data from FACT 22 2.4. Process schemes 22 2.5. Thermodynamic database developed for cement clinker and lime applications 25 2.5.1. Fuel definition, including chemistry and enthalpy 26 2.6. Phase chemical equilibrium calculations 27 2.6.1. Non-equilibrium cooling according to Gulliver-Scheil 27 2.6.2. Energy balance 28 2.7. Strengths and weaknesses of the developed models 28 3. Results and discussion 31 3.1. Validation of simulation models 31 3.1.1. Cement clinker process models 31 3.1.2. Process model of a lime rotary kiln 35 3.1.3. Development of the thermodynamic database 37 3.1.3.1. Validation of the clinker composition 38 3.1.3.2. Influence of phosphorus (P) in cement clinker production 42 3.2. Simulated scenarios with validated models 44

I 3.2.1. Oxygen-enriched combustion in cement clinker production 45 3.2.2. Influence of co-combustion on the lime production process and the lime product 48 3.2.3. Partial and full oxy-fuel combustion in cement clinker production 50 3.2.4. Oxy-fuel combustion in lime production 56 4. Conclusions 60 5. Abbreviations 62 6. References 63

II Abstract

The goal of the thesis is to evaluate if developed phase chemical process models for cement clinker and lime production processes are reliable to use as predictive tools in understanding the changes when introducing sustainability measures. The thesis describes the development of process simulation models in the application of sustainability measures as well as the evaluation of these models. The motivation for developing these types of models arises from the need to predict the chemical and the process changes in the production process, the impact on the product quality and the emissions from the flue gas. The main chemical reactions involving the major elements (calcium, silicon, aluminium and iron) are relatively well known. As for the minor elements, such as sodium and potassium metals, sulphur, chlorine, phosphorus and other trace elements, their influence on the main reactions and the formation of clinker minerals is not entirely known. When the concentrations of minor and trace elements increase due to the use of alternative materials and fuels, a model that can accurately predict their chemistry is invaluable. For example, the shift towards using less carbon intensive fuels and more biomass fuels often leads to an increased phosphorus concentration in the products. One way to commit to sustainable development methods in cement clinker and lime production is to use new combustion technologies, which increase the ability to capture carbon dioxide. Introducing oxy-fuel combustion achieves this, but at the same time, the overall process changes in many other ways. Some of these changes are evaluated by the models in this work. In this thesis, a combination of the software programs Aspen Plus™ and ChemApp™ constitutes the simulation model. Thermodynamic data from FACT are evaluated and adjusted to suit the chemistry of cement clinker and lime. The resulting model has been verified for one lime and two cement industrial processes. Simulated scenarios of co-combustion involving different fuels and different oxy-fuel combustion cases in both cement clinker and lime rotary kiln production are described as well as the influence of greater amounts of phosphorus on the cement clinker quality.

Keywords: Process modelling, phase chemistry, cement clinker, lime, sustainability, CO 2, energy.

III Acknowledgements

Sincere thanks to many people for making this thesis possible; friends and colleagues at Cementa, Cementa Research and Umeå University, especially to Rainer Backman, Dan Boström and Lars Bäckström at Umeå University, to Erik Viggh, Stefan Sandelin, Bo-Erik Eriksson, Thomas Lind and Anders Lyberg at Cementa, to Matias Eriksson and Kjell Dahlberg at Nordkalk, to Kristina Johansson and Johan Larsson at Cementa Research and to Marianne Thomaeus and Jan Bida at MinFo (Swedish Mineral Processing Research Association). I‘m grateful to the Swedish Energy Agency (No. 2006-06679, Project 30527-1), to MinFo and to the Swedish National Research Platform Bio4Energy for financial support. Last but not least; to great friends, to my family and especially to Liv, Emil, Dan and Hans, thank you!

IV Populärvetenskaplig sammanfattning

Tillverkningen av cement och kalk är energiintensiv och innebär utsläpp av stora mängder koldioxid (CO 2). CO 2 har två ursprungskällor, CO 2 kan antingen härstamma från råmaterialet eller från bränslet. Vid cement- och kalktillverkning består råmaterialet till stor del av kalksten. När kalksten värms upp erhålls den önskvärda kalken och samtidigt bildas CO 2. CO 2 från bränslet kan minskas genom att energieffektivisera processen, ersätta traditionella fossila bränslen med biomassa eller med bränslen med lägre kolhalt. Det finns färre möjligheter att minska CO 2 som härstammar från råmaterialet. Inom cementtillverkning finns möjlighet att ersätta en del av råmaterialet med restprodukter från järn- och stålindustrin där CO 2 redan är avdriven.

Genom att rena rökgaserna från CO 2 kommer man åt CO 2 från både råmaterial och bränsle. Tekniken heter CCS (Carbon Capture and Storage) och innebär att fånga CO 2 ur rökgaserna, komprimera och lagra dessa djupt ned i geologiska formationer. Oxy-fuel förbränning kan med fördel kombineras med CCS. Härvid ersätts förbränningsluften helt eller delvis med ren syrgas. Förbrännings- temperaturen ökar då avsevärt och kontrolleras genom att återcirkulera rökgaser. Rökgasernas koncentration av CO 2 kan därmed ökas från ca 25 till

75%, vilket ger möjligheter till effektivare CO 2-avskiljning. Målet med avhandlingen är att utvärdera hur väl det utvecklade simuleringsverktyget fungerar att använda för att förutspå förändringars påverkan på produktkvaliteten, tillverkningsprocessen och den yttre miljön. De använda termodynamiska processmodellerna beskriver de komplexa kemiska reaktionerna i både gasfasen och i materialet i respektive tillverkningsprocess. Modellerna baseras på naturvetenskapliga grundregler vilket ger en generell predikterbarhet för nya system som ännu inte testats i verkliga fall. Tillgänglig kemisk grunddata är utvärderad och till vissa delar förbättrad för att bättre passa för cementklinker- och kalktillverkning. Modellernas riktighet har verifierats med data för två cementfabriker och en kalkfabrik. De förändringar som utvärderats i detta arbete syftar till ökad miljömässig hållbarhet inom cement- och kalkindustrin. Modellerna kan användas till att studera inverkan av förändringar så som införande av nya slags bränslen, energieffektiviseringsåtgärder och nya förbränningsteknologier. Resultateten skulle ytterligare förbättras med utökade och förbättrade termodynamiska data och om modellen kompletterades med möjlighet att simulera stoftbildning i gasfasen.

V List of papers

This thesis is based on the following papers, which are separately appended and referenced by their corresponding roman numerals, I to VI, in the text.

I. A predictive chemistry model for the cement process B. Wilhelmsson Hökfors, E. Viggh and R. Backman. ZKG International 7 (2008) 60-70. In Chinese, ZKG China 3 (2008) 40-44. In Russian, ZKG Russia 2 (2009) 53-58.

II. Improved process modeling for a lime rotary kiln using equilibrium chemistry B. Hökfors, M. Eriksson and R. Backman. Journal of Engineering Technology (2012) Spring 8-18.

III. Oxy-fuel combustion in rotary kiln lime production M. Eriksson, B. Hökfors and R. Backman. Submitted to Energy Science & Engineering.

IV. Modelling the cement process and cement clinker quality B. Hökfors, M. Eriksson and E. Viggh. Advances in Cement Research (2014). DOI: 10.1680/adcr.13.00050

V. On the phase chemistry of Portland cement clinker B. Hökfors, D. Boström, E. Viggh and R. Backman. Accepted 30 th January 2014 to be published in Advances in Cement Research.

VI. Simulation of oxy-fuel combustion in cement clinker manufacturing B. Hökfors, E. Viggh, and M. Eriksson. Accepted 6 th January 2014 to be published in Advances in Cement Research.

Reprints have been included within this thesis with permission from the publishers.

1. Introduction

1.1. Sustainability in the cement and lime industry

1.1.1. Energy and carbon dioxide (CO 2) The world‘s cement production was 3.6 billion tonnes in 2012 and is expected to increase (IMF, 2013). Cement is the fundamental component of concrete, composing anywhere from 10 to 15% of the concrete mix. Other than cement, concrete is composed of a mixture of aggregates (sand and gravel or crushed stone) and water. Concrete is essential to building construction, infrastructure development and so on and is the second most consumed product on earth.

The production of cement and lime is energy intensive, and the CO 2- emissions are substantial. The emissions from the top 10% efficient installations in the EU are 0.766 kg CO 2 per tonne of cement clinker and

0.954 kg CO 2 per tonne of lime (2011/278/EU). The energy needed to produce 1 tonne of clinker is 3.6 GJ for a modern dry process (GNR Project reporting CO 2, 2012) and 5.7 GJ for a wet process (Taylor, 1997). Similarly, the energy needed to produce 1 tonne of lime is between 3.6 and 6.1 GJ. The lowest energy requirement is from a regenerative shaft kiln, and the highest value results from the production of lime in a rotary kiln (IPCC, 2009). In this thesis, only rotary lime kiln processes are considered. The cement and lime industry has traditionally used fossil fuels as their primary source of energy. In recent years, however, there has been a trend to increase the amount of biomass- and waste-based fuels. The total emissions from the cement industry in 2011 were 573 million tonnes of CO 2. The cement industry accounted for 5% of the total global man-made CO 2 (GNR Project reporting CO2, 2012). To produce cement and lime, the main chemical reaction required is the calcination of limestone (composed mainly of calcium carbonate). The calcination reaction evolves CO 2 from two sources. These sources are the combustion emissions and the process emissions, see Figure 1 . Combustion emissions are related to the combustion of fuels for energy production, and process emissions are related to the release of CO 2 from the raw materials

(CaCO 3 and MgCO 3).

Figure 1. CO 2 emissions during the production of cement clinker and lime originate from either combustion or process CO2. CaCO 3 is the main mineral in both cement clinker and lime production.

On average, the CO 2 resulting from combustion accounts for 40% and the process CO 2 accounts for 60% of the total emissions in cement clinker production. When lime is produced in a rotary kiln, the process emissions can be as high as 65%. The higher amount is due to the raw material content, which is close to 100% carbonates, which are needed for lime production. Within the European Emission Trading Scheme (EU ETS), a benchmark for each industrial sector is set starting at 10% of the most efficient installations from the year 2007-2008. This level is referred to as the free allocation level. Grey cement clinker production has 766 allowances or 766 kg CO2 per tonne of product, and the lime production sector has 954 allowances or 954 kg CO 2 per tonne of lime, (2011/278/EU). During the years from 2013 to 2020, the allowances will be multiplied annually by a factor (ranging between 0.800 and 0.300) reducing the amount of free allocations. The cement and lime industry is facing new challenges while trying to strengthen its environmental position. For instance, some of the options the cement and lime industry has to reduce its CO 2-emissions are to improve the energy efficiency of the processes and to increase the use of biomass-based fuels. For the cement sector, the options also include the ability to reduce the clinker content in the cement by adding more slag, fly ash, or other additives and to introduce low-energy clinker types (Damtoft et al ., 2008).

Possible methods to reduce the CO 2 combustion emissions in cement clinker and lime production are:

• To improve the energy efficiency of the production process. • To use less carbon intensive fuels. • To replace fossil fuels with biomass based fuels.

To further reduce the process emissions in cement clinker and lime production would require the use of raw materials lacking carbonate minerals included. One alternative option is to utilise slags from steel industry. Slags are produced as by-products when limestone and lime is added in the process to extract impurities and to flux the gangue oxides that accompany the ore. The slags are removed from the process and can be potentially used in other processes. The slags are, for the most part, composed of CaO, MgO, Al 2O3 and SiO 2 and often have a very high CaO content. Due to the high temperatures necessary to produce steel, the steelmaking slags are already calcined and do not carry any by-product of

CO 2. The usage of slags in cement production is possible only if the chemical composition of the slag is appropriate for the product quality. Metallurgical slag is already used to some extent in cement clinker production, but it is not an option in the lime industry. The aim of this work is to develop a simulation model as a tool that is capable of predicting the changes of the process, predicting the product and predicting environmental impact following the introduction of new fuel types and to evaluate new combustion technologies.

1.1.2. Carbon Capture and Storage (CCS) The concept of Carbon Capture and Storage denotes any technique that captures and permanently stores CO 2 instead of releasing it into the atmosphere. There are three capturing technologies for large point sources: pre-, post- and oxy-fuel combustion (IEA, 2013). A pre-combustion method is one where syngas is produced from the fuels, making it possible to separate the CO 2. The CO 2-free syngas is then used as a fuel in the kiln. In cement clinker and lime applications, a pre-combustion technique would only remove CO 2 combustion emissions, which is why post-combustion and oxy-fuel combustion are the only applicable technologies (Barker et al. , 2009). A typical scheme of a kiln adapted for oxy-fuel combustion is shown in Figure 2b .

Methods to reduce process emissions are:

• To use raw materials free of CO 2. • To remove CO 2 by using oxy-fuel carbon capture techniques. • To remove CO 2 by using post-combustion carbon capture techniques.

The CO 2 emissions captured from processes need further treatment. CO 2 can be utilised in some other process or cleaned, compressed and sent to permanent storage, which is known as Carbon Capture and Storage. One of the most promising post-combustion options is chemical absorption with amines (Barker et al. , 2009). Technical solutions for oxy- fuel combustion in the cement clinker and lime processes will be different compared to the conventional flue gas cleaning. Theoretically, flue gases from cement clinker production are more beneficial as a result of the higher

CO 2 concentration, typically 25 vol-% compared with 14 vol-% for coal fired power plants (IEA GHG, 2008). Extensive research is still required to fully understand all the potential impacts on the clinker burning process following the introduction of oxy-fuel combustion (IEA-WBCSB, 2009). Research of CCS in cement processes has been going on since 2007 by the ECRA (European Cement Research Academy) (Hoenig et al. , 2012). The focus between 2009 and 2011 was on amine-based post-combustion carbon capture and the oxy-fuel combustion process. Simulations were performed to size equipment for amine scrubbing, and laboratory studies were conducted to examine the critical substances degrading the CO 2 solvent mono-ethanol amine. A pilot plant has been developed and designed to make further tests at larger scales possible. The pilot plant has been built at the cement plant in Brevik, Norway. In addition to amine scrubbing, three other techniques will be tested: membrane separation, a solid sorbent technique and a regenerative carbonate cycle. Oxy-fuel research at the ECRA has involved laboratory furnace tests under oxy-fuel conditions. Other important factors such as the design of the burner and the two-stage cooler, the durability of the refractory material resulting from temperature profile changes, the techno-economic evaluation of false air ingress and the clinker cooler layout were studied. These laboratory tests found no negative impact on the clinker quality.

Figure 2a. A principal scheme of a cement clinker kiln burning fuel in air.

Figure 2b. A principal scheme of a cement clinker kiln using oxy-fuel combustion. The combustion oxidant is oxygen (O 2) instead of air, and the exhaust gases are recirculated.

Normally, combustion uses air as the oxidant, as shown in Figure 2a . The aim of oxy-fuel combustion is to reduce the concentration of nitrogen (N 2) in the combustion air. An Air Separation Unit (ASU) is utilised for the N 2 separation as is shown in Figure 2b . The combustion temperature will then increase as it approaches combustion with almost pure oxygen (O 2). By recirculating the exhaust gases to the combustion zone, the temperature is controlled.

1.2. Chemical calculations, thermodynamic data and process models Modelling is a versatile tool that can be used to gain valuable information and to make decisions about many process related issues, including investment costs and environmental concerns. A global equilibrium analysis of the process is of key interest because it provides a basis to consider the whole process, or just distinct parts of it, as an equilibrium reactor. Given the temperature or the enthalpy changes of the reactor along with the ambient pressure and a list of possible chemical compounds with their corresponding thermodynamic data, the chemical equilibrium composition can be found. If the total enthalpy is fixed, then the results produced are the amount and the composition of each phase and the equilibrium temperature. Thus, one or several reactors linked together with the proper mass flows can describe the entire process.

1.2.1. Chemical equilibrium calculations and thermodynamic data Chemical phase equilibrium calculations are valuable for the understanding of clinker reactions, as has been successfully shown by (Barry and Glasser, 2000). Their calculations were based on thermodynamic data from

MTDATA™. A system of CaO-SiO 2-Al 2O3-Fe 2O3 was modelled, but the thermodynamic data neither contain the properties of sodium and potassium solubility in the clinker mineral phases nor the properties of sulphates in liquid salt. A comparison of chemical equilibrium calculations of MTDATA TM and FACT data with systems of up to 7 oxide components have been made available (Ghanbari Ari et al. , 2004). The stable temperature ranges and the cooling of the products are shown, but because of the data system used, the results are slightly different. The FACT data used is not equivalent to the data used in this thesis. For example, C 3A and C 2F are treated as stoichiometric compounds, while in this thesis, they are defined as solution species. Both databases lack thermodynamic data on solid solutions of

Ca 3SiO 5. MTDATA is a software program developed mainly for metallurgical applications. The database available to use with MTDATA is SGTE. SGTE contains pure substances and solutions of all types (NPL, 2013). HSC is another program that includes databases of thermodynamic data that is applicable to mineral processing, hydro-metallurgy and pyro-metallurgy. Thermodynamic data in HSC does not include any solution phase properties (Outotec, 2013). It is the aim of this work to expand and to evaluate the chosen database and, where not possible, to state the need for further expansion.

1.2.2. Types of process models There are differences between predictive models and empirical models that only describe a single specified process. The latter is not able to fully predict changes of the process parameters. Some empirical models describe only the relationship between the input variables and the output results, but they do not model the chemical and the physical mechanisms involved. Often, so- called hybrid models are used where fully predicting sub-models are combined with empirical models. The usefulness of such models is entirely dependent on how the different models are combined. Thus, they should be used with great care when significant deviations from the original process exist. Models found in the literature for cement and lime kiln applications are chemical or physical models or a combination of both. They can be either dynamic or static. Dynamic models are time dependent, describing the start- up or shut-down of a process or the changes that occur when there is some

type of disturbance in the process. A static or steady-state model will reveal how the process reacts to process changes in the long run. For example, a steady-state model determines the differences between two different scenarios in terms of the raw materials, the fuel changes or the changes in the operational conditions. Different tools are used to make the calculations in each model. For example, there are commercial software programs that couple the equations of mass and energy conversion. There are numerous developed models, but none of them have fully adequate thermodynamic data for cement clinker and lime applications. Furthermore, none of them have high-quality fuel definitions. Below are examples of models for cement clinker and lime production in rotary kilns. In the literature, there are many physical models with limited chemistry. Once the chemistry is simplified, the chemical reactions that are considered involve the oxide systems of the major elements (calcium, silicon, aluminium and iron), the clinker minerals C 2S (belite), C 3S (alite), C 3A

(calcium aluminate) and C 4AF (calcium aluminoferrite), a few components in the gas phase and possibly liquid H 2O if the drying stage is modelled. Physical models are focused on the bed height of the solid material, the flame characteristics, the kiln rotation speed, the geometric measures of the equipment and the temperature profiles. There are also physical models developed using the mathematical equations of mass and energy conversion (Georgallis et al., 2005, Locher, 2002, Marias and Bruyères 2009, Martins et al., 2002, Mujumdar et al., 2006, Schwertmann, 2004. Models using CFD (Computational Fluid Dynamics) for the prediction of flow patterns are used as described by Mastorakos et al. 1999 and Yuan et al. 2007 and calculate the mass, the energy and the momentum conservation. Mikulcic et al. , 2012 show how the modelling of a cement plant, achieved with the CFD code, can help CO 2 reducing measures by optimising calciners. Kääntee et al. , 2004, Pajarre and Penttilä 2004 and Yokota 2004 have developed chemical models. Kääntee et al. 2004 have developed a process model for the cement clinker kiln by simulating the use of alternative fuels. The model is a steady-state model developed in Aspen Plus™. A chemical model with more extensive and more appropriate thermodynamic data for cement kiln applications are described (Yokota, 2004). Data from FACT are used in the model, but it seems that the model is using adjustable parameters for the amount of condensed material in the gas phase, making the model non-predictive. Aspen Plus TM has been used to simulate and to optimise the function of a calciner (Zhang, 2011). The model only considers the calcination reaction and uses assumptions to account for NO x, SO x and CO x. Pajarre and Penttilä, 2004 presents the simulation tool KilnSimu™, which is a combination of calculations of reaction kinetics, heat transfer and multi-

component thermodynamic calculations with ChemApp™. The presented results are limited. The models presented in this thesis are a continuation of the model developed by Kääntee et al . 2004. By using thermodynamic data, the enthalpy balance for the whole process was achieved. To proceed with measures for sustainability in the cement clinker and lime production processes, models that predict changes in the fuel and the raw material feed are of particular interest. The parameters to be controlled are the cement clinker and lime quality, the temperature profile of the process, the flue gas composition and the CO 2 emissions. Changes that must be evaluated are: • Introduction of new types of fuels. • Introduction of new types of raw material. • Energy efficiency measures. • Introduction of oxy-fuel combustion.

It is important that the models developed are able to predict such changes and, particularly, to evaluate the reliability of the results and to suggest future improvements.

1.3. The cement and lime production processes Described briefly below are the production processes of cement clinker and lime, the typical process schemes, the specific features of the rotary kiln processes, the types of fuel used and the main chemical reactions taking place during the production.

1.3.1. The cement process Cement is produced in rotary kilns with few exceptions. As much as 99% of the worldwide cement production is grey cement (ordinary Portland and various composite cements), and only 1% is white cement, which is mainly used for niche architectural applications. Cement is used, nearly without exception, in concrete. The production of grey cement uses raw materials with 80% CaCO 3, and the rest is mainly oxides of silicon, aluminium and iron (for more details, see the section 1.3.5.1 below). Figure 3 shows a typical cement production process.

Figure 3 . Schematic process scheme of the production of cement (www.cementa.se ).

The raw material is blasted in the quarry, crushed, homogenised, ground in a raw meal mill and then stored in silos. The material that is approximately less than 100 Om after grinding is then fed into the preheating tower at the top of the cyclones where it starts to calcinate when it meets the hot air. At the bottom of the cyclone tower, almost all of the calcite is decomposed. In the rotary kiln, clinkering reactions occur. The maximum material temperature of 1450°C is reached at the lower end of the rotary kiln where the main burner is located. Cement clinker leaves the rotary kiln hot end and is cooled in a clinker cooler. To improve energy efficiency, in some processes, the cooling gas is drawn back to the preheating tower. A magnified picture of clinker is shown in Figure 4 . Flue gases leaving the preheating tower are stripped of dust in an electrostatic precipitator, and sulphur is removed in a scrubber. The cooled clinker is ground in a cement mill together with gypsum and other additives. The cement is stored in silos before it is delivered to the customer.

Figure 4. Light microscopy picture of a cement clinker etched in ammonium chloride. The magnification is 1000 times. The main minerals are alite (angular, hexagonal crystals blue and brown in colour) and belite (round and grey in colour). The interstitial phase is calcium aluminate and calcium aluminoferrite. (Photograph from Cementa Research).

Each cement plant is unique in its process layout, its raw material composition and the type of fuel that it uses. In this thesis, full-scale industrial data from two processes are used. The production for these plants is between 2000 to 5500 tonnes of clinker per day (Paper I , Paper IV and Paper VI ).

1.3.2. The lime process Lime is produced in both rotary and shaft kilns. Depending on the application of the lime and the physical properties of the limestone (geological origin, particle size distribution), a specific production process is selected. The main products are lime, ground or in lumps; PCC (precipitated calcium carbonate); and slaked lime. The raw material for the production of lime is limestone, CaCO 3. Limestone naturally contains other minerals, such as dolomite, CaMg(CO 3)2, clays (aluminium-silicon-hydroxyl minerals with

water), quartz (SiO 2) and iron oxides (Fe 2O3). A typical production process of lime in a rotary kiln is shown in Figure 5 . The process is similar to the cement manufacturing process, except that there is no preheating tower. The maximum temperature in the material bed is approximately 1000°C, following the decomposition to CaO. The heated air from the cooler is used in the rotary kiln. One rotary lime kiln process is used for the reference data collection in this work ( Paper II and Paper III ).

Figure 5 . Schematic process scheme of the production of lime. (www.britishlime.org)

1.3.3. Specific features of the rotary kiln processes Cement clinker and lime production processes are special in some aspects. The most important are the fact that: (i) it is a counter-current gas and a material flow process and (ii) the fuel is fed directly into the process. Consequently the fuel ash is incorporated in the products. In the counter-current flow process, the condensed material moves opposite to the gas flow. The material entering the kiln system meets the gas that has already passed through the entire system. The counter-current flow principle may lead to the enrichment of elements in the system, which may cause deposit build up and clogging problems. Condensed compounds of potassium and sodium have high and temperature-dependent vapour pressures making them relatively volatile in the hotter part of the system.

When carried back with the gas to the colder parts of the system, they condense, resulting in internal accumulation. When the concentration is high, condensed material builds up inside the cyclones, inside the duct or inside the kiln, causing operational problems. Fuel for a cement clinker or lime process, is introduced at the lower part of the rotary kiln. In the cement clinker kiln, fuel can also be fed at the riser duct just above the upper part of the kiln or in the calciners. The calciners are situated between riser ducts and the lowest cyclones. The combustible part of the fuel burns if conditions are favourable. The fuel ash, the inorganic part of the fuel, is entirely or partially incorporated into the product. This means that the fuel ash represents a part of the raw material feed and no fuel ash is withdrawn from the process as a by-product. Another important aspect when comparing cement clinker and lime production in rotary kilns is the difference in their particle size distribution. The raw material fed to a cement kiln is usually ground to less than 100 µm, and the raw material used for producing lump lime is 15 to 40 mm. This is of importance when considering the time dependence of the decomposition reactions.

1.3.4. Types of fuel Historically the only fuels used in cement clinker and lime manufacturing have originated from fossil fuels, namely, oil and coal. More recently, alternative fuels have presented an opportunity to make use of by-products from other production processes. Wastes can have a fossil or biomass origin. The use of renewable fuels and biomass fuels has just recently started in the cement and lime industry. In 2010, the amount of biomass was approximately 3% of total fuel usage in the cement industry, see Figure 6 .

Figure 6. Thermal energy by fuel origin for cement production. Excluded in the figure is the thermal fossil energy that sums up to 100%. (GNR Project reporting CO 2, 2012).

The CO 2 from a fuel can be considered as either fossil or non-fossil in origin depending on the portion of the fuel used in the process that can be regarded as biomass. Because very high temperatures are needed in cement clinker and lime burning, a substantial portion of the fuels used worldwide in these processes still consist of coal, oil or natural gas. Today, different qualities of waste fractions, such as used car tires and spent oils, are substituted for parts of the fossil fuels in Scandinavian cement plants. The biomass fuels are considered as CO 2-neutral. Alternative fuels might be partly CO 2-neutral if, for example, the RDF (Refused Derived Fuel) contains wood or used car tires contain natural rubber. Instead of regarding them as wastes, they should be seen as useful sources of energy in cement clinker and lime production. The effects of utilising biomasses compared with fossil fuels have not been fully studied. Some differences are a lower heating value, a higher amount of volatiles, and oxygen in the biomass compared with fossil fuels (Sami et al., 2001). The expected effects on the production process are influence on product quality and changed combustion flame characteristics. Several types of fuels used in the processes are modelled and described in this thesis: coal of different origins, petcoke, used rubber tires, meat and bone meal (MBM), light hydrocarbon liquid fuel (AC) and converted fuel oil (KEO).

1.3.5. Cement clinker and lime phase chemistry The chemical reactions that occur when producing clinker and lime are complex and not entirely understood in detail. The simple fact that most of the chemical reactions occur in a rotary kiln makes it difficult to make direct observations. One possibility to predict the cement clinker and lime composition is by using a simulation tool that enables complex multiphase chemical calculations and provides thermodynamic data of good quality that are applicable to cement clinker and lime production.

1.3.5.1. Major elements (Ca, Si, Al, Fe and Mg) High temperature chemistry in cement clinker and lime production mainly involves calcium (Ca), silicon (Si), aluminium (Al) and iron (Fe). A simplified picture of some reactions that occur during the production of cement clinker and lime is described below. The clinker minerals are, for example, not pure condensed phases but are rather a variety of polymorphs often involving solid solutions. Magnesium (Mg) is of great importance because it occurs as a natural part of limestone and sometimes as pure dolomite, CaMg(CO 3)2, in the raw meal. Mg has similar properties to Ca. Mg affects, for example, the formation of the melt by increasing the amount of belite, while decreasing the amount of alite (Altun, 1999). When heating raw material of a typical composition for the production of ordinary Portland cement (77.1% CaCO 3, 14.0% SiO 2, 4.1% Al 2O3, 1.6% Fe 2O3, other 3.2% (Hewlett, 2004)), the clinker formation reactions, as shown in Figure 7, appear. The relationships between the four main oxides (i.e., CaO,

SiO 2, Al 2O3 and Fe 2O3) are described in cement manufacturing by the lime saturation factor (LSF), the silica ratio (SR) and the alumina ratio (AM). These ratios determine the main properties of cement and certain process related characteristics. The ratios are calculated from the raw material composition; LSF = CaO/(2.80SiO 2+1.18Al 2O3+0.65Fe 2O3), SR = SiO 2/(Al 2O3+Fe 2O3), and AR = Al 2O3/Fe 2O3.

Figure 7. Schematic representation of the clinker formation reactions. Cr in the figure is cristobalite. (Javed et al. , 2004).

The raw meal for lime production usually contains more than 95% CaCO 3.

The remaining composition is contaminants, such as SiO 2, Fe 2O3, Al 2O3, K and Cl. The dominant reactions during production of cement clinker are the calcination of limestone and formation of belite and alite. The reaction with highest heat demand is the calcination of CaCO 3 and MgCO 3 ( Table 1 ). The lime reacts with the decomposition products of clay minerals (for example kaolinite, Al 4[(OH) 8Si 4O10 ] or AS 2H2 and pyrophyllite Al(OH)Si 2O5 or AS 4H and quartz to belite (C 2S). In the same temperature regime intermediate products C 2(A,F) and C 12 A7 are formed from quartz and the decomposed calcite and clay. Eventually, they form C 3A and C 4AF. At approximately

1300°C, a liquid (oxide melt) is formed that promotes the formation of C 3S.

The main minerals in the clinker are abbreviated to alite (C 3S), belite (C 2S), aluminate (C 3A) and ferrite (C 4AF).

The same chemical reactions occur for lime production as for cement production, but CaCO 3 is more abundant. The heating stops when all the material is calcined at approximately 1000°C; consequently, no alite is formed. Both endothermic and exothermic reactions occur when producing clinker and lime. The dominant endothermic reactions for the major elements are specified in Table 1 .

Table 1. Endothermic reactions in clinker and lime formation (Taylor, 1997 and Javed et al. , 2004).

Chemical reaction Temperature Enthalpy needed for changes the reactions (25°C) ∆H to occur (°C) (kJ/kg reactant)

H2O(liq) H 2O(g) 0-100 +2 446

CaCO 3 CaO+CO 2(g) 550-960 +1 782 MgCO 3 MgO+CO 2(g) 550-960 +1 395

AS 4H W-Al 2O3 + 2SiO 2 + 2H 2O(g) +224

AS 2H2 W-Al 2O3 + 2SiO 2 + 2H 2O(g) +538

2FeO(OH) W-Fe 2O3 +H 2O(g) +254 CaO + X-C 2S C 3S +59* Formation of liquid phase melt >1280 +600 *). Valid below 1430°C.

First, heat is needed to evaporate water and to decompose the calcium carbonates, the magnesium carbonates and the clays. The formation of intermediate products releases some heat. To progress further with the chemical reactions, energy is needed to melt and to form the final products, except for the formation of C 3S from CaO and SiO 2, which is exothermic. Formation of C 3S from -C 2S and CaO is endothermic below 1430°C and exothermic above 1430°C (Jöns, 1980). The dominant exothermic reaction is the combustion of fuels, see Table 2 .

Table 2. Exothermic reactions in clinker and lime formation (Taylor, 1997 and Javed et al. , 2004).

Chemical reaction Temperature Enthalpy needed for changes the reactions (25°C) ∆H to occur (°C) (kJ/kg reactant)

C + O 2 CO 2(g) -32 803 2 CaO + SiO 2 X-C 2S 600-1300 -734 3 CaO + SiO 2 C3S 1330-1450 -495

3 CaO + W-Al 2O3 C3A -27

6 CaO + 2W-Al 2O3 + W-Fe 2O3 C6A2F -157 4 CaO + W-Al 2O3 + W-Fe 2O3 C4AF -105

Formation of C 2F 800-1250 Formation of CA 800-1250

When the material reaches the desired maximum temperature of approximately 1450°C, the clinker is cooled to approximately 1200°C, where the liquid phase solidifies. Then, cooling proceeds rapidly down to approximately 100°C to prevent the alite from disintegrating into belite and free lime. Although the number of chemical reactions presented in Table 1 and 2 are fairly large, they are not sufficient to accurately describe the complex chemical phase reactions that occur. Using a global equilibrium approach with the Gibbs energy minimisation method, advanced thermodynamic data for all of the pure phases as well as the solutions can yield results that describe the process in detail. The gas entering into the process is ambient air preheated by the hot cement clinker and lime. The gas enters the process via the burner. When the fuel burns in preheated combustion air in the kiln, it reaches a temperature between 2000 and 2400°C. As the combustion gases and excess air travels along the kiln, the temperature is decreased to 1000°C at the upper end of the kiln while transferring heat to the kiln charge and loosing heat to the kiln refractories and to the shell of the kiln. In the preheating tower, the gas heats the incoming material leaving the kiln system at approximately 350°C. Additional investigations of the cement clinker formation in the high temperature region together with kinetic factors and discussions of solid- solid and solid-liquid phase formation in the CaO-SiO 2-Al 2O3-Fe 2O3 system have been previously described (Telschow, 2012 and Telschow et al. , 2012).

1.3.5.2. Minor elements (K, Na, S, Cl, F, P, Zn and Ti) Elements defined as minor have concentrations higher than 100 µm but are present at less than 1%. At temperatures above 1300°C, reactions involving the major elements occur, but below 1300°C, minor phases are of great importance. Important elements with minor concentration are potassium (K), sodium (Na), sulphur (S) and chlorine (Cl), as they form salts as chlorides, sulphates and carbonates. The salts readily combine to form low- melting salt mixtures causing deposits in the kiln system. Fluorine (F) is also of importance because it possesses similar properties to chlorine; thus, it is involved in forming similar compounds. The addition of CaF 2 results in formation of alite at lower temperatures and changes the clinker mineral composition (Altun, 1999). With the increased use of biomass fuels, for example, wood, MBM (Meat and Bone Meal), and saw dust, the input of phosphorous (P) in the process will increase. The topic of ash reactions in biomass co-firing in power production as well as the influence of P 2O5 on clinker chemistry has been studied (Punkte, 2004, Sandström, 2006, Nastac et al. , 2007, Stan[k and Sulovskð, 2009, and Poulsen et al., 2010). These findings are included in this thesis, as is the laboratory work that was undertaken with the XRD evaluations and modifications of thermodynamic data that were adapted to the phosphorous chemistry related to alite and especially to belite formation, as described in Paper V . It is important to take zinc (Zn) into account since it is high in concentration in some alternative fuels like used tires. Titanium (Ti) is of importance due to its solubility in perovskite phases (Hewlett, 2004) and in the oxide melt.

1.3.5.3. Fuel elements (C, H, O and N) The main components in the fuel are carbon, hydrogen, oxygen and nitrogen; see Table 3 for the composition of different fuels.

Table 3. Some properties of the fuels defined in the process models.

Fuel Coal I Petroleum Light Meat bone Converted Rubber Coal-II (dry matter) Coke fuel oil meal fuel oil tires (Petcoke) (AC) (MBM) (KEO) (Tires) C (%) 67.8 86.2 77.7 39.5 73.0 67.6 75.8 H (%) 4.3 3.1 10.4 5.3 10.9 7.0 5.0 O (%) 11.4 0.9 7.2 17.3 13.9 1.3 8.8 N (%) 1.2 1.7 0.4 7.5 0.1 0.4 1.2 S (%) 0.3 7.5 3.8 0.2 0.9 1.6 0.8 Cl (%) 0.001 0.004 0.48 0.25 0.1 0.1 - Ash (%) 15.0 0.6 - 30.0 1.1 21.9 8.4 HHV (MJ/kg) 31.7 35.1 31.2 18.2 43.0 33.2 32.9 Moisture (%) 9.7 6.8 5.0 6.5 - 1.6 4.7

1.4. Scope of the thesis The aim of the present work is to develop simulation models for cement clinker and lime production. The simulations models should be able to predict changes of interest. The evaluated changes are the effects on the process, on the product and on the external environment. The changes are the sustainability measures, such as the introduction of new types of fuels and new combustion technologies. The models should include process-specific properties for cement clinker and lime processes and calculate the chemistry correctly. It is the aim of this work to evaluate and to expand the thermodynamic database and where not possible, to state the need of further expansion. Finally, the goal of this thesis is to determine if developed phase chemical process models for cement clinker and lime production processes are reliable to use as predictive tools to understand the changes that occur when sustainability measures are introduced.

2. Methods used in developed process models

The simulation tool developed uses two software programs, Aspen Plus™ and ChemApp™, as well as thermodynamic data from FACT. Together with the software FactSage TM , it is possible to define data for different fuels and also to modify thermodynamic data.

2.1. Aspen Plus™ Aspen Plus TM is a process simulation program that allows for the prediction of process behaviour by using basic engineering relationships, such as mass and energy balances. With the program, a process simulation model is built and simulated. The process flow sheet of the developed models is drawn by the Aspen Plus™ interface. A number of unit operation models together with streams are selected to describe the process. In the models presented here, material and gas streams are utilised to connect unit operations, such as mixers, splitters, manipulators, cyclones, electrostatic precipitators and the ChemApp-user models. When the complete flow sheet has been designed the programme automatically selects tear streams. Tears are streams involved in the loops of the flow sheet. It is the tear streams that are required to converge. When calculating a flow sheet, a tolerance is set as a criterion for convergence. Then, calculations and an iteration procedure starts, and the tear streams are given an initial guess value. The next value on the tear stream is compared with the former. The calculations proceed until two consecutive guesses are within the specified tolerance. The tear variables are the total mass flow and all component flows, pressures and enthalpies. Equilibrium calculations in the models are performed in the ChemApp user blocks.

2.2. ChemApp™ ChemApp™ is a software that enables the calculation of chemical equilibria. It is also possible to integrate ChemApp TM in Aspen Plus™. The main reason for choosing ChemApp™ modules instead of the RGibbs-calculator available in Aspen Plus™ is that it allows for the use of thermodynamic data from FACT as well as allows for the usage of user-defined component data. To initialise the ChemApp™ calculations, thermodynamic data on all possible compounds, elemental amounts, pressures and temperatures are required. The equilibrium pressure, temperature or enthalpy changes, and equilibrium components are then calculated in the ChemApp-block.

2.3. Thermodynamic data from FACT Data required to calculate chemical equilibria are the standard enthalpy of formation (∆ fH° 298.15 ), the absolute standard entropy (S° 298.15 ) and the coefficients of the C p-polynomial heat capacity at a constant pressure (Cp) and at appropriate temperature intervals. Data used in these models are collected from FACT-database using the FactSage™ software (Bale et al., 2002). The thermodynamic data accompanying Aspen Plus™ is insufficient for cement clinker and lime applications and cannot, for example, handle solution phases. Data from FACT includes solution species, and it contains a vast database of oxides and is the most complete database relevant to the cement clinker and lime chemistry. The thermodynamic data in FACT is transparent and therefore possible to modify for the users.

2.4. Process schemes In Figure 8 , the process scheme of a cement plant is shown as it appears in the graphical user interface of Aspen Plus™. The process consists of a double string preheater of which only one is shown in the figure. Each string has one calciner and 5 cyclones in the preheating tower. The gas from the preheating tower is cooled and dedusted in an ESP (Electro Static Precipitator). A gas by-pass is capturing gas from the riser duct. The clinker is cooled with ambient air and the preheated air from the cooler is partially circulated back to the kiln via the calcinators. In Figure 8 , the three ChemApp-blocks to the left constitute the rotary kiln. The five boxes on the top right represent the cyclones. The two ChemApp-blocks below to the right constitute the riser duct and, the upper and lower part of the riser. The ChemApp-block to the left of the upper part of the riser duct is a calciner. In Figure 9 , a simplified process scheme of the process in Figure 8 is shown.

Figure 8. Process scheme of a part of the cement clinker manufacturing process in the Aspen Plus™ graphical user interface. Hexagonal symbols coloured in green are ChemApp-blocks.

Figure 9. Schematic process scheme of a cement clinker production process that has been modelled. The ChemApp-blocks (green coloured) calculate the equilibrium chemistry. Three blocks calculate the chemistry in the rotary kiln, two represent the riser duct and one represents the calciner.

2.5. Thermodynamic database developed for cement clinker and lime applications The thermodynamic database throughout the work performed in this thesis has grown along with a number of system components and phases. Some phases have been modified to better suit cement clinker and lime chemistry. In its latest and present form, it consists of the following:

• 1 gas phase with approximately 100 components • Stoichiometric condensed phases o Some pure liquids o Approximately 160 pure solids • 9 solution phases o 2 mixed melts o 7 solid solutions • 7 user defined fuel phases

The phases are based on 4 fuel elements, 5 major elements and 8 minor elements; see Table 4 .

Table 4. System components included in the latest developed model.

Group System components Fuel elements C H O N Major elements Ca Si Al Fe Mg Minor elements K Na S P Cl F Zn Ti

By carefully including or excluding components in the solution phases, the thermodynamic data can be adopted to suit certain applications. For each solution phase, there is a set of components included as a result of the progress of data optimisation. In Table 5 , the components included in each solution phase are listed and divided between major and minor elements. A good representation for predicting the chemistry in cement clinker and lime applications is achieved by selecting the solution phases and associated components.

Table 5. Solution phases with the associated components included in the database

Solution Components of major Components of minor phase elements elements

Oxide melt CaO SiO 2 Al 2O3 Fe 2O3 MgO TiO 2 Ti 2O3 ZnO K 2O Na 2O

FeO CaSO 4 Ca 3(PO 4)2 CaSO 4 Ca 3(PO 4)2

Salt melt CaSO 4 KCl NaCl NaF KCl Na 2SO 4

K2SO 4 Na 2CO 3 K 2CO 3 CaSO 4

W‘-Ca 2SiO 4 Ca 2SiO 4 Fe 2SiO 4 Ca 3(PO 4)2 Zn 2SiO 4 Ca 3(PO 4)2

W-Ca 2SiO 4 Ca 2SiO 4 Fe 2SiO 4 Ca 3(PO 4)2 Ca 3(PO 4)2

C2(A,F) Ca 2Fe 2O5 Ca 2Al 2O5

C3(A,F) Ca 3Fe 2O6 Ca 3Al 2O6

Perovskite Ca 2Ti 2O6 Ca 2Ti 2O5

ASC Na 2SO 4 K 2SO 4 Na 2CO 3 K 2CO 3

KCSC K 2SO 4 K 2CO 3 CaSO 4 CaCO 3

2.5.1. Fuel definition, including chemistry and enthalpy The existing fuel definition method in Aspen Plus™ uses both proximate analysis (weight-% of coal, ash and moisture) and ultimate analysis (weight- % carbon, hydrogen, nitrogen, sulphur, oxygen and chlorine). The ash content of the fuel is regarded as inert, which means that the ash does not react chemically. To consider the ash as inert in cement clinker and lime applications is a rough assumption because the ash is almost completely incorporated in the products. In fact, the fuel ash is a source of raw material, as it is mainly composed of calcium-, silicon-, aluminium- and iron-oxides. In this thesis, a procedure has been developed that takes into account the complete chemistry of the fuel as well as its energy content. The fuel definition treats the ash as an active part of the fuel. By understanding the complete chemical analysis (C, H, O, N, S and Cl and the ash composition) together with the calorimetric fuel value, it is

possible to define a component with the properties of the analysed fuel. By reproducing the chemical equilibrium reactions that occur in the calorimetric experiment with FactSage TM to determine the heating value, it is possible to determine the enthalpy of formation for each fuel. With the enthalpy of the products and the enthalpy of the reaction known, the enthalpy of the fuels is determined. The Cp of the fuel is assumed to be constant. The bomb calorimeter gives U. H = U + (P*V) and the volume is constant; therefore, (P*V) is zero. Gives that H= U, and H(reaction) = H(products)-H(reactants) = H(products)-H(fuel), it is possible to calculate H(fuel). With this definition in FactSage™, it is possible to add the different fuels in the process scheme of Aspen Plus™. For the moment, 7 fuels are defined; see Table 3 . The defined fuels are traditional fossil fuels, alternative fuels and wastes.

2.6. Phase chemical equilibrium calculations Equilibrium calculations are central in the models presented in this work. The equilibrium calculations provide information on the most stable state of the system. Equilibrium can be calculated using the Gibbs minimisation method which determines the lowest possible value of the Gibbs energy of a specified system (Moore, 1983). The equilibrium state is dynamic because that all reactions are reversible. Gibbs energy is a measure of the available energy of the system. The equation of total Gibbs energy includes the concentration and the molar Gibbs energy for each species and the phases present in the system. The minimum value is determined using an appropriate mathematical method. The pressure needs to be set as well as the temperature or the enthalpy changes for each equilibrium that needs to be calculated. To calculate the energy balance, the actual components in both the input and the output flows need to be known, whereas if only the composition of the equilibrium system is required, it is enough to define only the amount of each element. The calculation method to solve phase chemical equilibrium involves the solution of strongly non-linear equations, and it is performed with advanced programming. The equilibrium algorithm ChemSage™ is the precursor to ChemApp™ (Eriksson and Hack, 1990). Today ChemApp TM is the core program also in FactSage TM .

2.6.1. Non-equilibrium cooling according to Gulliver-Scheil The models are assumed to be in chemical equilibrium between and within phases. To calculate a relevant composition of the final product, a non- equilibrium calculation procedure is performed. The calculation procedure is

called Gulliver-Scheil and is described in detail in Paper II (Gulliver, 1913 and Scheil, 1942). The cooling of the products in the industrial process are rapid and thus not in equilibrium. By using the option Gulliver-Scheil cooling in FactSage™, rapid cooling is calculated. The calculation method allows for precipitated compounds from selected solution phases to be withdrawn from further equilibrium calculations. The oxide melt is chosen as the target phase for Scheil cooling. By introducing this calculation, a realistic cooling procedure will be predicted. The end product is the accumulated compounds that gradually have been withdrawn from the calculations. This procedure can only be used after a careful selection of solution species to include in the calculations.

2.6.2. Energy balance The equilibrium calculations are calculated by defining the pressure and the enthalpy changes of the system. Either the enthalpy changes or the equilibrium temperature can be set. With an enthalpy controlled model, it is possible to determine changes in the temperature due to the change of fuels, raw materials or combustion technologies. The equilibrium calculations in the models are enthalpy controlled. The enthalpy changes are calculated from industrial measurements of energy losses to the surroundings, namely, convection and radiation. The calculations of the energy losses are described in Paper II. The calculated results are added to the ChemApp-blocks included in the model and are iteratively changed to find reasonably temperatures in each process step. There is a discrepancy of the calculated energy losses and the ones added in the model, although this is not significant considering the whole picture. The energy balance for the whole process is calculated from the Gibbs equation with the thermodynamic data of all the included components and phases; the enthalpy of the fuels and the energy losses are defined in each ChemApp-block.

2.7. Strengths and weaknesses of the developed models There are many advantages to using chemical phase equilibrium models, but there are also some limitations accompanying them. Being aware of the limitations and using the models with this knowledge is necessary to use the models as effective tools in process modelling. The strengths of the developed models are that they are predictive with no adjustable parameters. The chemical sub-models included are based on scientific principals, meaning that no validity boundaries exist. However, there are both chemical and physical phenomena in the processes that need attention.

A chemical equilibrium model assumes that there is no time dependence (kinetics) for the reactions to proceed. The factors influencing kinetics are inhomogeneity and granulometry. Ideal mixing and isothermal conditions are assumed. Equilibrium is usually achieved if the temperature is high enough, if the retention time is long enough and if the mixing is complete. In the cement clinker process, the temperatures differ between the material (condensed phases) and the gas phase in some parts of the process; see Figure 10 . When the temperatures of all phases are equal, one can assume that equilibrium has been reached. This is the case in the preheater and at the maximum temperature of the material in the rotary kiln. The quality of the cement clinker is, to a high degree, established at the maximum temperature of the material, meaning that the assumption of chemical equilibrium is reasonable. If stoichiometric chemical reactions were chosen for the calculations instead, then all of the reactions would need to have been stated. This is not currently possible because they are not entirely known. If time factors and kinetics are used, then all chemical reactions and chemical components as well as their kinetic functions would need to be known. In fact, all of the chemical reactions in the cement clinker and lime production are not known, nor are there complete data on the kinetics. Thus, one advantage of using the Gibbs energy method for equilibrium calculations is that no reactions need to be stated.

Figure 10. Temperature of the gas and material in different parts of the process. (Cembureau, 1997).

As mentioned above, the counter-current gas and condensed material characteristics causes the enrichment of metals in the system. The accumulation cycle is reduced in the industrial process due to vapour condensation of fine particles moving with the gas. The calculated results of accumulation phenomena in the process model therefore show the worst case scenario.

3. Results and discussion

The results and corresponding discussions are divided into two parts. The first part concerns the validation of the simulation models as well as the validation of the chemical equilibrium calculations based on thermodynamic data development. Validation of the simulation models is achieved by comparing the output process data from the industrial plants with results from the simulation models. Input data are equivalent to the measured input data from the industrial process whilst in stable running conditions. Validation of the equilibrium calculations with their corresponding thermodynamic database is accomplished by comparing the calculated results with the literature, experimental data and other relevant calculation methods. The second part of the results concerns the simulation of different scenarios with the validated models. The simulated scenarios evaluate different measures for improving environmental sustainability.

3.1. Validation of simulation models The developed models have been verified by comparing the calculated results with full-scale industrial data from three production plants. Some parameters that have been compared are the flows of the material and gas, the flue gas and the product composition. The characteristics of the processes and more parameters are described below.

3.1.1. Cement clinker process models The first model developed in this thesis is described in Paper I . It simulates a cement clinker plant that produces 2000 tonnes of clinker per day. The model has a thermodynamic database composed of 11 system components, a gas phase with 59 constituents, 99 stoichiometric condensed phases consisting of 22 liquids and 77 solids. The fuels modelled are defined according to the routine developed in this work. The fuels used are coal, petcoke, MBM, AC, KEO and tires. The preheating portion of the system, consisting of a double cyclone string with four cyclones in each, is thoroughly modelled. In Figure 11 , the simulated amounts of clinker and raw material in the flue gas are slightly higher than those measured in the industrial process, most likely a result of the ideal conditions assumed in the model. The measured gas flow at the top cyclone is higher than the simulated gas flow, indicating that there is a lack of false air intake in the model. The raw material in the gas phase is a consequence of the separation efficiency of the upper cyclone. The raw material in the gas phase is not a measure of dust in the gas phases because it only originates from the incoming raw material.

Dust originates from several sources such as fuel ash, clinker, and condensate, which implies that the raw material from the upper cyclone is only a minor part of the total dust amount.

Figure 11. Amount of the clinker and raw material in the flue gas and gas volume that is measured from a full industrial-scale plant and in the simulated model.

Figure 12. Simulated results of the major components in the gas phase, from the burning zone to the top cyclone in the cement process model. N 2(g) is excluded.

The gas composition varies along the process. Figure 12 shows that the concentration of CO 2(g) is relatively constant throughout the entire process, 33œ36% in the cyclones and 21œ23% throughout the kiln. In the kiln inlet, there are high concentrations of KCl(g) and K 2Cl 2(g). These components are accumulated in this temperature interval due to an evaporation and a condensation cycle. The second cement process model developed is a 5-stage preheater kiln with calciners, and is more complex. The cement process has calciners and a duct connecting the air from the clinker cooler to the calciners and has more cyclones compared to the model described in Paper I . The production is approximately 5700 tonnes of clinker per day. Full scale industrial data from the plant was recorded during stable running conditions as well as during a full scale test. These were used to validate that the cement model is capable of evaluating process changes ( Paper IV ). Figure 9 shows an outline of the process.

The thermodynamic database is extended to include 16 system components. Following elements are added; Mg, Na, P, Zn and Ti. Phases that are added or increased with number of components are; the oxide melt

(CaO, SiO 2, Al 2O3, Fe 2O3, MgO, FeO, TiO 2, Ti 2O3, ZnO, K 2O, Na 2O, CaSO 4,

Ca 3(PO 4)2), the salt melt (KCl, NaCl, Na 2SO 4, K 2SO 4, Na 2CO 3, K 2CO 3 and CaSO 4), solid solutions of several polymorphs of Ca 2SiO 4, calcium aluminate ferrites, sodium and potassium salts and calcium titanium oxides. Results from the validation of a reference case are shown in Table 6 . Industrial process data from normal stable running conditions are compared with the simulation results of input values equivalent to industrial data. There is good agreement between the industrial and the simulated data regarding the amount of cement clinker produced, the temperature in the burning zone and the oxygen concentration in the kiln inlet. The oxygen amount in the top cyclone is lower than the actual amount. Leak air is most likely the cause for this finding. The simulated clinker composition is close to the industrial process data.

Table 6. Full-scale industrial reference data compared with the simulated data of a cement plant

Property Industrial Simulation process data data Clinker (tonnes/h) 223 227 Temperature in burn zone ( °C) 1370 1399

Concentration of O 2 (vol-%): Top cyclones 6.0 5.0 Kiln inlet 2.4 2.3 Chemical composition of clinker:

C3S (wt-%) 69.7 65.2 C2S (wt-%) 10.7 11.7

”C 4AF‘ (wt-%) 13.3 14.6 Free lime (wt-%) 0.8 0.6 Miscellaneous (wt-%) 5.5 7.9

”C4AF‘ for the industrial process data represents the amount of liquid phase from a microscopy determination of clinker phases, and for the simulated data, it represents the sum of the calcium aluminate ferrites.

3.1.2. Process model of a lime rotary kiln One lime plant has been used to collect data during stable running conditions and during changes in the process to validate the lime model (Paper II and Paper III ). The process layout of the lime plant, including the raw materials, the fuel intake, the air intake, the rotary kiln, the exhaust gas outlet and the product cooling, is shown in Figure 13 .

B8 AIRLICOL COLDLIME Exhaust gas MTRLBZ Lime B1

LIMEOUT

B4 FALSEAIR HOTAIRS1 Lime Kiln GASBZ

PRIMAIR BURNZONE LIMESH20 CALCINZ KILNIN GASKI HOTAIR LIMESTON Chemapp Chemapp COALH2O Chemapp Limestone PARCOAL B3 B5 MTRLCZ GASCZ MTRLKI Fuel FUEL

Figure 13 . Process scheme of the lime model in Aspen Plus™ with three ChemApp-blocks; burning zone, calcination zone and kiln inlet.

The main developments compared with earlier cement models are the introduction of solution phases in the thermodynamic database, the non- equilibrium cooling of the product according to Gulliver-Scheil and the energy losses calculation form the industrial process. The validation of the model includes the comparison of mass flows. The sum of the industrial measured condensed phases and the gas phase is 1.3% less than for the simulated values, which can be regarded as a good agreement. At this stage of the development of the thermodynamic database, 11 system components were included. Because Mg was not a component in the thermodynamic database but was present in limestone and the fuels, the Mg content in the limestone and the fuels were recalculated to equal Ca on molar basis. The solution phases included in the database were an oxide melt with

6 oxide species (CaO, SiO 2, Al 2O3, Fe 2O3, FeO and K 2O) and a salt melt with potassium species (KCl and K 2SO 4). At the maximum temperature of the process, the material consists of 12.7% oxide melt and 0.5% salt melt. By calculating the composition of the cooled product (with the Scheil cooling procedure), the final simulated product is compared to the industrial

produced lime product; see Table 7 . The industrial process data for the lime contains 3.2% less CaO than the simulated values. Some CaCO 3 is still present, most likely due to kinetic factors that inhibit its ability to totally convert into CaO. If totally converted, there would be 0.4% more CaO; 1.3% MgO corresponds to 1.8% CaO. Including these recalculations, the recalculated CaO content of the industrial product would be (81.6+0.4+1.8) 83.8%, which is close to the simulated Scheil cooled product. Recalculating to 1.1% of the simulated data, potassium phases occur, with no corresponding phase observed in the industrial produced product.

Table 7. The chemical composition of lime from the simulations of the equilibrium product, Scheil cooled product and industrial produced product

Compound Simulated Simulated Composition equilibrium Scheil of the composition Cooled industrial product produced product* CaO (wt-%) 84.8 84.8 81.6

CaCO 3 (wt-%) - - 0.7 MgO (wt-%) - - 1.3

Ca 2SiO 4 (wt-%) - 8.5 1.3 Ca 3Al 2O6 (wt-%) - 4.3 -

Ca 2Fe 2O5 (wt-%) 1.3 1.3 -

K2SO 4 (wt-%) - 0.6 - K2Ca 2(SO 4)3 0.7 0.4 - KCl (wt-%) - 0.1 - Oxide melt (wt-%) 12.7 - - Salt melt (wt-%) 0.5 - - Miscellaneous - - 15.1 *). Deviates from the results in Paper II due to error in calculations.

To control the model with enthalpy changes, the heat transfer mechanisms and heat flow needs to be known. For the lime process, measurements were undertaken to determine the heat losses to the surroundings. The heat losses were then defined in the process model, and the temperatures could be calculated at three different parts of the process, in the kiln inlet of the rotary kiln, calcination zone (middle of the rotary kiln) and burning zone. Energy losses to the surroundings are less than 10% of the enthalpy content in the exhaust gases.

3.1.3. Development of the thermodynamic database The thermodynamic database has gradually throughout the work been expanded with a number of system components and phases to a rather complete dataset suitable for cement clinker and lime production. The full dataset includes Ca, Si, Al, Fe, Mg, K, Na, S, P, Cl, F, Ti, Zn, C, H and O. Figure 14 shows the simulated results of a system of four components,

CaO, SiO 2, Al 2O3 and Fe 2O3 per 100 grams of material that is suitable as cement clinker raw meal. Figure 15 shows the results from the calculations with the full database per 100 grams of raw meal. Comparing the predicted results of the full system with the four component system reveals the following differences: there are more phases formed in the full system (MgO, salt melt, etc.), the formation of oxide melt starts at a lower temperature, the amount of oxide melt constitutes a larger portion, and so on.

Figure 14 . Amounts of the major phases in chemical equilibrium as a function of temperature. The input is 100 grams of a raw meal consisting of the four components (CaO, SiO 2, Al 2O3, Fe 2O3). The composition is suitable for cement clinker production. LSF is 92.5, SM is 2.9 and AM is 1.5.

Figure 15 . The amounts of the major phases in chemical equilibrium as a function of temperature. The input is 100 grams of a typical raw meal for cement clinker production with the system components Ca, Si, Al, Fe, Mg, K, Na, S, P, Cl, F, Zn, Ti, C, H and O. LSF is 92.5, SM is 2.9 and AM is 1.5. (The equilibrium composition at 1280 °C is a calculated average value of compositions at 1279 and 1281 °C. This is due to the condition that no oxide melt is formed at 1280 °C. This result is unlikely and the result of data that was not properly optimised).

3.1.3.1. Validation of the clinker composition The industrial production of cement includes the rapid cooling of the clinker. Rapid cooling implies that equilibrium conditions have not been reached. To replicate this cooling procedure, the Gulliver-Scheil method is used. The method is described in Paper II and Paper V .

The calculated results of cooling the four-component system is shown in Figure 16 and of cooling a complete system is shown in Figure 17 . Initially, the calculations yield Ca 3SiO 5(s), which is unchanged throughout the cooling calculations for both systems. The differences of cooling a four component system compared to a complete system are the amount and the retention of the oxide melt and the proportion of belite polymorphs. During cooling, the

oxide melt decreases while the phases W‘-C 2S, W-C 2S, C 2(A,F) and C 3(A,F) increases. Formation of C 3(A,F) and C 2(A,F) occur at higher temperatures in the four component system in comparison to the complete system. In the four component system, the oxide melt totally crystallised at 1300 °C. In the complete system, some oxide melt still remain at 700°C. According to Figure 15 , the oxide melt starts to form at 1200 °C during heating and should not be present during cooling after. The unrealistic existence of the oxide melt at 700 °C is likely due to non-optimised thermodynamic data for the sodium and potassium components in the oxide melt. The concentration of K 2O and Na 2O are enriched in the oxide melt because their total amounts are constant throughout the temperature interval from 1400 to 700 °C. The other components precipitated from the melt to form solids.

Figure 16 . The amounts of the major phases during Scheil cooling. The input is identical to the four components system (CaO, SiO 2, Al 2O3, Fe 2O3) used for the calculations in Fig. 14. LSF is 92.5, SM is 2.9 and AM is 1.5.

Figure 17 . The amounts of the major phases during Scheil cooling. The phases holding less than 1 g are excluded. The input is identical to the full system used for the calculations in Fig. 15. LSF is 92.5, SM is 2.9 and AM is 1.5.

The simulated equilibrium composition, the Scheil cooled simulated product and the products calculated from the raw material according to Bogue (Taylor, 1997) are shown in Figure 18 . A temperature of 1360 °C was chosen because at this temperature, the amount of C 3S is at its maximum. Bogue calculations only consider the main oxides of Ca, Si, Al and Fe. Thus, the methods are not completely comparative. In addition, Bogue calculations do not take into consideration the formation of the melt and its composition.

Figure 18 . The simulated equilibrium composition at 1360 °C, the composition of the products formed with Scheil cooling calculations and the composition of clinker calculated with Bogue. Bogue phases are calculated as a comparison (only the major phases are included).

C-A-F for the Scheil product and the Bogue calculated product represents the sum of all of the calcium aluminate ferrites. As shown in Figure 18 , the oxide melt in the equilibrium product primarily forms calcium aluminate ferrites and W‘-C 2S. The amount of the salt melt triples during cooling, as observed in the Scheil cooled product. A comparison of the results from the predicted Scheil cooled product and the Bogue calculations yields approximately 6% more Ca 3SiO 5(s) and 6% less C 2S. These results were expected because the Bogue calculation overestimates the amount of alite and underestimates the amount of belite (Taylor, 1997).

3.1.3.2. Influence of phosphorus (P) in cement clinker production The influence of P on the cement clinker chemistry was examined in laboratory produced cement clinkers. Industrial raw meal with the addition of phosphorus (as Ca 3P2O8) was used. The clinkers were analysed for their mineral composition with XRD, and the XRD patterns were semi- quantitatively analysed by the Rietveld method. The results are described in Figure 19 . Sample 0, is the reference sample and samples 1 to 9 have increasing amounts of Ca 3P2O8 added. The concentration of P 2O5 in sample 0, is 0.03%; in sample 1 it is 0.50%, and it increases gradually to 5.13% in sample 9. It is clearly shown that from sample 3 and higher, the C 3S phase decreases with increasing P-content in the clinkers. At the same time, W‘-

C2S(P) and CaO are increasing.

Figure 19 . The development of phases as a function of the increased P content from sample 0 to sample 9 for a typical cement clinker raw meal. The values are expressed as mole per 100 g.

A structure with the exchange of Si for P and Ca 2+ vacancies (Va) is suggested for the belite solid solution (Poulsen et. al., 2010). The chemical reactions

occurring when P is incorporated in the solid solution may be described by Equation 1 :

24Ca 3SiO 5+Ca 3(PO 4)2↔{4[(Ca 1.75 Va 0.25 )(Si 0.5 P0.5 )O 4]+22Ca 2SiO 4}(ss)+24CaO Eq. 1.

Considering the molar variation of C 3S, W‘C 2S(P) and CaO displayed in Figure 19 , these phases roughly correspond to the development of the stoichiometric relationship given in Equation 1 for samples 3 to 7.

By adopting the description of the C 2S-rich end of the C 2S-C 3P binary (Fix et al. , 1969) and the P 2O5 solubility observations in alites (Diouri et al. , 1997) a tentative outline of sub-solidus phase relations is described in Figure 20 .

Figure 20 . Pseudo-ternary sub-solidus phase diagram for C 2S-C 3P-CaO. The solid solutions are represented by bold blue lines. The red circles represent the compositional positions of sample 0 to 9 with the approximate subtraction of the free lime in the clinker.

Samples 0, 1 and 2 may reside in two-phase area I, which is characterised by

C3S and XC 2S(P) solid solutions. Samples 4 to 7 are located in three-phase area III, where C 3S, W‘C 2S(P) and CaO coexist. Samples 8 and 9 are located in two-phase area IV, where W‘C 2S(P) coexists with CaO. This schematic representation does not include any other clinker components besides SiO 2- CaO-P 2O5.

3.2. Simulated scenarios with validated models The developed models are used to simulate scenarios, such as varying fuel mixtures and introducing oxy-fuel combustion, which are of interest for process development for increased sustainability. The introduction of oxy- fuel combustion implies a change from a combustion oxidant containing 21 vol% O 2 to one containing almost 100% O 2. In turn, this will increase the CO 2 concentrations from approximately 25 to 75 vol% in the off-gas from the cement clinker and the lime processes. The effects of the altered composition of the oxidant gases on the fuel combustion characteristics, such as the adiabatic flame temperature, are of interest. Theoretical calculations can be made with FactSage TM . The adiabatic flame temperature of the combustion of coal is 2177 °C in air and 2935 °C in pure oxygen. Such a temperature increase influences the wear of the equipment, the formation of NO x, etc.

Increasing the CO 2 concentration in the ambient atmosphere will increase the calcination temperature, see Figure 21 . The theoretical calcination temperature is the equilibrium temperature at which the complete conversion of CaCO 3 to CaO is achieved at ambient atmosphere. At a CO 2 partial pressure (p(CO 2)) of 0.25, the calcination temperature is 831 °C, and at p(CO 2) = 0.75, the temperature is 852 °C. The increase is 22 °C, which is not drastic. The temperature measured in laboratory tests, obtained at p(CO 2) = 0, is 670 °C and increases to 900 °C at p(CO 2) = 1 (Zeman, 2009).

Figure 21. Theoretical equilibrium temperature for the total conversion of CaCO 3 to CaO, as a function of the partial pressure of CO2.

However, there are other issues regarding increased p(CO 2) in the cement clinker processes that must be evaluated before large-scale implementation is possible, such as the reduced velocity in cyclones due to the reduced quantity of gas and the mass and volumetric flow (Zeman, 2009). The models presented here give valuable information on the phase chemistry and energy efficiency for the evaluation of the introduction of oxy-fuel combustion into the cement clinker and lime production processes.

3.2.1. Oxygen-enriched combustion in cement clinker production A full-scale trial test with oxygen-enriched air at a cement plant aimed at increasing the production of clinker is simulated (Paper IV ). Oxygen was introduced at 1500 m 3/h into the burn zone. To reproduce the scenario and optimise the fuel addition in the calciners and the raw meal feed, two sets of process model simulations were performed. The objective of the first set was to simulate the demand for extra fuel in the calciners to reach the same oxygen level as in the reference case. As observed in Figure 22 , this was achieved at 108%.

Figure 22. Oxygen concentration in various parts of the cement clinker process at the reference level, with oxygen addition and with additional fuel feed into the calciners of 6, 8 and 10 wt-%.

With extra fuel fed into the calciner, more energy can be utilised to produce more cement clinker. To determine the level of extra raw material that it is possible to feed into the process, the following simulations were performed. The previously optimised simulation shown in Figure 22, which resulted from 108% fuel addition to the calciners and 1500 m3/h oxygen addition to the burn zone, is designated as the reference in Figure 23 . From this base scenario, the raw meal addition was increased up to a temperature in the burn zone, rendering a good-quality product, which implies that the temperature in the burn zone is maintained at a level similar to that in the reference case. The simulations show that it is possible to add 16% more raw meal with 108% fuel in the calciners and 1500 m 3/h oxygen addition, see Figure 23 . With 118 and 120% raw meal feed, the temperature decreases drastically in the burn zone. The drastic temperature drop is due to the large energy losses defined in the burn zone and the fact that the calcination reaction is highly endothermic. The equilibrium assumption predicts that all carbonate decomposes at a specific temperature, not gradually as occurs in a real rotary kiln.

Figure 23. Equilibrium temperatures in various parts of the cement process at the reference level including oxygen addition and 108% fuel feed in the calciners and raw meal feed rates of 100 œ 120%.

In Table 8 , the industrial process data on clinker production amount and energy efficiency are compared with simulated data. The clinker production increase shows good agreement, the industrial data gives 20% and the simulated data gives 17.5%. The energy consumption is based on fuel usage per tonne product produced. The energy needed to produce oxygen is not included. The industrial process data on energy consumption should therefore be beneficial for the oxygen enrichment test because more raw materials are added and the amount of added fuel is lower. This was not the case, which is most likely caused by an insufficient testing time, and limited knowledge of the expected effects of oxygen enrichment. The simulated data gives an energy decrease, which is the expected result.

Table 8. Full-scale industrial process data from a trial test compared with the simulated data for oxygen enrichment in the cement process

Property Industrial process Simulated data data change change Clinker (tonnes/h) +20% +17.5% Energy (kJ/kg clinker) +2% -11.8%

The quality of the cement clinker from the simulated reference case and that under oxygen-enriched air conditions are shown in Table 9 . It is clear that the oxygen-enriched air scenario gives similar or even better quality than the reference clinker. The amount of C 3S is 1.3% higher, and the amount of C 2S is 1.0% lower.

Table 9. Comparison of the simulation results of the reference cement clinker quality and the clinker quality achieved under oxygen-enriched air conditions (”Clinker oxygen-enriched air‘).

Clinker component Reference Clinker oxygen-enriched clinker air

C3S (wt-%) 65.2 66.5

C2S (wt-%) 11.7 10.7 ”C 4AF‘ (wt-%) 14.6 14.5 Free lime (wt-%) 0.6 0.6 MgO (wt-%) 2.5 2.5

K2SO 4+Na 2SO 4 (wt-%) 2.3 2.3

K2O+Na 2O (wt-%) 0.6 0.5

CaSO 4 (wt-%) 1.7 1.6

Ca 3Ti 2O7 (wt-%) 0.5 0.5 K2Ca 2(SO 4)3 (wt-%) 0.4 0.4

3.2.2. Influence of co-combustion on the lime production process and the lime product The validated lime model described in Paper II was used to simulate the influence of co-combustion of coal (Coal II in Table 3 ) with KEO on the temperature profile of the rotary kiln and on the lime product, see Figure 24 .

Figure 24. Influence of six different fuel mixtures of coal and KEO (converted fuel oil) on the temperature profile of the rotary kiln for lime production.

The reference scenario uses 100% coal as fuel. With an increasing amount of KEO and a decreasing amount of coal, the temperature rises in the calcination zone from 850 to 862 °C (1.4%) and at the kiln inlet from 400 to 412 °C (2.9%). The temperature in the burn zone is reduced by 4 °C or 0.4% under these conditions. The temperature changes are thus low and due to the altered chemical composition and the amount of condensed material.

Figure 25 . Influence of six different fuel mixes of coal and KEO on the chemical composition of the lime product.

In Figure 25 , the quality of the lime product is shown with increasing amount of KEO in the fuel mixture. The quality is defined as the amount of CaO in the product. The lime product quality is improved with an increased amount of KEO in the fuel mixture due to a decreased amount of ash in KEO.

3.2.3. Partial and full oxy-fuel combustion in cement clinker production Two different process layouts for the introduction of oxy-fuel combustion into cement clinker production were evaluated (see Paper VI ). The first process is full oxy-fuel combustion, as described in Figure 2b . With this option the flue gases are recirculated back to the burn zone thus affecting the entire process. The recirculation is needed to control the temperature in the burn zone.

Figure 26. Simulated temperature in various parts of the model for the reference case and for 20, 40, 60 and 80% recirculation of the flue gases from full oxy-fuel combustion of a cement clinker process. The oxidant is air for the reference scenario, and the oxidant is oxygen for the scenarios with 20 - 80 % recirculation.

The aim is to recirculate a minimum amount of gases. As shown in Figure 26 , at 60% recirculation of flue gases, the temperature in the burn zone is comparable to that for the reference case. The CO 2 concentration in the flue gases increased from 33% for the reference case to 76% at 60% recirculation.

If water is excluded, the CO 2 concentration is 90%. The product quality is influenced mainly by the temperature in the burn zone, see Figure 27 .

Figure 27. Composition of the product obtained from simulation of the reference case and 20, 40, 60 and 80% recirculation of flue gases. The results are from full oxy-fuel combustion of a cement clinker process. The oxidant is air for the reference scenario, and the oxidant is oxygen for the scenarios with 20 - 80 % recirculation.

The recirculation ratio must be 60% to reach the same level of Ca 3SiO 5(s) in the oxy-fuel combustion simulations as for the reference cases. Retrofitting the process to full oxy-fuel combustion requires extensive rebuilding and new piping. To reduce this burden, a partial oxy-fuel combustion scenario is modelled. The process layout is described in Figure 28 . The aim is to divide the process into two separate parts with regard to the gas circulation. The pre-heater tower, including two parallel cyclone strings, is divided into two separate parts: one part, the CO 2 string, will be fed with oxygen instead of air to exclude nitrogen from the gas phase. The

CO 2 string includes one cyclone string and both calciners. This makes possible the capture of combustion CO 2 from the calciners and process CO 2 from the raw material. Combustion CO 2 derived from the burn zone enters the N 2 string because all gases from the rotary kiln will leave the process via the cyclones in the N2 string.

Figure 28. Flow sheet of the process scheme in Aspen Plus™. The layout simulates partial oxy-fuel combustion by dividing the pre-heater tower into two parts; the CO2 string and the N2 string. The solid lines in the figure are material flows and the dashed lines are gas flows.

Figure 29. Energy distribution between the burn zone and the calciners for the reference case and for partial oxy-fuel combustion with 20 and 40% recirculation in the CO 2 cyclone string and the total energy input to the cement process. The reference scenario uses air as the oxidant. The 20 and 40% recirculation scenarios use oxygen as the oxidant.

Two scenarios are modelled; 20% and 40% recirculation of gases in the CO2 string. The aim is to have all raw material calcined before it enters the kiln. The gas stream that is not intended for capture is designated the N2 string. Therefore the energy added to the calciners is adjusted to reach full calcination at the desired location. The results of the energy distribution that give maximum CO 2 in the CO2 string from the raw material are showed in Figure 29 . The results for the flue gas composition in the two strings are described in Figure 30 .

Figure 30. Flue gas composition in the CO2 cyclone string and in the N2 string for the reference and the partial oxy-fuel combustion with 20 and 40% recirculation.

If all of the CO 2 in the CO2 string is captured and removed, the CO 2 emissions per tonne clinker will decrease from 716 kg CO 2 to 119 kg CO 2 for the 20% recirculation scenario.

Table 10. CO 2 emissions from the reference and the cases with 20% and 40% recirculation of the flue gases in the CO2 cyclone string

Reference 20% 40%

recirculation recirculation

Total CO 2 (tonnes/hour) 47.75 49.73 50.80

CO 2 in the CO 2 string - 41.82 42.63 (tonnes/hour)

CO 2 per tonne clinker (kg) 716 747 760

CO 2 per tonne clinker if CO 2 716 119 122 captured in the CO 2 cyclone string (kg)

3.2.4. Oxy-fuel combustion in lime production In the lime process model, oxy-fuel combustion is simulated with 50, 60, 70 and 80% recirculation of the flue gases (see Paper III) . The temperatures in the burn zone when recirculating at 50 - 80% of the flue gases are all higher (1229 - 1774 °C) than in the reference simulation (940 °C). To determine if it is possible to perform oxy-fuel combustion with less fuel, the same recirculation scenarios are simulated, for which the energy input is 90% of that in the reference case. With 90% energy input, the oxygen input is also reduced to 90%, and the temperatures for the simulation scenario with a 50 to 60% recirculation rate give the appropriate temperature in the burn zone as well as good quality products. The results of the major components in the gas phase for the flue gases are shown in Figure 31 .

Figure 31. Flue gas composition for the reference simulation and a 50 to 80% recirculation ratio at 1oo% energy and at 90% energy when simulating full oxy-fuel combustion in lime production.

The CO 2 concentration increases directly to 75.7% when recirculation is applied. With recirculation of flue gases, the intake of air is changed to pure oxygen, giving immediate results on the flue gas composition. At 90% energy input, all CO 2 concentrations are less than for simulations with 100% fuel, due to the lower quantity of fuel combusted. One identified risk with the recirculation of flue gases is the accumulation of harmful species such as corrosive chlorides. Evaluation of the gas composition of chlorides shows that with an increasing recirculation rate, the sum of HCl and KCl increases, and it is lower for 90% energy than for 100%. This result is expected because one source of chloride is the fuel. In Figure 32 , the amounts of HCl and KCl are shown for reference in all simulation scenarios.

Figure 32. Influence of the recirculation rate of flue gases and the amount of energy input on the amount of KCl and HCl in the flue gases.

The results show that the largest quantity of chlorides (KCl+HCl) is obtained in the simulation with the highest energy input and recirculation rate. With 90% energy input, the concentrations are very low, less than 0.00002 molfractions. The distribution of chloride between potassium (K) and hydrogen (H) varies, which can be partly attributed to the nonuniform temperatures of the flue gases. In Figure 33 , the amount of KCl in the flue gas partly follows the temperature curve. With higher temperatures, Cl combines with K in the gas phase.

Figure 33. Influence of the extent of recirculation and the amount of energy input on the distribution of chlorides. The secondary y-axis indicates the temperatures of the flue gases.

4. Conclusions

Several conclusions can be drawn based on (i) the process models development, (ii) full-scale industrial process measurements, (iii) chemical equilibrium predictions as well as (iv) laboratory experiments. The process models developed provide useful information on introducing sustainability measures in the cement clinker and lime production. The models predict the amounts of products (cement clinker and lime) and the gas flow with good accuracy. The prediction of the chemical composition of the gas and of the condensed phases in the process and the products are good for cement clinker and relatively good for lime production. The simulation results are significantly different for the four component system (Ca, Si, Al and Fe) than for the system also including Mg, K, Na, S, P, Cl, F, Ti, and Zn. The introduction of new types of fuels broadens the use of the process model and gives new and useful information on temperature and product quality in future processes with different fuel mixtures compared to the present. Energy-efficiency measures such as using oxygen-enriched air are introduced into the model and they are compared to industrial full-scale data. It is shown that introduction of oxy-fuel combustion is possible in both cement clinker and lime production processes. The quality of the products is maintained and high concentrations of CO 2 in the flue gases are achieved. The simulations show that it is possible to decrease fuel addition in lime processes. Partial oxy-fuel combustion in the cement process is possible with maintained cement clinker quality and a slight increase in energy consumption. Simulation of full oxy-fuel combustion in the lime process shows that at high recirculation rates the concentrations of KCl and HCl in the flue gases are likely to increase.

For future development of the model, the following improvements in the thermodynamic model are needed: • An optimised oxide melt phase with regards to K and Na. • Data on the solubility of minor elements (especially Mg, P, K, Na) in alite. • Data on solubility of minor elements in belite polymorphs. • Additional types of fuels defined.

Further, it is necessary to include predictive submodels for handling of the dust carried on with the gases in both the cement and lime process in order to account for potassium and sodium enrichment and transport to cyclones and electrostatic precipitator. These mechanisms are probably different for the two processes and additional process modules may be required.

5. Abbreviations

C = CaO

S = SiO 2 A = Al 2O3

F = Fe 2O3 M = MgO

K = K 2O

N = Na 2O

T = TiO 2

P = P 2O5 H = H 2O . C2S = (CaO) 2 SiO 2 = clinker mineral belite . C3S = (CaO) 3 SiO 2 = clinker mineral alite . C3A = (CaO) 3 Al 2O3 = clinker mineral aluminate . . C4AF = (CaO) 4 Al 2O3 Fe 2O3 = clinker mineral aluminoferrite GJ = giga Joule = 10 9 J= 1 000 000 000 J

Hf = enthalpy of formation S = entropy

Cp = specific heat capacity at constant pressure G = Gibbs energy P = Pressure V = Volume H = Enthalpy U = Internal Energy ChemApp TM FactSage TM Aspen Plus TM

6. References

2011/278/EU Commission decision 2011/278/EU. Official journal of the European Union . 17.5.2011.

Altun A I (1999) Effect of CaF 2 and MgO on sintering of cement clinker. Cement and Concrete Research 29 : 1847-1850.

Apen Plus, AspenTech, Burlington, Massachusetts, USA, www.aspentech.com

Bale C W, Chartrand S A, Degterov S A, Eriksson G, Hack K, Ben Mahfoud, R, Melancon J, Pelton A D, Petersen S (2002) FactSage Thermochemical Software and Databases Calphad 2 (26): 189-228.

Barker D J, Turner S A, Napier-Moore P A, Clark M, Davison J E (2009) CO2 Capture in the Cement Industry Energy Procedia 1: 87-94.

Barry T I, Glasser F P (2000) Calculations of Portland Cement Clinkering Reactions Advances in Cement Research , 1 (12): 19-28.

Cembureau (1997) Alternative fuels in cement manufacture, Technical and Environmental Review , Cembureau.

Cementa, www.cementa.se

Chemapp, GTT-Technologies, Herzogenrath, Germany. www.gtt-technologies.de

Damtoft J S, Lukasik J, Herfort D, Sorrention D, Gartmer E M (2008) Sustainable development and climate change initiatives. Cement and Concrete Research 38: 115-127.

Diouri A, Boukhari A, Aride J, Puertas F, Vazquez T (1997) Stable Ca 3SiO 5 solid solution containing manganese and phosphorus. Cement and Concrete Research 27(8): 1203-1212.

Eriksson G, Hack K (1990) A Computer Program for the Calculation of Complex Chemical Equilibra. Metallurgical Transactions B 21 (6): 1013- 1023.

FACT, www.factsage.com.

Factsage www.factsage.com.

Fix W, Heyman H, Heinke R (1969) Journal of American Ceramic Society œ Discussions and Notes 52(6): 346.

Georgallis M, Nowak P, Salcuedan M, Gartshore I S (2005) Modelling the rotary lime kiln. The Canadian Journal of Chemical Engineering 83: 212- 223.

Ghanbari Ahari K, Thompson D, Argent B B, Sharp JH (2004) Phase Equilibra Predictions by Computational Calculations on Portland Cement Clinker. Key Engineering Materials 264-268: 2131-2136.

GNR World Business Council for Sustainable Development (2012) The cement sustainability initiative: getting the numbers right, Cement Industry Energy and CO2. http://wbcsdcement.org/GNR-2011/world/GNR- Indicator_3211a-world.html (Accessed 15 December 2013).

Gulliver G H (1913) The Quantitative Effect of Rapid Cooling upon the Constitution of Binary Alloys. Journal of the Institute of Metals 9: 120-157.

Hewlett P C (2004) Lea‘s Chemistry of Cement and Concrete, 4 th Edition, 208-210.

Hoenig V, Hoppe H, Koring K, Lemke J (2012) ECRA CCS Report œ Report on Phase III. Technical Report, TR-ECRA-119/ http://www.ecra- online.org/fileadmin/redaktion/files/pdf/ECRA_Technical_Report_CCS_P hase_III.pdf (Accessed 16 December 2013).

IEA GHG, IEA Greenhouse Gas R&D Programme (2008) CO 2 Capture in the Cement Industry 3.

IEA WBCSB (2009) International Energy Agency World Business Council for Sustainable Development. Cement Technology Roadmap, Carbon Emissions Reductions up to 2050.

IEA (2013) Technology road map ”Carbon Capture and Storage.

IMF (2013) IMF World Economic Outlook 2013, April 2013.

IPCC, Integrated Pollution Prevention and Control (2009) Draft Reference Document on Best Available Techniques in the Cement, Lime and Magnesium Oxide Manufacturing Industries. European Commission.

Javed B, Miller I, MacGregor F, Kosmatka S H (2004) Innovations in Portland Cement Manufacturing , CD400, Portland Cement Association, Skokie, Illinois, U.S.A. ISBN: 0-89322-235-1

Jöns E S (1980) Thermodynamics driving force for formation of tricalciumsilicate. Cement and Concrete Research 1 (10): 103-106.

Kääntee U, Zevenhoven R, Backman R, Hupa M (2004) Cement Manufacturing Using Alternative Fuels and the Advantages of Process Modelling. Fuel Processing Technology 85 : 293-301.

Locher G (2002) Mathematical models for the cement clinker burning process œ Part 5: Complete plant. ZKG International 7 (55): 1-11.

Marias F, Bruyères B (2009) Modelling of biomass fired furnace for production of lime. Chemical Engineering Science . 64: 3417-3426.

Martins M A, Oliveira L S, Franca A, S (2002) Modelling and simulation of limestone calcination in rotary kilns. Part 1: Pilot kiln. ZKG International 4 (55): 76-87.

Mastorakos E, Massias A, Tsakiroglou C D, Goussis D A, Burganos V N, Payatakes A C (1999) CFD predictions for cement kilns including flame modelling, heat transfer and clinker chemistry. Applied Mathematical Modelling 23: 55-76.

Mikulcic H, Vujanovic M, Fidaros D K, Priesching P, Minic I, Tatschl R, Duic N, Stefanovic G (2012) The application of CFD modelling to support the reduction of CO 2 emissions in cement industry Energy 45: 464-473.

Moore J W (1983) Basic Physical Chemistry. Prentice/Hall International.

Mujumdar S K, Arora A, Ranade V V (2006) Modeling of Rotary Cement Kilns: Application to Reduction in Energy Consumption. Ind. Eng. Chem. Res. 45: 2315-2330.

Nastac DC, Kääntee U, Liimatainen J, Hupa M, Muntean M (2007). Influence of P(V) on the characteristics of calcium silicates and the hydration of clinkers. Advances in Cement Research. 19(3): 93-100.

NPL (2013) http://resource.npl.co.uk/mtdata/databases.htm . (Accessed 18 December 2013).

Outotec (2013) http://www.outotec.com/en/Products--services/HSC- Chemistry/Databases . Accessed 18 December 2013.

Pajarre R, Penttilä K (2004) KilnSimu-A ChemSheet based simulator for rotary kilns. GTT Annual Workshop . www.gtt-technologies.de .

Poulsen LS, Jakobsen HJ, Skibstedt J (2010) Incorporation of phosphorus guest ions in the calcium silicate phases of Portland cement from 31P MAS NMR Spectroscopy. Inorganic chemistry 49: 5522-5529.

Punkte, S., Auswirkungen des Phosphateintrages in Drehofenanlagen der Zementindustrie auf Klinkermineralogie und Zementeigenschaften. Doctoral Dissertation, Technischen Universität Clausthal, Germany, 2004.

Sami M, Annamalai K, Wooldridge M (2001) Co-firing of coal and biomass fuel blends. Progress in energy and combustion science (24): 171-214.

Sandström, M., Structural and Solid State EMF Studies of Phases in the

CaO-K 2O-P 2O5 System with Relevance for Biomass Combustion. Doctoral Dissertation, Umeå University, Umeå, Sweden, 2006.

Schwertmann T (2004) Thermodynamic aspects of the counterflow lime burning process. ZKG International . 9 (57): 64-77.

Scheil E Z (1942) Bemerkungen zur Schichtkristallbildung. Metallk 34: 70- 72.

Stan[k T, Sulovskð P (2009) The influence of phosphorous pentaoxide on the phase composition and formation of Portland clinker. Materials Characterization 60: 749-755.

Taylor H F W Cement chemistry 2 nd edition, 1997, Thomas Telfors, ISBN 0 7277 2592 0.

Telschow S (2012) Clinker burning kinetics and mechanism. Ph.D. Thesis. Technical University of Denmark (DTU).

Telschow S, Frandsen F, Theisen K, Dam-Johansen K (2012) Cement Formation œ A success story in a Black Box: High temperature phase formation of Portland cement clinker. Industrial & Engineering Chemistry Research 51: 10983-11004.

Yokota M (2004) Application of Aspen/Fact/ChemApp interface for clinker burning process simulation. Aspen Simulation World User Group 2004, Orlando, Florida, USA, 14 th October .

Yuan J, Darabi P, Salcudean M (2007) Modelling of flow, combustion, clinker formation and NOx emissions in long rotary cement kilns. ZKG International . 12 (60): 54-67.

Zeman F (2009) Oxygen combustion in cement production. Energy Procedia 1: 187-194.

Zhang Y, Cao S, Shao S, Chen Y, Liu S, Zhang S (2011) Aspen Plus-based simulation of a cement calciner and optimization analysis of air pollutants emission Clean Techn. Environ Polic . 13(3): 459œ468.