Physical and computational studies of slag behavior in an entrained flow gasifier

This work details an investigation of how to modify slag flow so as to maintain a clear line of sight across the reaction section of an entrained-flow coal gasifier. Physical and computational models were developed to study methods of diverting the molten slag that flows vertically down the walls of the reactor. The physical models employed silicone oil of varying viscosity. The computational models were developed using the Fluent software package. Based on the insight gained from the results of the models, two devices were created and tested in a pilot scale gasifier located at the University of Utah. The first method of slag diversion studied employed a gas jet to impact the slag film and cause it to flow around a sight port in the gasifier wall. By studying the film and jet interactions, it was discovered that the resulting behavior of such a system can be described by a dimensionless ratio of the kinetic energy of the jet and the surface energy of the film. The development of the dimensionless number, called a Lotte number in this work, is presented in detail. Generally, viscous films will be broken by a jet when the Lotte number is greater than 5 and will reclose when the Lotte number falls below a value of 1.5. The second slag diversion method studied used a round alumina tube protruding horizontally into the reaction section to break up the film. As the film impacts the tube, it progresses horizontally along the length of the tube before resuming the downward flow.

PHYSICAL AND COMPUTATIONAL STUDIES OF SLAG BEHAVIOR IN AN ENTRAINED FLOW GASIFIER by Randy Pummill A dissertation submitted to the faculty of The University of Utah in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Chemical Engineering The University of Utah August 2012Copyright © Randy Pummill 2012 All Rights ReservedThe University of Utah Graduate School STATEMENT OF DISSERTATION APPROVAL The dissertation of Randy Pummill has been approved by the following supervisory committee members: Kevin Whitty , Chair 6/8/2012 Date Approved Eric Eddings , Member 6/8/2012 Date Approved Thomas Fletcher , Member 6/11/2012 Date Approved Robert Stoll , Member 6/8/2012 Date Approved Constance Senior , Member 6/8/2012 Date Approved and by JoAnn Lighty , Chair of the Department of Chemical Engineering and by Charles A. Wight, Dean of The Graduate School. ABSTRACT This work details an investigation of how to modify slag flow so as to maintain a clear line of sight across the reaction section of an entrained-flow coal gasifier. Physical and computational models were developed to study methods of diverting the molten slag that flows vertically down the walls of the reactor. The physical models employed silicone oil of varying viscosity. The computational models were developed using the Fluent software package. Based on the insight gained from the results of the models, two devices were created and tested in a pilot scale gasifier located at the University of Utah. The first method of slag diversion studied employed a gas jet to impact the slag film and cause it to flow around a sight port in the gasifier wall. By studying the film and jet interactions, it was discovered that the resulting behavior of such a system can be described by a dimensionless ratio of the kinetic energy of the jet and the surface energy of the film. The development of the dimensionless number, called a Lotte number in this work, is presented in detail. Generally, viscous films will be broken by a jet when the Lotte number is greater than 5 and will reclose when the Lotte number falls below a value of 1.5. The second slag diversion method studied used a round alumina tube protruding horizontally into the reaction section to break up the film. As the film impacts the tube, it progresses horizontally along the length of the tube before resuming the downward flow. iv The models helped to establish how far the tube should protrude into the reactor in order to successfully break up the slag flow. Slag diversion devices were constructed and installed on a pilot scale gasifier. The jet diversion method was found to require an unreasonably large amount of purge gas to be successful and the metal jet suffered from the high temperature of the reactor despite the cooling effect of the gas. The tube diversion method worked very well for a series of experiments. However, erosion of the alumina tube in the reaction section remains an impediment to using such a device in an industrial setting. A design using a water-cooled tube is suggested.To Rachel, my wonderful partner and friend and our four terrible, funny, and delightful children CONTENTS ABSTRACT ……………………………………………………………………………. iii LIST OF FIGURES …………………………………………………………………….viii LIST OF TABLES …………………………………………………………………….... xi ACKNOWLEDGEMENTS ……………………………………………………………. xii Chapter 1 INTRODUCTION ………………………………………………………………. 1 1.1 Gasification ………………………………………………………………….. 1 1.2 Research Motivation ………………………………………………………… 4 2 REVIEW OF LITERATURE ………………………………………………….... 6 2.1 Slag Formation and Structure ………..……………………………………… 6 2.2 Temperature/Viscosity Relationships …………………………..…………… 9 2.3 Other Slag Property Studies ………………………………………...……… 23 2.4 Computational Modeling of Slag in a Reactor ………………………….….. 27 2.5 Jet Impinging on Thin Film Studies ………………………………….…….. 30 2.6 Summary ……………………………………………………......................... 31 3 OBJECTIVE AND APPROACH ……………...…………...………………...... 33 3.1 Objective…………………………………………………………..………... 33 3.2 Proposed Methods ………………………………………………..……….... 34 3.3 Approach …………………………………………………………………… 36 4 EXPERIMENTAL ……………………………………………….…………….. 38 4.1 Introduction .…………………………………………………..……………. 38 4.2 Silicone Oil Model .………………………………………………………… 38 4.3 Computational Model ………………………………………………………. 53 4.4 Gasifier Application …………………………………………….………..… 68 vii 5 RESULTS AND DISCUSSION ……………………………………………….. 81 5.1 Silicone Oil Experiments ……………………………………….…..................... 81 5.2 Computational Experiments …………………........……………...……….. 103 5.3 Gasifier Application ………………………………………………………. 125 6 SUMMARY AND CONCLUSIONS ………………………………………… 135 APPENDICES A: FALLING FILM THICKNESS CALCULATION ………………………. 139 B: MAXIMUM SHEAR RATE OF SILICONE OIL ……………………….. 145 C: SLAG VISCOSITY MODEL IN UDF FORM …………………………… 148 D: PENDANT DROP METHOD ……………………………………………. 149 E: GASIFIER FEED RATE EQUATION ………………………………..….. 153 REFERENCES ……………………………………………………………………….. 155 LIST OF FIGURES Figure 2.1 Illustration of temperature of critical viscosity …..…………………………….. 10 2.2 Glassy and crystalline slag viscosity versus temperature …...…………………. 12 3.1 Diagram of jet diversion method ………………………………………………. 35 4.1 Schematic of physical model setup ………………………………………….…. 40 4.2 Wall geometry for oil tests …………………………………………………….. 41 4.3 Jet impingement angle measurement …………………………………………... 42 4.4 Tube within elliptical slot …………………………………………………….... 43 4.5 Dimensions of refractory cast ………………………………………………….. 45 4.6 Structure of polydimethylsiloxane ……………………………………………... 47 4.7 Pendant droplet technique for measuring surface tension ……………………... 47 4.8 Silicone oil on impact with the tube …………………………………………… 51 4.9 Geometry for impinging jet computational model ……………………………... 55 4.10 Detail of hole and tube …………………………………………………………. 56 4.11 Top boundary arrangement of slag inlet boundary …………………………….. 57 4.12 Tube diversion computational model geometry ……………………………….. 59 4.13 Geometry wireframe front view ……………………………………………….. 61 4.14 Geometry wireframe isometric view …………………………………………... 62 4.15 Slag viscosity experimental data versus model predictions ……………………. 67 ix 4.16 Cut-away section of the gasifier ……………………………………………….. 70 4.17 Jet diversion device ………………………………………………………..…… 72 4.18 Optical access window design …………………………………………………. 73 4.19 Steel window holders with sapphire pieces ……………………………………. 75 4.20 Assembled alumina tube diversion device ……………………………………... 76 4.21 Retaining ring of tube diversion device ………………………………………... 76 4.22 Window assembly of tube diversion device ………………………………….... 77 4.23 Mounted assembly of tube diversion device …………………………………… 78 5.1 Low flow rate jet and silicone oil film …………………………………………. 82 5.2 Increased flow rate with larger oil bubble ……………………………………... 83 5.3 Increased flow rate with fluctuating bubble …………………………………… 84 5.4 Silicone oil hood formation ……………………………………………………. 85 5.5 Reformed oil bubble …………………………………………………………… 88 5.6 Jet mass flow rate data for film burst ………………...……………………….... 95 5.7 Lotte number versus impingement angle for film burst ………………...……… 96 5.8 Jet mass flow rate for film closure …………………………………………...… 98 5.9 Lotte number for film closure ………………………………………………..… 98 5.10 Jet flow rate versus film thickness for burst and closure ……………………... 100 5.11 Lotte number versus film thickness for burst and closure ……………………. 100 5.12 Tube length versus oil flow rate …………………………………….………… 102 5.13 Jet diversion model at low Lotte number ……………………………………... 104 5.14 Jet diversion for Lo = 8.0 ……………………………………………………... 106 5.15 Jet diversion results for high film flow rate, Lo = 8.0 ………………………... 107 x 5.16 Jet diversion results for high film flow rate, Lo = 13.5 ………………………. 108 5.17 Baseline simulation results at 250 time steps ………………………………… 111 5.18 Baseline simulation results at 500 time steps ………………………………… 113 5.19 Baseline simulation results at 500 time steps, side view ……………………... 114 5.20 Baseline simulation results at 750 time steps ………………………………… 115 5.21 Baseline simulation results at 750 time steps, side view ……………………... 116 5.22 Variable jet flow rate simulation results ……………………………………… 117 5.23 Variable jet temperature simulation results …………………………………... 119 5.24 First tube simulation results …………………………………………………... 122 5.25 Second tube simulation results ………………………………………………... 123 5.26 Third tube simulation results ………………………………………………..... 124 5.27 Jet diversion device after use in gasifier ……………………………………… 128 5.28 Sight port after jet diversion experiments …………………………………….. 129 5.29 Alumina tube and slag in the gasifier ………………………………………… 132 5.30 Cracked alumina tube in the gasifier …………………………………………. 133 5.31 Recovered alumina tube from gasifier experiments ………………………….. 135 A.1 Momentum fluxes …………………………………………………………….. 140 D.1 Surface tension measuring device …………………………………………….. 151 D.2 Droplet chamber with suspended droplet …………………………………….. 152LIST OF TABLES Table 2.1 Summary of mineral matter present in coal …..…………………………………. 7 2.2 Fitted values for bi ……………………………………………………………… 19 4.1 Silicone oil properties ....……………………………………………………….. 48 4.2 Ash composition in weight percentage of Sufco coal ………………………….. 64 4.3 Under-relaxation factors for computational model …………………………….. 68 4.5 Ultimate analysis of Skyline and Sufco coals ………………………………….. 79 4.6 Proximate analysis of Skyline and Sufco coals ………………………………... 79 4.7 Ash composition analysis of Skyline and Sufco coals …………………………. 80 5.1 Tube diversion results ………………………………………………………… 102 5.2 Conditions for jet diversion simulations ……………………………………… 110 5.3 Conditions for tube diversion simulations ……………………………………. 121 B.1 Critical shear rates for silicone oil ……………………………………………. 145 B.2 Calculated critical shear rates ………………………………………………… 145 B.3 Critical shear rates and maximum shear rates ………………………………… 147ACKNOWLEDGEMENTS As with any accomplishment, this work is truly the result of the efforts of many people. The author wishes to thank the many people who have lent their time and talents. Thanks to Eric Berg, Mike Burton, Gabe Hansen, David Ray Wagner, Travis Waind, and Dan Sweeney for their help with developing the physical model and in operating the gasifier. Thanks to Kevin Whitty and Dave Wagner for all their advice and guidance. It surely saved my life more than once. Thanks to Robin Hughes for helping me to understand that sometimes a simple solution is best. Thanks also to Jay Jeffries, Kai Sun, Ritobrata Sur, and Stanford University for the opportunity to work on this project. It was a pleasure. Thanks also to my committee for their advice and valuable feedback during this project. And finally, thanks to my wife and family for their patience and steadfast support. This work was supported by a U.S. Department of Energy Cooperative Agreement via Stanford University, DE-FE0001180. CHAPTER 1 INTRODUCTION 1.1 Gasification Coal gasification is a chemical process similar to combustion. In both processes, the coal is oxidized and gives off useful heat. During the combustion of coal, the carbon and hydrogen present in the coal reacts with oxygen from an air feed. The amount of oxygen fed is enough to completely react with all of the available carbon and hydrogen. This reaction produces carbon dioxide and water as the major products, as seen in Equation 1.1: (1.1) In contrast, gasification is usually carried out with a pure oxygen feed instead of an air feed. In gasification, the oxygen is fed in substoichiometric amounts so that the major product of the reaction is a mixture of hydrogen and carbon monoxide gas, as shown in Equation 1.2. This product gas is often called syngas, as it can be used to synthesize a great many other useful chemicals or used directly in gas turbine to produce electrical power. (1.2) 2 Carbon-based fuel gasification is a very old technology, but, despite its age, its use for electricity production is still relatively new. Only a small number of integrated gasification combined cycle (IGCC) power plants are operating today. Historically, the syngas product has had many uses. Coal gas or town gas was produced via coal gasification in the United States and Europe. The gas was primarily used for streetlights and home heating before the Second World War; its use died out as natural gas gained infrastructure and popularity (Everard, 1949). From the 1940s to the present, coal gasification has been used to create liquid fuels. In his review, Dry (1996) notes that this is done by converting the syngas to petroleum substitutes via Fischer-Tropsch reactions. More recently, coal gasification has been investigated as a possible power source via Integrated Gasification Combined-cycle (IGCC) systems. The thermodynamic cycles of an IGCC system are much more efficient than those of a standard coal combustion system. For example, Beér (2007) estimates that while a coal combustion-based, steam-generator electrical plant will have an efficiency rate of 32%, a gasification-based electrical plant is estimated to be about 47% efficient. Coal gasifiers operate at elevated temperatures. Steels are not able to withstand such excessive temperatures and so any gasification reactor must either be insulated or actively cooled while running. The reactor at the University of Utah Advanced Gasification Research Center that was used for this project relies on insulation, with thick pieces of refractory lining the interior of the reactor walls. The refractory is designed so that the reactor shell temperature is maintained below 530 K during operation. 3 Gasification also varies from combustion in the area of operating pressure. Most common combustion processes are performed at pressures just under that of the ambient environment. This ensures that the reactants and products will not leak out of the reaction areas. Gasification is generally performed at high pressures, between 30 and 70 bar. The high pressure allows gasification reactors to be compact when compared to combustion reactors of similar energy output, and increases process efficiency, particularly in the clean up and mercury removal phases. Beér (2007) also notes that the high outlet pressure also lends an overall efficiency advantage when carbon sequestration is considered. Overall, gasification offers many advantages over traditional coal combustion as an energy source but has yet to be fully embraced by utility companies. This reluctance is mostly due high capital and research costs and to uncertainties about how the process and facilities will fare in the long term. Most of the long term data on the gasification process has been collected by private companies and is proprietary. To bring this information into the public sphere and encourage adoption of the technology, the U.S. Department of Energy has made many efforts to fund research projects to increase the viability and attractiveness of coal gasification as a source of electrical energy production. One example is the TECO Polk Power Station located in Polk County, Florida, USA. This plant represents a joint effort between TECO and the U.S. government, which contributed $120 million to the project (TECO Energy 2012). The 250 MW plant began operation in 1997 uses an IGCC to maximize efficiency. It has been in full-scale, commercial operation since 2002 (McDaniel 2002). 4 Despite the efforts of the U.S. Department of Energy, industry as a whole has been slow to embrace gasification. In their report to the U.S. Department of Energy, Clayton et al. (2002) identified many possible areas for improvement to motivate large scale adoption of gasification technology. This list included better injector design, fuel flexibility, refractory wear and improvement studies, slag removal techniques, and instrumentation. Improving these areas will result in increased confidence in gasification technology and could lead to wider adoption of the technology. 1.2 Research Motivation Fast and reliable instrumentation has been identified as a critical need for gasification technology (Clayton et al. 2002). The aggressive environment inside a high-temperature, slagging gasification reactor can make getting reliable temperature and gas composition data from the reactor very difficult. Any physical probes used to obtain data are subject to the extremely high temperatures and the reducing environment of the reactor and will consequently have a very short lifetime. One way to overcome this limitation is to use a noninvasive measuring device such as an optically-based measurement system. Work by Sanders et al. (2001), Zhou et al. (2003), Jeffries (2009), and Ortwein et al. (2010) has shown that by directing a tunable laser through the hot gases inside the reactor and measuring the absorbance spectra at different wavelengths, data such as temperature and major species concentrations can be obtained. In order to function properly, such a measurement device would require a clear line of sight across the entire reactor. One impediment to an optically-based system is layer of slag on the reactor walls. As coal particles react, much of the mineral matter present in the coal remains as ash. 5 The amount of ash depends on the coal's origin but is often a significant portion of the original coal mass, generally making up more than 20% of the mass of the coal. In a gasification environment, the high temperatures cause the ash to become molten, which is then called slag. The molten slag will stick to the walls of the reactor and, over time, will accumulate and flow downward. This slag flow would likely flow over any sight ports cut into the reactor refractory and interfere with the line of sight necessary for the laser measurements. CHAPTER 2 REVIEW OF LITERATURE 2.1 Slag Formation and Structure Slag is crystalline mineral matter in either the molten or solid form and is the byproduct of many processes including coal combustion, coal gasification, and steel manufacturing. In solid form, slag is a brittle crystalline structure with dark coloration. Slag often contains entrapped bubbles of gas. These bubbles are entrapped while the slag is in molten liquid form and fixed in place as the slag cools. This gas generally consists of hydrogen, nitrogen, and carbon monoxide which are in high concentration during the gasification process. Slag consists of the residual mineral matter from the coal after the more volatile chemicals have been consumed. From the work done by Smith et al. (1994), of the standard Argonne coals analyzed, the mineral matter present ranged from 4.6% to 29.6% on a dry, sulfur free basis. This mineral matter predominantly consists of silicon, aluminum, calcium, and iron oxides in majority, and some trace amounts of other minerals and metals such as titanium, potassium, magnesium, sodium, and phosphorus. Table 2.1 gives information on mineral matter in various coals from the Argonne coal depository (Smith et al. 1994). As can be observed, the amount of mineral in coal 7 Table 2.1. Summary of composition of the mineral matter present in a wide range of coals. These data were taken from the Argonne coal samples and shows the wide range that mineral matter in coal can have. Min (%) Max (%) Avg (%) SiO2 21.5 60.8 43.3 Al2O3 13.5 36.0 22.6 Fe2O3 3.8 19.9 12.0 TiO2 1.0 2.6 1.66 P2O5 0.0 1.5 0.42 CaO 1.0 21.8 9.83 MgO 1.2 5.4 2.55 Na2O 0.0 8.3 1.39 K2O 0.2 3.7 1.59 SO3 0.0 25.7 5.23 Total MM 4.6 29.6 11.77 can vary greatly. Silicon is the most common element present in the mineral matter of coal and is followed by aluminum and iron. The mineral matter present in coal affects the slag properties in different ways. Song et al. (2010) divide the minerals into two groups, fluxing oxides, which include CaO, MgO, Na2O, K2O; and glass forming oxides, SiO2, Al2O3, Fe2O3, and TiO2. Fluxing oxides result in a lower viscosity slag and melting point and a lower temperature of critical viscosity. Glass forming oxides tend to have the opposite effect, increasing the viscosity of the slag at a given temperature and increasing the temperature at which the slag begins to flow. Most other works take a more complicated approach to cation categorization, based on the work of Zachariasen (1932). Zachariasen proposes that glass is made up of polyhedra of oxide anions surrounding a metal cation. The crystal lattice forms from the cations located at the corners of the polyhedral. Using x-ray crystallography, Zachariasen 8 was able to show the formation of the predicted three-dimensional structures in glass. Later, Sun (1947) extrapolated on Zachariasen's work by including consideration of the bond strength and setting out criteria for the polyhedral chain formation. He found that stronger bonds between an oxide and a cation created stronger lattices. Applying this to slag would mean that higher bond strength results in a more viscous slag. Also discovered was evidence that nonbridging oxygen, or NBOs, act as chain terminators within the crystalline lattice and weaken the overall structure of the glass. Mysen and Richet (2005) then determined that the ratio of NBOs to the silica present in the melt was a good indicator of the extent of depolymerization of the structure. Higher depolymerization results in a lower viscosity slag. Based on these works, other efforts to understand slag structure and behavior have the minerals divided into three groups: glass formers, modifiers, and amphoterics (Senior & Srinivasachar 1995; Urbain 1981). Glass formers contribute to the backbone or base of the structure. They are the cations with a coordination number of four or greater, and include Si4+, Ti4+, Ge4+ and P5+. Modifier ions, identified as Ca2+, Mg2+, Fe2+, Na+, V5+, Ba2+, Cr3+, Sr2+, and K+, have less network bonding potential and so cause the lattice to weaken. These cations create NBOs and act as chain terminators in the structure. This results in lower viscosity and melting point for the slag. Amphoterics, identified as B+, Fe3+, Zn2+, and Al3+, can act as either glass formers or modifiers, depending on the presence of other cations. If there is a high concentration of modifiers in the slag, the amphoterics can bond with the modifiers and the resulting molecule may act as a glass former resulting in a higher viscosity slag than would otherwise be expected. In the reverse case, if there are few modifiers, the influence of 9 amphoterics will be similar to that of the modifiers, weakening the glass network and lowering the viscosity of the slag. 2.2 Temperature/Viscosity Relationships Perhaps the most defining characteristic of slag is the temperature-viscosity relationship of the material. Many of the slag structure studies cited in the previous section were couched in terms of the effect that structure would have on a slag's viscosity. This was of great interest as the slag viscosity was a good predictor of potential fouling in combustion and gasification systems. One of the major goals of slag research is to develop a robust model to describe how slag behaves across a range of temperatures and slag elemental compositions. Reaching this goal is a complicated endeavor. Due to the structural issues discussed in the previous section, slag behavior can vary greatly depending on its composition. Another complication is the potential for a hysteresis effect in the temperature-viscosity behavior; the measured viscosity of the slag may vary depending on whether it is being heated or cooled. Since a large amount of slag is created during steel manufacturing, the steel and mining industries have been a primary driver in slag research. The literature has many studies on slag formation and behavior, some coming from as early as 1940. Reid, Nicholls, and Cohen worked on a special committee for the Bureau of Mines in 1940 and 1941 during which time they did much of the foundational studies on the subject. A high temperature viscometer had been developed by the Bureau of Mines in 1940 and this technology permitted the study of slag behavior that had not been previously possible. 10 Reid and Cohen established the idea of a temperature of critical viscosity, Tcv; the temperature above which the slag behaves approximately as a Newtonian fluid (Reid and Cohen 1944a). Refer to the illustration in Figure 2.1. The molten liquid form of slag behaves as a non-Newtonian fluid as it is heated until a certain temperature is reached, after which the viscosity/temperature relationship can be considered to be approximately Newtonian. The temperature when this transformation in behavior occurs is called the temperature of critical viscosity, or Tcv. This property also varies with the composition of the slag. Figure 2.1. Illustration of the concept of Tcv or temperature of critical viscosity. Slag behaves as a non-Newtonian fluid until the temperature of critical viscosity is reached. This occurs at an inflection point of the temperature-viscosity relation. Illustration adapted from figure found in Benyon (2002). 11 Slag itself can be generally classified as either being crystalline or glassy (also called vitreous or amorphous in some literature). Glassy slag generally has a lower Tcv and a shallower viscosity/temperature relationship; changes in temperature will cause a gradual change in viscosity. As a glassy slag cools, it maintains its homogeneity. No solids precipitate from the slag. Meanwhile, crystalline slag tends to have a higher Tcv and even small drop in temperature below the Tcv will raise the viscosity dramatically. This is due to crystals forming in the slag as it cools. The crystals precipitate from the slag and create a heterogeneous, two-phase mixture within the slag. The viscosity is greatly increased as the liquid phase interacts with the newly formed solid phase. Figure 2.2 shows a plot of glassy and crystalline slag viscosities versus temperature. Oh et al. (1995) theorize that this difference in behavior is due to the way the mineral crystals in the slag form. Glassy slag tends to form long, linear crystals when cooled. In contrast, the crystals formed in a crystalline slag tend to be dendritic, having many branches. The branches of the crystals interact and bond with each other, causing the slag to have greater viscosity and an increased melting point. Cohen and Reid (1940) developed the Ferric Percentage, which they used as a marker to identify crystalline slag, or slag of which the viscosity would increase greatly with a small drop in temperature. The equation they empirically developed for this was where all percentages are mass percents. They determined that any slag with a Ferric Percentage over 20 would be very high viscosity (crystalline) and that the amount of iron 12 Figure 2.2: Glassy and crystalline slag viscosity versus temperature. The crystalline slag has a much steeper viscosity curve and higher inflection point (around 1425 C) than the glassy slag (around 1325 C). in the slag became the main determinant of the slag viscosity. The first attempt to create an empirical equation to directly describe slag viscosity was done by Nicholls and Reid (1940). They created a concept called the Equivalent Silica Percentage. This was calculated by using the mass percents of SiO2, Fe2O3, CaO, and MgO in This percentage was then inserted into another empirical equation to get the viscosity at a given temperature. The correlation had good agreement with data from slag with a silica to aluminum oxide ratio between one and four over a narrow temperature band. 0 500 1000 1500 2000 2500 3000 1200 1300 1400 1500 1600 Viscosity (Poise) Temperature (C) PMB (Glassy) PMA (Crystalline)13 Later, the work of Cohen and Reid (1944) was used to create a prediction of the thickness of the slag layer on a furnace surface. Later still, they worked to develop empirical relations to describe the build-up of slag on tubes inside of a coal combustion furnace (Reid and Cohen 1944a). Cohen and Reid (1940) also developed a viscosity equation, basing it on a power function. The equation they created to describe their observations was ( ) where Z, and A were empirically derived constants, B is a composition-dependent constant, ηis the viscosity in poise and T is temperature in Kelvin (Reid and Cohen 1944b). Another noteworthy model of slag critical viscosity temperature was developed by Sage and McIlroy (1959). They took the innovative approach to divide the minerals present in slag into acidic or basic groups. The ratio of these two groups and the temperature of the slag were then inserted into an empirical equation to predict the slag viscosity. In this model, silicon, aluminum and titanium are considered acidic while iron, calcium, magnesium, sodium and potassium are considered to be basic. The basic minerals all only bond with 1 or ½ oxygen atoms (such as calcium or magnesium), which would theoretically contribute to a weaker crystal matrix within the slag. The acidic minerals each bond with 1 ½ or more oxygen atoms (such as iron, silicon, or titanium), thereby creating a stronger matrix as more possibilities to bond will result in a more 14 networked arrangement of molecules. The ratio of the mass percent of acidic and basic minerals (A/B) is inserted into, ( ) to get the Tcv value. Streeter et al. (1984) examined the accuracy of this equation by employing it on a large sample of low-rank coals. According to their work, after some modifications of the constants, the equation gives good results for coals with iron content of less than 10% by mass of the total mineral content and an acid/base ratio between one and seven. The next major milestone in the slag viscosity literature is the Watt and Fereday (1969) model. They had the goal of creating a model that would be better suited to a wide range of British coals. They were concerned that the previously developed models, which used American coals as their basis, would not accurately describe the behavior of wide-ranging British coals. The Watt-Fereday model focused on the effect of silica, aluminum oxide, calcium oxide, iron oxide. They analyzed the viscosity curves of 113 different coals and used a regression fit to create an equation similar to the Arrhenius equation in form: where m and c are constants based on composition, 15 where all oxides are evaluated as mass percents and the iron equivalent was determined as per eq 2.1. The Watt-Fereday model was determined to be very useful over a wide range of slag compositions. In previous models, calcium, magnesium, and iron were all considered to have equal fluxing efficiency. Through their research they discovered that not all fluxing minerals have an equal effect on the final slag viscosity. In 1981, Urbain set out to further improve viscosity predictions by applying the concepts of network formers and modifiers as found in glass formation theory. Urbain considered the work of both Weymann (1962) and Frenkel (1955) who had both done work on modeling silicate glass viscosities. The Weymann-Frenkel model for viscosity was similar to the Watt-Fereday model in that it used an Arrhenius-type equation, ( ) where T is temperature in Kelvin, η is the viscosity of the slag in units of poise and A and B are constants that are dependent on the slag composition. Urbain (1981) postulated that it should be possible to simplify the model by creating dependency between the A and B constants. By studying 60 different silicate melts, they determined that the two constants could be linked by 16 The group then determined B would be related to the mole fraction of silica in the melt, as per where N is the silica mole percent in the melt. The empirically determined values of the Bi values were influenced by the ratio of calcium oxide to the total of calcium oxide and aluminum(II) oxide, . For example, B0 was given by the parabolic equation Three additional equations were given for the other Bi constants in the literature but are not presented here. Although the Urbain model was fairly successful at predicting slag viscosities, it was really the simplicity and flexibility of the model that set it apart. The Urbain model is important in that many modern models are variations or modifications of it. Kalmanovitch and Frank (1990) tested several of the existing models and found that the Urbain model was the most accurate. They also refined the Urbain equations and expanded them to be able to consider slag containing iron, magnesium, sodium, potassium and titanium, which had previously been neglected in calculations. They modified the alpha value to be 17 rather than the simple ratio between calcium oxide and the sum of calcium oxide and aluminum(II) oxide. In another example of modifying the Urbain model, Senior and Srinivasachar (1995) used the Urbain model and expanded it to apply to slag at much higher viscosities. Previous models were only applicable to the relatively low viscosity range of 102-103 Pa·s, but the Senior-Srinivasachar model could be applied to the 104-108 Pa·s range. This was important as it helped to make better predictions of slag behavior at lower temperatures, before the Tcv of the slag had been reached. Much of the slag that sticks to furnace pipes and surfaces is below the Tcv, so this model could be directly applied to improve deposition models. The basic form of the Senior-Srinivasachar model is an empirical equation of the form ( ) where A and B are composition and temperature dependent variables. Each The B variable is a function of N, the mole fraction of SiO2 in the slag, and  which is defined as 18 where xi is the mole fraction of the indicated mineral. In order to add validity to their model at lower temperatures, Senior and Srinivasachar divided the viscosity-temperature into two parts at the Tcv point. The A and B variables each have two values, one for use at temperatures above the Tcv and the other for lower temperatures. The equation for the B value was set determined to be where each bi is a curve fit parameter. The final value for these constants is given in Table 2.2. The high and low-temperature B values were calculated to be 12.03 and 77.66 respectively. The A value is dependent on the B value and the ratio of non-binding oxygen to tetrahedral oxygen (NBO/T), The NBO/T value for the coal used was calculated to be 0.87. The equation for A in the high temperature region was determined to be 19 Table 2.2. Fitted values for bi Constant High Temperature Low Temperature b1 -224.98 -7563.46 b2 636.67 24431.69 b3 -418.70 32644.26 b4 823.89 -103681.0 b5 -2398.32 74541.33 b6 1650.56 -46484.8 b7 -957.94 146008.4 b8 3366.61 -104306.0 b9 -2551.71 21904.63 b10 -1722.24 -68194.8 b11 1432.08 48429.31 and the low-temperature equation for the related NBO/T value was The high and low-temperature A values were determined to be -8.70 and -69.94 respectively. 20 The Browning (BBHLW) model has taken a different approach to predicting slag viscosity. Browning et al. (2003) started with the assumption that at a given viscosity, the gradient of a slag temperature/viscosity curve is equal to any other at that same viscosity and that therefore all slag viscosity curves could be normalized to fit a standard curve. This is based on the observation of Nicholls and Reid (1940) that the change in viscosity with respect to temperature is related to the viscosity and not the temperature, or in equation form, To normalize the slag viscosity curves, they relate the viscosity to a modified temperature by subtracting from the measured temperature a temperature shift, labeled TS. The TS term is calculated from The TS value is dependent on composition, as the A term is calculated from The resulting equation is 21 ( ) The results of this unique approach had better accuracy than the Kalmanovitch/Urbain model, particularly for natural slag, but it was still limited to temperatures and viscosities above the Tcv of the slag. Another idea to improve model predictions was first proposed by Van Dyk et al. (2009). They postulated that a way to improve the model accuracy would be to take into account the formation of solids inside the slag as the formation of these solid crystals would result in a small change in the composition of the liquid phase. Van Dyk et al. employed the FactSageTM thermochemical database software package to determine any solidification. The composition of the predicted liquid phase was plugged into the KF model, developed by Kalmanovitch and Frank (1990). The results were an improvement in accuracy over the base model without using FactSageTM. Kondratiev and Jak (2001) also attempted to improve on other models by taking the solid phase into account. They postulated that as a slag crystallized, certain minerals were more likely to precipitate than others. Thus the remaining liquid slag may have a different composition than the overall slag composition. As Van Dyk et al. did, Kondratiev and Jak used the F*A*C*T* software program (later integrated into the FactSage software) to predict the leftover liquid composition in the slag after crystallization had been taken into account. For the viscosity predictions they employed a modified version of the Urbain equation. The α term was modified to include all modifiers and amphoterics as per 22 where the X terms represent the mole fractions of modifier or amphoterics in the slag. They developed a global B term to describe the interactions of all the modifiers and amphoterics at once, Σ ΣΣ( ) ( ) where XA, XC, XF, and XS represent the mole fractions of aluminum(II) oxide, calcium oxide, iron(II) oxide and silica. The b terms are empirical parameters that were used to fit the model to experimental data. By combining the modified Urbain model with the use of F*A*C*T* they were able greatly increase the accuracy of viscosity predictions over those of the original Urbain model. More recently, as computational power has expanded, attempts have been made to use artificial neural network models to describe the slag viscosity behavior, such as the one developed by Folkedahl (1997). Artificial neural networks are designed to act similarly to an organic brain. The network model has various input nodes that feed the data to a set of "hidden" nodes. The hidden nodes then interact to produce the output value, similar to neurons interacting to remember or interpret information. The input nodes in a slag viscosity model would be the temperature and composition and the output would be the fluid viscosity. Neural networks are more flexible than a purely empirical relation, but may be less accurate and must be trained with large volumes of information. 23 For example, Folkedahl used a data set of 18,000 points of slag composition, temperature, and viscosity data. The neural network he created gave results that better matched the experimental results than those predicted by the Urbain or Senior-Srinivasachar models over certain temperature ranges. Duchesne et al. (2010) also developed an artificial neural network model. Their model was created solely to predict the viscosity of Genesee coal, but over a wide range of temperatures. They compared the results of the neural network model with several other empirical correlations including the Watt-Fereday, Urbain, and S2 models. The neural network model outperformed all of the empirically derived models by a very significant margin. In some cases the empirical models mis-predicted by orders of magnitude, while the neural network model only had an absolute error of 5. Each model presented above has strengths and weaknesses and no model can be easily applied to all cases. Artificial neural network modeling is very interesting and can achieve very good results; however the amount of data and time required for training the model will be prohibitive to its wide use. For most cases, a modified Urbain or Weymann model will be sufficiently accurate. 2.3 Other Slag Property Studies Although the viscosity-temperature behavior of slag has been focused on the most by the research community, additional studies have been done to determine other properties of slag and silicate melts. As preliminary research into using slag as a building material, much as fly ash is used as an additive for concrete, Aineto et al. (2006) worked to determine the thermal expansion of coal slag. They found that slag has a relatively high thermal expansion compared to fly ash of the same particle diameter and determined 24 that this was due to the expansion and devolution of the trapped gasses inside the cooled slag particles. Bottinga and Weill (1970) studied the density of magmatic liquids. Assuming silicate slag to be an ideal solution, they found that using a simple summation of partial molar volumes of the minerals in the liquid yielded results as good as physical measurements with an average error of less than 8%. Their equation was Σ ̅ where Xi is the mole fraction of component i, Mi is the molecular weight of component i, and ̅ is the partial molar volume of component i. This work has been successfully applied to coal slag. Additional work identifying slag properties was done by Mills and Rhine (1989). Earlier, Grau and Masson (1976) theorized that in order to accurately calculate the density of a silicate melt, the intermolecular forces needed to be considered. However the model that they produced was not as successful at predicting density as the ideal solution model of Bottinga and Weill. Mills and Rhine (1989) built up the work done by Bottinga and Weill by taking into account the molecular interactions between different mineral oxides. This resulted in an empirical equation with adjustments made to the molar volumes of silica and aluminum oxide, which provided more accurate estimates of the melt density. Mills and Keene (1987) also developed a simple equation for the density of slag, 25 which was found to be accurate within 5%, despite its simplicity. Mills (1986) also provided augmentations to equations to calculate the surface tension and thermal expansion coefficients of slag. Up to that time, most equations to predict the surface tension of silicate melts were based on a simple summation of the contribution of the components, similar to eq 3.19, Σ where i is the surface tension of component i (Boni and Derge, 1956; Popel, 1962). Based on Mills' own previous work, the slag components were divided into surface active and surface inactive groups. The contribution of surface active components to the slag surface tension was taken into account with a set of parabolic equations that were inserted into the empirical equation. For example, the mole fraction of iron(II) oxide was adjusted as per Corrections were also offered for the oxides of sodium, potassium, phosphorous, chromium, and sulfur. This new, adjusted equation successfully resulted in more realistic results for surface tension calculations. 26 Methods of calculating heat capacity in slag have also been developed. According to Mills et al. (2002), the Kopp-Neumann Law gives an accurate estimation of the heat capacity of slag. This law states the heat capacity of a mixture is equal to the sum of the product of the mass fraction of each component multiplied by the heat capacity of the pure component, Σ where Cp,i is the heat capacity of component i. Gohil and Mills (1981) developed an empirical model to improve the accuracy of the results, but found little improvement over the simple Kopp-Neumann Law. Also found in the work of Mills and Rhine (1989) is an equation to describe the effective thermal conductivity of slag, where is the effective thermal conductivity, and aeff is the effective thermal diffusivity. Many viscosity tests for slag employ a crucible and spindle viscometer. By measuring the inner and outer temperatures, Mills and Rhine were able to determine the thermal diffusivity, which they found to be 4.5x10-7 m2/s. Zbogar et al. (2005) state that the conductivity should fall in the range of 0.6 ± 0.2 W/m K. Finally, Mills and Rhine also determined the slag emissivity was equal to They found the value to be independent of slag composition and determined 27 that it was constant for any temperature between 1070 K and 1800 K. Other work, such as that performed by Bohnes et al. (2005), found a strong composition and temperature dependence for the emissivity. Calcium, in particular, greatly increased the emissivity of the slag. 2.4 Computational Modeling of Slag in a Reactor Many groups have attempted to apply the previously discussed property models to reactor simulations. The goal of these simulations is to accurately predict slag build-up, general distribution, temperatures, and viscosity in order to minimize problems during reactor operation. For example, in operating gasifiers, slag must be removed from the reactor through a slag tap. In order to flow from the tap, the slag must be at a viscosity of 25 Pa·s or lower, otherwise the slag will fail to flow quickly enough and may plug the gasifier (Song et al. 2009). Such an occurrence would mean having to shut down the gasifier to remedy the problem. There have been several groups that have successfully created computational models of slag flow inside of a gasifier and their models will be discussed here. For each model, a general overview is given and then some attention is paid to the details of what viscosity model was used, how the Tcv was predicted and how in depth the treatment of energy transfer was. Seggiani (1998) created a model for a Prenflo gasifier to study what effect the addition of fluxing agents would have on the slag thickness. A Prenflow gasifier differs from an entrained-flow gasifier in that it is upward fired. Slag collects on the walls and flows downwards toward a slag tap where it is collected. In order to keep the slag at a viscosity low enough to be tapped, it is common to add limestone to act as a fluxant. 28 Seggiani's model solved the Navier-Stokes equations in three dimensions using the finite volumes method. To do this, he divided the gasifier into 15 volumes of different heights, with smaller volumes filling the top and bottom of the gasifier for increased accuracy in those zones. To predict the slag Tcv values, Seggiani used the model created by Sage and McIlroy, given previously in eq 3.6. For the temperature/viscosity relation, Seggiani chose to use the BCURA S2 model, discussed in a previous section and summarized in eqs 3.14 and 3.15. He states that the slag density and specific heat was calculated based on the work of Bottinga and Weill (see eq 2.1). These were then used to calculate the thermal conductivity, based on a thermal diffusivity of α = 4.5 x 10-7 m2 s-1, as reported by Mills & Rhine (1989). The emissivity of the slag was set to 0.83, also based on the work of Mills and Rhine. Seggiani included an ash deposition and slag build-up model. Apart from the needed mass and momentum balances, Seggiani's model also included an energy balance. This included radiation to and from the slag and included heat loss through refractory via conduction. Seggiani's model was important for many reasons. First, it was the first full simulation performed for slag flow in a gasifier, and it was very successful at predicting slag levels inside the gasifier with varying amount of added fluxant. Second, even though it was the first model of its kind, it was a complete three-dimensional model. Third, it was brought several other empirical models together and demonstrated that it was possible to couple these equations into a full reactor simulation. Seggiani's work continues to be studied and replicated even today and has been cited over 250 times. 29 Bockelie et al. (2001) modeled the slag flow in a downward-fired, entrained-flow gasifier as a submodel for a larger simulation. The larger simulation included fluid dynamics for the reactor, char gasification and devolatilization, and ash deposition on the reactor walls. For the slag flow submodel, the reactor walls were divided into two-dimensional vertical strips. For each strip, first a heat transfer calculation was made to determine the temperature distribution through the slag layer. From this, thickness at which the slag was at its Tcv was determined. Any slag cooler than the Tcv was assumed to be solid and immobile while the liquid slag was assumed to be at a uniform average temperature. The two-dimensional Navier-Stokes equation was solved for the liquid slag for each strip simultaneously. Fluid was allowed to move between the strips, however, as the reactor walls were clear of any impediments the fluid simply flowed downward to a great degree. These results for each strip were then patched together into a pseudo-three-dimensional result. The model was able to qualitatively predict slag thickness and to point to sections in the gasifier where the slag was prone to build up. The model by Bockelie et al. included an energy balance that accounted for heat loss through the refractory and allowed the slag to freeze into a solid mass. All heat transfer to the slag layers was assumed to only come from heat radiated from the flame. Convective heat from the hot gases inside the reactor was ignored, as were the convective losses from the slag flowing over the refractory or frozen slag. Temperature across the slag layer was calculated as a gradient. Bockelie et al. used the model developed by Patterson et al. (2001) to calculate the Tcv and viscosity of the slag. The Patterson model was developed to predict the 30 viscosity behavior of Australian coals, but was applied here due to similar ash composition. No information was given on how the density, specific heat, or thermal conductivity of the slag was calculated. Ni et al. (2010) created a model of a slagging gasifier. This model was significant for two reasons. First, they removed the assumption of an average temperature distribution in the slag layer, which had been used in many of the previous models. Second, they considered the convective heat transfer of the reaction gases to the slag. It is also noteworthy that the model was developed in the Ansys Fluent program. Ni et al. modeled both refractory-lined and actively cooled gasifiers. As with other models, slag below the Tcv was considered to be solid. Liquid slag temperature was not averaged over the entire layer, but was instead calculated for each cell. Temperature dependent properties were also calculated in this manner. The team used the Weymann viscosity model for their slag as it fit the compositions of the coals they were testing. Comparison of the slag deposition thickness as predicted by their model results to the results of real experiments showed very good agreement. 2.5 Jet Impinging on Thin Film Studies One component of the research performed in this study concerns the penetration of a gas jet through a falling viscous liquid film. Efforts were made to identify literature related to such studies. However, it appears that no studies on this phenomenon have ever been reported. There is a large amount of literature dedicated to the study of gas jets in cross flow (JICF); the phenomenon has been studied in great detail, but the work is 31 concerned with the behavior of the jet as it penetrates the fluid where the present research is concerned with the behavior of the fluid as it is penetrated by the gas jet. Banks and Chandrasekhara (1963) performed a study on surface deformation of a fluid when impacted by a gas jet. They found that the size of the cavity a jet would cause to form in a fluid surface is related to the fluid density and surface tension and the jet geometry and velocity. The concept of a Weber number is very prominent in the area of jet and fluid behaviors. The Weber number is the ratio of the kinetic energy of a droplet and its surface energy. Weber numbers are often used to quantify the behavior of falling droplets upon impact with a surface. The Weber number has been modified to be applicable to a droplet being struck by a jet stream. Tambe (2004) and Joseph et al. (1999) used such a modified Weber number to determine the conditions at which droplets would be broken up when exposed to a cross jet. Senecal (1999) and Nagabhushana Rao and Ramamurthi (2010) also modified the Weber number to apply the breakup of liquid sheets. They studied liquid sheets being broken up by parallel flowing gas jets. They found that to produce a particular droplet size, the gas jet velocity had to be adjusted so that the resulting Weber number was equal to 27/16. 2.6 Summary For the current study, information regardinFor the current study, information regardin For the current study, information regardin For the current study, information regardin For the current study, information regardin For the current study, information regardin For the current study, information regardin For the current study, information regardinFor the current study, information regardin For the current study, information regardin For the current study, information regardin For the current study, information regardinFor the current study, information regardinFor the current study, information regardin g coal slag properties is very g coal slag properties is very g coal slag properties is very g coal slag properties is very g coal slag properties is very g coal slag properties is very g coal slag properties is very g coal slag properties is very g coal slag properties is very g coal slag properties is very g coal slag properties is very g coal slag properties is very g coal slag properties is very important. Particularly, the viscosity models were very carefully considered. Due to important. Particularly, the viscosity models were very carefully considered. Due to important. Particularly, the viscosity models were very carefully considered. Due to important. Particularly, the viscosity models were very carefully considered. Due to important. Particularly, the viscosity models were very carefully considered. Due to important. Particularly, the viscosity models were very carefully considered. Due to important. Particularly, the viscosity models were very carefully considered. Due to important. Particularly, the viscosity models were very carefully considered. Due to important. Particularly, the viscosity models were very carefully considered. Due to important. Particularly, the viscosity models were very carefully considered. Due to important. Particularly, the viscosity models were very carefully considered. Due to important. Particularly, the viscosity models were very carefully considered. Due to important. Particularly, the viscosity models were very carefully considered. Due to important. Particularly, the viscosity models were very carefully considered. Due to important. Particularly, the viscosity models were very carefully considered. Due to important. Particularly, the viscosity models were very carefully considered. Due to important. Particularly, the viscosity models were very carefully considered. Due to important. Particularly, the viscosity models were very carefully considered. Due to important. Particularly, the viscosity models were very carefully considered. Due to important. Particularly, the viscosity models were very carefully considered. Due to important. Particularly, the viscosity models were very carefully considered. Due to important. Particularly, the viscosity models were very carefully considered. Due to important. Particularly, the viscosity models were very carefully considered. Due to important. Particularly, the viscosity models were very carefully considered. Due to important. Particularly, the viscosity models were very carefully considered. Due to important. Particularly, the viscosity models were very carefully considered. Due to important. Particularly, the viscosity models were very carefully considered. Due to important. Particularly, the viscosity models were very carefully considered. Due to important. Particularly, the viscosity models were very carefully considered. Due to important. Particularly, the viscosity models were very carefully considered. Due to ability of the Seniorability of the Senior ability of the Senior ability of the Seniorability of the Senior ability of the Senior ability of the Seniorability of the Senior -Srinivasachar model to predict slag viscosities below the TSrinivasachar model to predict slag viscosities below the T Srinivasachar model to predict slag viscosities below the T Srinivasachar model to predict slag viscosities below the T Srinivasachar model to predict slag viscosities below the TSrinivasachar model to predict slag viscosities below the T Srinivasachar model to predict slag viscosities below the T Srinivasachar model to predict slag viscosities below the T Srinivasachar model to predict slag viscosities below the T Srinivasachar model to predict slag viscosities below the T Srinivasachar model to predict slag viscosities below the T Srinivasachar model to predict slag viscosities below the TSrinivasachar model to predict slag viscosities below the TSrinivasachar model to predict slag viscosities below the T Srinivasachar model to predict slag viscosities below the T Srinivasachar model to predict slag viscosities below the T Srinivasachar model to predict slag viscosities below the T Srinivasachar model to predict slag viscosities below the TSrinivasachar model to predict slag viscosities below the TCV , this , this 32 model was applied to the computational model was applied to the computational model was applied to the computational model was applied to the computational modelmodel was applied to the computational model model was applied to the computational model was applied to the computational model was applied to the computational model was applied to the computational modelmodel was applied to the computational model model was applied to the computational model was applied to the computational model was applied to the computational model was applied to the computational models. The empirical correlations for other slag s. The empirical correlations for other slag s. The empirical correlations for other slag s. The empirical correlations for other slag s. The empirical correlations for other slag s. The empirical correlations for other slag s. The empirical correlations for other slag s. The empirical correlations for other slag s. The empirical correlations for other slag s. The empirical correlations for other slag s. The empirical correlations for other slag s. The empirical correlations for other slag s. The empirical correlations for other slag s. The empirical correlations for other slag s. The empirical correlations for other slag s. The empirical correlations for other slag s. The empirical correlations for other slag properties such as density and heat capacity are also very important to this work properties such as density and heat capacity are also very important to this work properties such as density and heat capacity are also very important to this work properties such as density and heat capacity are also very important to this work properties such as density and heat capacity are also very important to this work properties such as density and heat capacity are also very important to this work properties such as density and heat capacity are also very important to this work properties such as density and heat capacity are also very important to this work properties such as density and heat capacity are also very important to this work properties such as density and heat capacity are also very important to this work properties such as density and heat capacity are also very important to this work properties such as density and heat capacity are also very important to this work properties such as density and heat capacity are also very important to this work properties such as density and heat capacity are also very important to this work properties such as density and heat capacity are also very important to this work properties such as density and heat capacity are also very important to this work properties such as density and heat capacity are also very important to this work properties such as density and heat capacity are also very important to this work properties such as density and heat capacity are also very important to this work properties such as density and heat capacity are also very important to this work properties such as density and heat capacity are also very important to this work properties such as density and heat capacity are also very important to this work properties such as density and heat capacity are also very important to this work properties such as density and heat capacity are also very important to this work properties such as density and heat capacity are also very important to this work properties such as density and heat capacity are also very important to this work properties such as density and heat capacity are also very important to this work properties such as density and heat capacity are also very important to this work properties such as density and heat capacity are also very important to this work properties such as density and heat capacity are also very important to this work properties such as density and heat capacity are also very important to this work properties such as density and heat capacity are also very important to this work several of them were used to estimate the slag properties for model. several of them were used to estimate the slag properties for model. several of them were used to estimate the slag properties for model. several of them were used to estimate the slag properties for model. several of them were used to estimate the slag properties for model.several of them were used to estimate the slag properties for model. several of them were used to estimate the slag properties for model.several of them were used to estimate the slag properties for model.several of them were used to estimate the slag properties for model. several of them were used to estimate the slag properties for model. several of them were used to estimate the slag properties for model. several of them were used to estimate the slag properties for model. several of them were used to estimate the slag properties for model. several of them were used to estimate the slag properties for model. several of them were used to estimate the slag properties for model. several of them were used to estimate the slag properties for model. several of them were used to estimate the slag properties for model. several of them were used to estimate the slag properties for model. several of them were used to estimate the slag properties for model.several of them were used to estimate the slag properties for model. The concept of a Weber number, while not directly The concept of a Weber number, while not directly The concept of a Weber number, while not directly The concept of a Weber number, while not directly aThe concept of a Weber number, while not directly The concept of a Weber number, while not directly aThe concept of a Weber number, while not directly aThe concept of a Weber number, while not directly aThe concept of a Weber number, while not directly aThe concept of a Weber number, while not directly aThe concept of a Weber number, while not directly The concept of a Weber number, while not directly The concept of a Weber number, while not directly The concept of a Weber number, while not directly aThe concept of a Weber number, while not directly The concept of a Weber number, while not directly aThe concept of a Weber number, while not directly aThe concept of a Weber number, while not directly applicable to this work, did pplicable to this work, did pplicable to this work, did pplicable to this work, did pplicable to this work, did pplicable to this work, did pplicable to this work, did greatly influence the dimensional analysis of physical model. The work done by greatly influence the dimensional analysis of physical model. The work done by greatly influence the dimensional analysis of physical model. The work done by greatly influence the dimensional analysis of physical model. The work done by greatly influence the dimensional analysis of physical model. The work done by greatly influence the dimensional analysis of physical model. The work done by greatly influence the dimensional analysis of physical model. The work done by greatly influence the dimensional analysis of physical model. The work done by greatly influence the dimensional analysis of physical model. The work done by greatly influence the dimensional analysis of physical model. The work done by greatly influence the dimensional analysis of physical model. The work done by greatly influence the dimensional analysis of physical model. The work done by greatly influence the dimensional analysis of physical model. The work done by greatly influence the dimensional analysis of physical model. The work done by greatly influence the dimensional analysis of physical model. The work done by greatly influence the dimensional analysis of physical model. The work done by greatly influence the dimensional analysis of physical model. The work done by greatly influence the dimensional analysis of physical model. The work done by greatly influence the dimensional analysis of physical model. The work done by greatly influence the dimensional analysis of physical model. The work done by greatly influence the dimensional analysis of physical model. The work done by greatly influence the dimensional analysis of physical model. The work done by greatly influence the dimensional analysis of physical model. The work done by greatly influence the dimensional analysis of physical model. The work done by greatly influence the dimensional analysis of physical model. The work done by greatly influence the dimensional analysis of physical model. The work done by greatly influence the dimensional analysis of physical model. The work done by greatly influence the dimensional analysis of physical model. The work done by greatly influence the dimensional analysis of physical model. The work done by previous authors on the Weber number also helped to clarify which variables would be previous authors on the Weber number also helped to clarify which variables would be previous authors on the Weber number also helped to clarify which variables would be previous authors on the Weber number also helped to clarify which variables would be previous authors on the Weber number also helped to clarify which variables would be previous authors on the Weber number also helped to clarify which variables would be previous authors on the Weber number also helped to clarify which variables would be previous authors on the Weber number also helped to clarify which variables would be previous authors on the Weber number also helped to clarify which variables would be previous authors on the Weber number also helped to clarify which variables would be previous authors on the Weber number also helped to clarify which variables would be previous authors on the Weber number also helped to clarify which variables would be previous authors on the Weber number also helped to clarify which variables would be previous authors on the Weber number also helped to clarify which variables would be previous authors on the Weber number also helped to clarify which variables would be previous authors on the Weber number also helped to clarify which variables would be previous authors on the Weber number also helped to clarify which variables would be previous authors on the Weber number also helped to clarify which variables would be previous authors on the Weber number also helped to clarify which variables would be previous authors on the Weber number also helped to clarify which variables would be previous authors on the Weber number also helped to clarify which variables would be previous authors on the Weber number also helped to clarify which variables would be previous authors on the Weber number also helped to clarify which variables would be the most important to consider in physical model. the most important to consider in physical model. the most important to consider in physical model. the most important to consider in physical model. the most important to consider in physical model. the most important to consider in physical model. the most important to consider in physical model. the most important to consider in physical model. the most important to consider in physical model. the most important to consider in physical model. the most important to consider in physical model. the most important to consider in physical model. the most important to consider in physical model. the most important to consider in physical model. the most important to consider in physical model. CHAPTER 3 OBJECTIVE AND APPROACH 3.1 Objective Three major objectives were identified for this work. The first objective was to study the behavior of a viscous film and an impinging jet. It was desired to understand how a viscous film, such as the slag film present in a coal gasifier, would respond to an impinging jet. It was anticipated that the results of these studies could be reduced to a single parameter to describe the system, as a Reynolds number does for fluid flow or a Weber number does for a falling droplet. The second was to develop physical and computational models to describe the slag behavior so that different methods of slag diversion could be tested. The models should serve to isolate the most important properties of the slag flow, such as the variable viscosity, so that each property can be studied independently. The final objective was to use the insight gained from the completion of the first two objectives to develop a reliable method for maintaining a clear line of sight across a real reactor. The diversion method should create a completely clear line of sight for many hours at a time in order to be considered successful. Interference with normal operation of the gasifier should be minimized. 34 3.2 Proposed Methods Two different methods of diverting the slag were considered: A purge gas jet and a solid tube to interrupt the slag film and divert it around the sight port. The purge gas jet method will be discussed first. The first method, called the jet diversion method, involves employing a purge gas jet set inside the reactor-side face of the sight port and aimed upward to impact the slag as it arrives at the top of the sight port opening. It was theorized that the momentum of the jet would stop the slag flow and force it to flow around the hole instead. A diagram is shown in Figure 3.1 to illustrate the idea. This method has several points in its favor, such as high adjustability. As the slag rate changes, the flow rate of the jet can be adjusted to compensate. Also, it is projected that the purge flow would keep the device relatively cool, allowing it to be a permanent fixture in the gasifier. There were also disadvantages to the jet diversion method. The geometry of the gasifier would make precisely aiming the device difficult. Also, using a high flow rate of purge would dilute the gasifier reaction and lower performance. Lower temperatures would result in higher slag viscosity, which could cause serious complications such as gasifier plugging. The second method studied, called the tube diversion method, consists of creating a physical diversion of refractory material or alumina set above the sight port. A search of the literature revealed that slag can wear refractory at very high rates. According to Bakker (1984), this may exceed 0.20 mm/hr under extreme conditions. Creating a diverter of refractory that would protrude into the reaction section of the reactor enough 35 Figure 3.1. Diagram of the gas jet diversion method. The slag flows down the wall from above and, as it reaches the opening in the wall, is struck by a gas jet. The momentum of the jet causes the slag to move to the side. This keeps the sight port clear for optical access. to impact the slag flow patterns would leave it exposed to additional erosive forces of the reactant stream and the thermal stress from the flame. There was great concern that such a structure would not last long in a reactor due to the erosive force of the slag flow and the chemical attack such a device would endure from the slag. As an alternative, a hollow alumina tube was chosen to be inserted a short distance into the reaction section. The tube would pierce the slag film and provide a line of sight into the reactor. 36 The tube diversion method presented advantages in that it would be simpler to implement than the jet diversion method and it would have a reduced purge gas usage, which would interfere less with gasifier performance. However, a lack of adjustability during gasifier operation is a potential drawback. 3.3 Approach Performing experiments with real slag samples would be prohibitively expensive. Slag has a very high melting point, 1500 K or above depending on the mineral composition. It also hardens quickly when cooled, making it difficult to work with. To avoid the costs and dangers inherent in live slag experiments, it was proposed to use a simplified physical model and a computational model to study the feasibility of the two proposed diversion methods before implementing them into the gasifier. A physical model using viscous silicone oil as the slag was proposed to study the impacts of film viscosity, surface tension, and flow rate. Specifically for the jet diversion method, the jet flow rate and impingement angle would also be studied. The flow rates at which a clear line of sight was established for these variables would be recorded and analyzed. For the hollow tube method, the approximate length of tube required to maintain the line of sight would be recorded. To study the interaction of fluid flow and the heat transfer effects, particularly the heat transfer between the slag and the jet, a computational model was proposed. The computational model would use the Fluent software package to model a small part of the interior of the gasifier and be applied to each proposed method. The possibility of slag cooling and freezing would be studied. The model would be validated by comparing 37 computational results without heat transfer to those of the physical, silicone oil-based results. It was predicted that by combining the results of the physical and computational models, an efficient and effective method of creating slag diversion around the sight ports in the reactor could be found. It was also expected that an empirical correlation could be developed that will predict the fluid diversion behavior for a particular system whose inputs, such as fluid velocity, viscosity and purge jet mass flow rate, are known. It was proposed to fabricate devices and apply each of them to the entrained-flow gasifier available at the University of Utah facilities for live tests. CHAPTER 4 EXPERIMENTAL 4.1 Introduction To gain a better understanding of the physics involved in altering the slag flow, several experimental approaches were made. A physical model was used to act as a proof of concept for each diversion method. Computational models were also used to examine how heat transfer effects could affect the diversion devices. The computational model also provided a convenient way to optimize the operating conditions of each method. Finally, each method was tested in a pilot-scale gasifier. In this chapter the models and devices created are detailed. 4.2 Silicone Oil Model 4.2.1 Apparatus The silicone oil experiments were devised as an inexpensive and safe way to test the two slag diversion methods. Using silicone oil as a substitute for slag allowed careful examination of the physics involved with both diversion methods without risk of handling very high temperature material. With these physical models, the effects of film mass flow rate, viscosity, and the jet flow rate and impingement angle could be considered and precisely measured. 39 Tests were performed for both the jet and the tube diversion methods. A unique experimental apparatus was created for each. For each of the setups, the scale of the geometry was set to be true to that of the actual reactor. However, the physical model for the jet method experiments was performed using a flat wall rather than a curved one. The experimental setup for the jet method tests, as shown in Figures 4.1 and 4.2, consisted of a vertical wall for the liquid film to flow down, a liquid recirculation system, and a gas jet control system. The vertical wall was made of a 55 x 37.5 cm smooth surface board. An elliptical hole measuring 3.5 x 20 cm was cut out of the wall for the jet to pass through. The hole is centered in the board, and the top of the hole is located approximately 22 cm down from the top of the wall. The wall was suspended on a support structure over the liquid reservoir so that the fluid falling from the wall would be captured and reused. A variable-speed, two stage, positive-displacement pump (Moyno Industrial Products, Model 34408, 2.8 lpm max flow rate) capable of pumping high-viscosity fluids circulated the liquid from the reservoir to a distributor at the top of the wall. The pump was powered by a three-phase electric motor (Baldor Electric Company, Model M3538, 370 W max output). The liquid distributor is made from a 1 cm OD tube attached horizontally to the top of the wall with 2 mm holes drilled at a regular spacing. To get an even film distribution, the spacing is set so that more holes exist near the center of the tube. The liquid flow is directed downward against the board at a 30 degree angle, thus creating a thin film that flows down under the influence of gravity. The film flows down the wall 40 Figure 4.1. Schematic of physical model setup. The hatched area represents the wall for the film formation. The nozzle of the gas jet was attached to a leveled protractor so that the angle could be easily adjusted. 41 Figure 4.2. Geometry of the wall for the oil tests. All units are in mm. The elliptical slot is shown in the wall and within the slot the tube and protractor can be seen. surface, and across the elliptical hole, creating a suspended film for the impinging jet. The film drips off the bottom of the wall into the reservoir. The gas jet nozzle is a simple capillary made of a 4.6 mm ID (5.5 mm OD) steel tube. It is situated inside the elliptical slot in the wall and recessed 13mm from the liquid film surface as shown in Figures 4.3 and 4.4. The back of the hole was left open to allow the gas to escape after impacting the film. 42 Figure 4.3. Jet impingement angle measurement. The tube is attached to a leveled protractor so that the angle of the jet can be measured and freely adjusted from 0 to 60 degrees. A zero degrees angle of impingement corresponds to the jet being set perpendicular to the flow of the oil film, as illustrated in Figure 4.3. The gas jet flow rate is measured using a pair of high-precision rotameters calibrated to measure gas flow rates from 0-50 SCFH and 0-200 SCFH (0-24 and 0-94 SLPM) (Dwyer Instruments, Michigan City, IN). The measured volumetric flow rates were converted to mass flow rates using the ideal gas law. For the experiments presented here, dry, filtered compressed air was used for the gas jet. 43 Figure 4.4. Tube within the elliptical slot. The tube was recessed 13 mm from the board face when the angle was set to zero degrees. 44 The jet flow rate ranged from 10.0 SCFH (4.7 SLPM) to 36.0 SCFH (17.0 SLPM). Assuming the ideal gas law, this is equivalent to 0.096 g/s to 0.35 g/s. The Reynolds number was calculated for these flow rates using where Q is the volumetric flow rate in m3/s, r is the radius in m, and v is the kinematic viscosity of the air in m2/s (de Nevers 2005). The Reynolds number ranged from 140 to 500 and so the flow was assumed to be laminar for all conditions. The second apparatus used for the tube method experiments was very similar to the first and followed the schematic shown in Figure 4.1. However, for these tests, the curved geometry of the gasifier was included in the design, as it was theorized that in such a scenario that the curved surface would have more influence than in the jet method. To create the curved surface the flat board was replaced with a cast piece of refractory. The piece was cast to have the same curvature and to have a sight hole the same size as that of the University of Utah's coal gasifier. The dimensions of the unit are shown in Figure 4.5. A fluid distributor was made for the refractory wall. It is made of 1.25 cm OD (1.1 cm ID) steel tubing and has been bent to fit the curvature of the cast. The distributor is butted up against the top of the refractory and held in place with clamps. The tube has 2 mm holes drilled at regular intervals for the silicone oil to pass through. The holes are arranged so that a greater concentration of holes is at the center of the tube. This was done to achieve an even distribution of silicone oil across the refractory face. To promote film formation, the holes are angled downward about 30o when the distributor is attached. 45 Figure 4.5. Dimensions of the refractory cast. The cast represented here was used for the hollow tube experiments. All units are in cm. Curvature of the refractory was set to match that of the inner refractory layer of the pilot-scale gasifier. This prevents the oil from accumulating in the small gap between the tube and the refractory and spilling into unwanted places. 4.2.2 Samples Silicone oil (Clearco, Bensalem, PA) was chosen for these experiments because its properties are well known and documented and it is available in a wide range of viscosities. It was theorized that the viscosity of the film would be significantly affect the behavior of the film when struck with the impinging jet. For this reason, it was important 46 to have films of variable viscosity without greatly changing the other parameters, such as density or surface tension. Silicone oil is made of polymerized dimethylsiloxanes, structure shown in Figure 4.6. The viscosity of the oil can be controlled by altering the length of the polymers in the oil. Longer, higher molecular weight polymer chains make a more viscous fluid than shorter chains (Friedman 1975). By controlling the average molecular weight of the chains, a broad range of viscosities of oil can be made. Oils with viscosities of 30,000 cSt, 12,500 cSt, 5,000 cSt, and 100 cSt were used. The density of each sample was calculated using a 50 mL volumetric flask and a balance. For each calculation, the mass of the clean volumetric flask was measured on the balance. The flask was then filled with 50 mL of an oil sample and again the mass was measured. The original mass of the flask was subtracted from the result and then divided by the 50 mL volume to obtain the sample density. This was repeated three times for each sample. As expected, the density of the samples had little variance and averaged 976 kg/m3. The surface tension of each sample was also measured using a pendant drop technique. A fellow student, Vasiliy Chernyshev constructed the machine used to conduct the surface tension measurements. Details of this measurement method are given in Appendix D and an illustration of the technique is shown in Figure 4.7. The properties of each sample are summarized in Table 4.1. The surface tension averaged 21.1 mN/m and did not show much variation. This value is in line with other values reported in the literature (Schurch et al. 1976; Carré and Woehl 2006). 47 Figure 4.6. Structure of polydimethylsiloxane. Two units of the molecule are shown. This molecule is the main component of the silicone oil used in the physical model. The R-groups can be additional dimethylsiloxanes or methyl chain termination groups. Figure 4.7. Pendant droplet technique for measuring surface tension. The diameter of the tube and gravitational force are known quantities. The angle of contact between the tube and the droplet and the volume of the droplet are measured. 48 Table 4.1, Silicone oil sample properties. Table includes values for the kinematic viscosity, density, and surface tension. Sample Kinematic Viscosity (cSt) Density (kg/m3) Surface Tension (mN/m) 1 100 975 21.3 2 5000 978 21.0 3 125000 976 21.4 4 30000 975 20.8 4.2.3 Procedure 4.2.3.1 Impinging Jet Tests Prior to each series of measurements, the liquid circulation system was cleaned and the silicone oil with the desired viscosity was loaded into the pump and reservoir system. The pump was primed and the system was run to remove any residual air bubbles from the oil lines. The angle of the impinging gas jet was adjusted and confirmed using the attached protractor. The distance between the jet tip and outer wall surface was confirmed to be 13 mm. For each experiment, the pump was turned on and the liquid was allowed to flow down the board until a uniform film flow had been established. The liquid flow rate was varied between 0.03 and 0.10 l/min by adjusting the speed of the pump motor in order to produce the desired film mass flow rate or thickness. For example, a majority of the experiments were performed with a film thickness 5.0 mm, which corresponded to a liquid film flow rate of 0.25 g/cm-s. Less viscous oils required higher mass flow rates to achieve the desired film thickness. 49 After any adjustment of the pump speed, the mass flow rate of the oil was confirmed, and thereby the thickness of the film was also confirmed. As the oil fell off the board into the reservoir, it would accumulate into a single falling stream. This stream was collected in a clean beaker, the mass of which had been previously measured. The time of collection was measured with a stopwatch so that the mass flow rate could be calculated. For the purge jet experiments, once the desired oil film flow had been achieved and the flow appeared to be stable, the air jet was slowly opened, at an approximate rate of +0.25 SLPM/sec as the behavior of the two fluids was observed. The jet flow rate was increased until the oil film burst open and the film began to divert around the jet flow. Once the liquid film burst open, the corresponding jet flow rate was noted. The air was then shut off and the oil film was allowed to return to a uniform flow. The experiment was repeated several times, except that in the subsequent tests, extra care was taken near the previously recorded bursting flow rate. The flow rate of the jet was increased very slowly so as to get a precise measure of the jet flow rate when film bursting occurred. Approximately 5 seconds time was given between each increase in the jet flow rate. After finding the gas jet flow rate required to burst the oil film, a similar process was used to determine the flow at which the oil film would flow over the jet flow path and reform a closed film. The gas jet flow rate was first adjusted to twice that which had been required to burst the film, and was slowly reduced at -0.25 SLPM/sec until the opening in the oil film closed. Again, once the flow rate was determined, experiments were repeated with more careful decreases in the gas flow rate. The system was allowed 50 to operate for several seconds between adjustments until the precise flow rate at which film diversion closure occurred had been precisely determined. The impingement angles tested varied across the different oil viscosities. Each of the four oils was tested at 0o, 10o, 15o, 30o, 45o, 50o, and 60o angles. Some oils were tested at intermediate angles when more data points were desired. 4.2.3.2 Tube Diversion Tests The objective of the performing the tube diversion method tests was to gain a qualitative understanding of how the fluid film would behave upon impacting the tube. Thus the tube tests were done in a less precise manner. Before each set of experiments, the pump and associated tubing were cleared of residual oil and the desired viscosity oil was loaded into the pump fluid reservoir. A hollow alumina or PVC tube was placed in the sight port of the refractory cast so that a 4 cm length protruded from the inner curved surface of the refractory into the path of the oil film. The tubes had an outer diameter of 3.2 cm. The pump was then initiated and the rate was adjusted to produce a film of the desired thickness. After any pump speed adjustment, the oil film mass flow rate was confirmed as described previously in this section. The behavior of the film as it met the tube was observed. At the impact area, the oil accumulated and formed a lip on the tube, as shown in Figure 4.8. Once this lip had stabilized, the thickness of the lip from the refractory face was measured. After measuring the silicone oil lip on the alumina tube, the tube was slowly, in 0.25 cm increments, pulled back out of the refractory, reducing the exposed section. 51 Figure 4.8. Silicone oil on impact with the tube. As the oil impacted on the tube, a lip formed on top of the protrusion. The thickness of this lip was measured from the refractory face. After each adjustment, the flow was allowed to stabilize. This was continued until the oil overran the lip of the alumina tube and spilled over into the line of sight. These tests were repeated for three oil film mass flow rates. The first flow rate was based on an estimation of the slag flow rate on the walls of the gasifier. The second and third flow rates were set to be double and quadruple of the first to test the robustness of the design of the tube diversion method. 52 It should be noted that the thickness of the produced film was not directly measured in any of the experiments because insertion of any measurement device would disturb the oil flow. Instead the film thickness was calculated based on the measured film mass flow rate. From the work of Bird et al. (2002) the thickness of a falling film such the one presented here is √ ̇ where  is the thickness of the film in m,  is the kinematic viscosity, ̇ is the mass flow rate of the film in kg/s, W is the width of the film flow in m, and  is the angle between the direction of flow and gravity (zero in this case, thus cos(0) = 1). The method executed to arrive at this equation is presented in Appendix A. Such an equation depends on the assumption that the fluid is Newtonian. Although silicone oil is non-Newtonian, studies of the material by Carré & Woehl (2006) determined that it behaves as a Newtonian fluid until shear rate exceeds a critical value, in this case approximately 1000 s-1. During such shear, the fluid becomes shear-thinning, exhibiting less viscosity as increasing shear rate is applied. In the system under consideration here, the shear rate on the fluid was not measured. However, Appendix B shows a quick exercise to prove that the fluid stayed well within the Newtonian regime. 53 4.3 Computational Model The physical silicone oil models provided good information about the feasibility of the proposed diversion methods. They also helped to determine optimal operating conditions for both of the diversion methods. However, the physical models were not able to predict the effects of heat transfer on the slag behavior. In particular, the tendency of the slag viscosity to greatly increase with decreasing temperatures was of great concern. Several questions were raised: - As the impinging jet cooled the slag, would the heat transfer from the reaction gases and flame be sufficient to maintain the slag in a liquidus state or would the slag freeze into a solid block? - If the slag froze, would it maintain a constant position or continue to slowly flow? - Would any solid slag help or hinder the diversion of the slag film? These questions could not be answered using the silicone oil model. Thus the computational models were developed to take the heat transfer effects of the system into account. Two separate geometries were created to mimic the apparatus used for each of the silicone oil tests. This was done so that the results of the silicone oil tests could serve as validation of the computational model. 54 4.3.1 Geometry and Boundary Conditions 4.3.1.1 Jet Diversion Model The geometry for the jet diversion method computational model is shown in Figure 4.9. It consists of a flat, vertical plane measuring 200 x 385 mm. A cylindrical, 22 mm diameter hole is set centered horizontally in the board, 181 mm from the top. Within the hole, a 5.5 mm OD (4.4 mm ID) tube is set at a 45 degree angle so that the tube is aimed at the point where the topmost part of the cylinder intersects with the vertical plane. The tube is recessed 13 mm from the vertical plane surface. All of the wall which the film traverses was a no-slip wall boundary. . In order to reduce demand for computation time, the geometry was split in half with a plane that sits perpendicular to the vertical wall. This new plane was assumed to be a symmetry boundary for the experiments. Also in Figure 4.9, the jet inlet boundary can be seen. This is shown in greater detail in Figure 4.10. The semicircle at the bottom of the jet was appointed as a mass flow inlet boundary for the jet. This allowed the flow rate of the jet to be easily adjusted for different computational runs. At the top boundary, along the edge where the top and film wall boundaries meet, a 1.25 cm-thick stretch was designated as the slag inlet boundary. The slag inlet stopped short of the side boundary by 1.5 cm. This arrangement is illustrated in Figure 4.11. The vector for the slag flow was set to be at 45o into and down the film wall. The bottom boundary was set as an outflow boundary condition. Also, the semi-circle at the rear of the sight port was also set as an outflow boundary, to allow the jet pressure to escape if the opening in the film wall became blocked. Figure 4.9. Geometry for the impinging jet computational model. All units are in mm. 56 Figure 4.10. Detail of hole and tube. The red arrow indicates the flow of the air jet from the bottom of the tube boundary. Units are in mm. Angle between tube and horizontal is 45 degrees. Tube is recessed 13mm from the surface of the vertical plane. 57 Figure 4.11. Top boundary arrangement of the slag inlet boundary. The hatched area is the slag inlet boundary. The blank area is the top boundary. All units are in cm. The meshes for the geometry were created using Gambit 2.4.6 (Ansys, Canonsberg, Pennsylvania, USA). The mesh totaled 10,300 nodes for the entire geometry. The mesh was created by first meshing each of the vertices of geometry. The successive ratio option was used to concentrate the nodes of the mesh around the sight port. The nodes were clustered near the film wall and became less dense drawing away from the wall. A particularly fine mesh, 0.025 cm, was used on the edges of the two mass flow rate inlet boundaries to ensure good development of the flow out of the 58 boundaries. Once the edges were meshed, the faces and volume were meshed. The face meshes used the Quad/Tri elements with the Map type option. The volume mesh used tetrahedral and hybrid elements and the TGrid type option. Initial meshing produced several (~25) highly skewed elements. The mesh was examined in Gambit to identify the affected elements. The vertices of these elements were adjusted to eliminate the problem. 4.3.1.2 Tube Diversion Model The second geometry was similar to the first; however the second was more complicated to create due to the curved refractory surface. This geometry is shown in Figure 4.12. The model is based around a curved surface, 38.1 cm tall with a diameter of 20.4 cm. Centered 21 cm from the top of the curved surface, a 1.27 cm/0.95 cm (OD/ID) tube juts 3.55 cm into the inner surface so that the length of the tube is perpendicular to the length of the curved surface. The tube continues out the back of the surface for an additional 6.36 cm. The curved surface and the walls of the tube were set as no-slip wall boundaries. The bottom was designated as an outflow boundary and the face of the rear of the tube was set as a mass flow inlet to act as the purge stream during reactor operation. A 1.25 cm strip along the intersection of the top and the curved surface was set to be the slag inlet, a mass flow rate boundary. The face opposite the curved surface was set as a symmetry plane. The mesh for the second geometry totaled 23,000 nodes. As before, the successive ratio option was used on the edge meshes to concentrate the nodes near the location where the curved surface and the tube meet. The mass flow inlets were finely meshed. 59 Figure 4.12. Tube diversion computational model geometry. In the picture, the shape and dimensions of the model are given, including the inner and outer radii of the tube and the inner radius of the wall. All units are in cm. 60 In order to have greater control over the overall form of the mesh, the volume was divided into nine sub-volumes. This was done by first inserting four concentric cylinders into the volume. The cylinders are centered on the tube and increase in diameter and length. The leftover space was subdivided by perpendicular planes into four separate sections and the tube itself made to be the last subvolume. This arrangement is illustrated in Figures 4.13 and 4.14. The tube had a very fine mesh applied to it and each succeeding cylinder had a slightly larger volumetric mesh size. This allowed the model to be more precise and accurate at the most important area: where the slag impacted on the tube protrusion. Subdividing the volume in this manner thus allowed for a more detailed analysis near the tube and wall intersection. This subdivision method also helped to prevent highly skewed mesh elements. This is desirable because mesh elements with high amounts of skewedness can cause instability in the model solution. 4.3.2 Slag Model Properties 4.3.2.1 Slag Composition As the properties of slag can change drastically depending on the mineral content of the melt, most of the slag property equations are composition dependent. Single oxidation number components such as calcium and sodium will lower the viscosity of a silicate melt. Under reducing conditions, iron ions will also contribute to a low viscosity slag. Therefore, in order to achieve the most realistic results in the computational model, the slag composition for the model was based on the coal that would be used for the 61 physical experiments in the gasifier, a locally obtained Utah coal from the Sufco coal mine with 2 wt% limestone added as a fluxant. Figure 4.13. Geometry wireframe, front view. The arrangement of the cylinders and planes used to subdivide the main volume into sections that were easier to work with. Figure 4.14, Isometric view of the geometry wireframe. Illustration shows the four dividing cylinders.63 This coal is useful in this instance because it has very low iron content. A great majority of the slag property models were developed under oxidizing conditions where the iron is likely to be in the Fe3+ form. However, in a gasification environment, a strong reducing atmosphere is present and the iron in the slag is more likely to be in the form of Fe2+. Studies by Hurley et al. (1996) and Song et al. (2010) both conclude that the surrounding atmosphere can greatly impact the slag properties, particularly the viscosity. However, another study by Folkedahl and Schobert (2005) showed that magnitude of the effect is directly related to the amount of iron in the slag. Due to the different oxidation states of iron, a low-iron slag is much less affected than a slag with high iron content. Iron under reducing atmosphere is likely to be in the FeO form. As such, it would contribute to a weak crystal lattice. Under oxidizing conditions, iron is more likely to be found in the Fe2O3 form, which with its multiple cross linking capabilities, is more likely to produce a more cross-linked lattice and, as a result, a more viscous slag. However, due to the low iron content of the Sufco coal, it was assumed that the models used to describe the coal properties would be reasonable estimations, despite the difference in the surrounding atmosphere. The coal composition was analyzed by Huffman Laboratories (Golden, Colorado, USA) and their analysis of the mineral matter present is presented in Table 4.2. The composition of the ash including the 2 wt% limestone fluxant additive is also given in the table. As can be seen, the amount of iron was 5.25 wt% without the fluxant and only 4.24 wt% after the addition of the limestone. 64 Table 4.2. Ash composition in weight percentage of the Sufco coal. The mass percentage of each mineral is shown with and without the limestone fluxant additive. Mineral Original Composition With Limestone Fluxant SO2 5.96 4.81 SiO2 48.85 39.42 Al2O3 8.34 6.73 CaO 18.21 34.00 FeO 5.25 4.24 MgO 2.84 2.29 Na2O 3.09 2.49 K2O 0.33 0.27 TiO2 0.64 0.52 4.3.2.2 Temperature-independent Properties The slag density was calculated from eq 3.2, from the work of Mills and Keene, The heat capacity was calculated using the Kopp-Neumann Law, as given in eq 3.4, Σ Data for the individual oxide heat capacities were obtained from the Condensed Phase Thermochemistry Section of the NIST Chemistry Webbook, data compiled by Damalski and Hearing (2012). Heat capacity values at 1500 K were used when available. When a specific value at 1500 K wasn't present in the data, nearby values were linearly interpolated. 65 Equation 3.5 was used to calculate the thermal conductivity, from the work of Mills and Rhine, where  and Cp are the values obtained from eqs 3.2 and 3.4. The value for aeff was set to 4.5x10-7 m2/s, as recommended by Mills and Rhine. It is possible to calculate the heat capacity, and thereby the thermal conductivity, based on temperature. However, to do so requires a large amount of effort, expressly calculating the values as given by the eight-term Shomate equation. For each component, the eight terms would have to be gathered and the equation recalculated for the given temperature. Such an effort would have increased computational time and possibly led to instabilities in the model in exchange for questionable benefit. 4.3.2.3 Viscosity Model The viscosity model proposed by Senior and Srinivasachar (1995) was applied to the computational model. This viscosity model was chosen because it is valid over a wide range of temperatures. Where most models are limited to temperatures above the Tcv of the slag, the model developed by Senior and Srinivasachar has a separate equation for predicting the slag viscosity in the high-viscosity region below the Tcv. It was postulated that in a computational model, the greatly increased slag viscosity calculated by the viscosity model could allow the computational model to take the slag freezing into account without the need to develop an explicit frozen slag layer. 66 The A and B values presented in Section 2.2 were implemented in Eq. 4.5 and applied to the Fluent simulation by way of a user-defined function or UDF. A UDF is a short C++ script that is interpreted by Fluent. The UDF used has been included in Appendix C. The model was verified by applying it to a similar Sufco slag for which viscosity-temperature data was available. Groen et al. (1998) collected such data for slag from coal from the same origin and a very similar mineral composition. After some adjustment of the parameters, the results were very favorable, shown in Figure 4.15. 4.3.3 Solution Setup The solution for the model was calculated using the Ansys Fluent package, version 6.3.26 (Ansys, Canonsberg, Pennsylvania, USA). For each of the geometries, a steady-state solution was produced using the pressure-based, implicit solver. To handle the fluid/gas interactions, the implicit volume of fluid (VOF) multiphase model was selected. According to the Fluent documentation (Fluent Inc. 2006), the VOF model works well for modeling the behavior of two immiscible fluids, and the user guide suggests that it would work well for jet-breakup and liquid-gas interface studies. The VOF model is also relatively easy to implement. These strengths made it an ideal choice for this case. The Fluent software requires that the PRESTO! discretization method be used for the pressure calculations with the VOF multiphase model. The momentum calculations were done using a second-order upwind discretization while the volume fraction calculations were completed with a first-order upwind discretization scheme. The energy equation was also discretized using the first-order upwind method. 67 Figure 4.15, Slag viscosity experimental data versus model predictions. Comparison of the experimental results of Groen et al. (shown as squares) and the Senior-Srinivasachar model (gray line). To promote overall model stability, under-relaxation was used. Under-relaxation values are a common method of ensuring that a system remains stable during the solution process. These values are used to scale the relative error during the solution and can help to dampen any oscillations in the solution values. The under-relaxation values used for the pressure, density, body forces and other equations for the Fluent simulation are given in Table 4.3. Most of the values are the defaults in the Fluent program, however it was found that reducing the energy under-relaxation factor from a value of 1.00 to 0.85 greatly improved the stability of the solution. Relaxing the pressure to 0.3 also helped the solution to converge more rapidly. 0 1000 2000 3000 4000 5000 6000 7000 1450 1500 1550 1600 1650 1700 1750 1800 Viscosity (poise) Temperature (K) Exp. Data (Groen et al.) Senior-Srinivasachar Model68 Table 4.3. Under-relaxation factors used for the computational models. The factors are used in the solution to promote solution stability. Equation Under-Relaxation Factor Pressure 0.3 Density 1.0 Body Forces 1.0 Momentum 0.7 Volume Fraction 0.20 Energy 0.85 4.4 Gasifier Application 4.4.1 Apparatus 4.4.1.1 Reactor The reactor used for these experiments is a pilot-scale entrained flow gasifier, located at the University of Utah Gasification Research Facility (Salt Lake City, Utah, USA). The cylindrical reactor stands approximately 7.3 m in height and has a maximum thermal input of 600 kWth. This corresponds to a maximum coal feed rate of 70 kg/hr. The reactor is rated to a maximum pressure of 450 psi (3100 kPa). A rendering of the gasifier is shown in Figure 4.16. The reactor can be considered as three distinct functional sections: the insulated reaction section, a pre-quench section, and the quench section. Fuel and oxidizer are fed into the reaction section and the product gases and slag move into the pre-quench section. There they are 69 sprayed with water to stop the reactions before passing through the quench bath at the bottom of the reactor. The reaction section is 160 cm in length and is refractory lined. The empty section has a radius of 40.4 cm, but with two, 15.2 cm thick refractory layers, a space with a radius of 10.2 cm remains for the reactions. The cold face refractory is Plicast LWI 28 and the hot face refractory is Plicast Si-Bond SP, both from the Plibrico Co. (Chicago, IL). The cold face is rated at a maximum temperature of 1350 oC and the hot face refractory is rated to 1760 oC. Both refractory sections consist primarily of Al2O3 and SiO2. Down the side of the reactor, along the reaction section, are six pairs of opposing sample ports. Each port is a two-inch (5.1 cm) nominal pipe. The third and fourth sets of ports, counting from the top, were used for optical access through the reactor. For coal gasification operations, the reactor is fed coal slurry (an approximately 60 wt% micronized coal and water mixture) and oxygen. It is downward fired and uses a concentric two stream burner. In such a burner, the coal slurry is injected through a small-diameter tube and the oxygen is fed through a surrounding annulus. The velocity of the oxygen leaving the annulus helps to atomize the coal slurry and increases conversion. During operation, the feed rates of the fuel and oxidizer are adjusted to maintain a set residence time in the reaction section. The coal mass flow rate is governed by ̇ ( ) ( ( )) 70 Figure 4.16. Cut-away section of the gasifier. The orange section represents the refractory of the reaction section. The three sections are indicated, A - reaction section, B - pre-quench, and C - quench bath. 71 where  is the residence time, V is the reactor volume, P is the pressure with gage and atmospheric indicated by g and atm subscripts,  is the desired fuel to oxygen ratio, and  is the mole percent of carbon, hydrogen, oxygen, and sulfur in the coal. The equations used to develop this operation are shown in detail in Appendix E. 4.4.1.2 Gas Jet Diversion Device Based on the results from the physical and computational models, a jet diversion device was created to divert the slag inside the operating gasifier. The device consists of a high-pressure sapphire window, two purge inlets, and a long tube to act as the purge jet. One purge inlet serves to blow onto the inner surface of the sapphire window to prevent particulate deposition and to keep water from invading the sight port tube. The other purge inlet feeds the purge jet to divert the slag. The unit includes a 2-inch, 300 class flange so that it could be directly mounted to the sample ports on the reactor. Two units mounted at the same height provide a clear line of sight across the reaction section of the gasifier. All of these features can be seen in the photograph of the device, shown in Figure 4.17. The jet is a simple capillary jet, fashioned from a 5.5 mm OD/4.4 mm ID stainless steel tube. The length of the tube is approximately 33mm and is straight for all but the last 7.5 mm where end of the purge jet tube is angled upward 30o. The tube is oriented so that it sits suspended along the top of the sight port. From the results of the silicone oil experiments, it was determined that the optimal point to aim the jet was at the top of the sight port hole in the refractory. The length of the tube was measured to just sit inside the hole and direct the jet at that point. 72 Figure 4.17. Jet diversion device. The major features of the device are labeled. The unit was mounted in the sample port on the gasifier. The windows were designed and fabricated especially for this project. The windows are wedge shaped so that the pressure inside the reactor presses the window further against its seal. This allows the windows to maintain a seal under high pressure and to maintain that seal even after multiple temperature cycles. The windows sit in a threaded mount so they can be easily removed for cleaning. A cut away view of the window design is shown in Figure 4.18. Each window has two seals. A Teflon gasket sits at the bottom of the threaded section to provide a seal at the threads. To seal the sapphire piece, the sides of the glass were coated with Ultra Copper by Permatex (Solon, Ohio), a high-temperature, copper-infused RTV silicone for making gaskets. A ring of silicone was also deposited around 73 Figure 4.18. Optical access window design. The windows sit in a steel holder that screws into a base attached to a flange. There are two points for a seal: at the bottom of the screw down section and between the sapphire and the steel holder. As the pressure in the reactor increases, the glass presses into the steel holder with greater force, increasing the strength of the seal. All units are in cm. 74 the perimeter of both ends of the sapphire once it had been inserted into the steel holder. A photograph of the window holders with the sapphire pieces and silicone seals is shown Figure 4.19. 4.1.1.3 Alumina Tube Diversion Device Devices were created to hold an alumina tube in the reaction section of the gasifier. The device consisted of the body and flange assembly, the window, and the alumina tube. A single purge inlet blows across the window face to keep any particles from depositing on the relatively cool sapphire face. Photographs of the completed assembly are shown in Figures 4.20 - 4.22. The photographs show the tube is held perpendicular to the flange at the base of the device using a compression fitting scheme. The fitting was packed with ceramic rope for the compression media. The total tube length was 47 cm. This length was chosen so that 3.8 cm of the tube penetrated through the refractory and protruded into the reaction section of the gasifier. To assemble the device, a small layer of ceramic rope is first coiled into the retaining hole of the device. Once the ceramic rope has been compacted into the orifice, the alumina tube is inserted into the center of the rope coil and butted up against the window section. A steel ring is slipped over the tube and tightened into the retaining section. The ring compresses the rope and holds the tube tightly in place. No seal is produced by the rope; however it does cause that the great majority of the purge gas directed at the inner face of the sapphire travels down the length of the interior of the tube rather than escaping around the tube exterior. This keeps water and particulates from making their way into the tube and keeps the device relatively cool during gasifier operation. The retaining ring and body are shown in Figure 4.21. 75 The windows used for this device were identical to those used in the jet diversion device. The design works very well in high-temperature and pressure situations and maintains a seal even after several thermal cycles. A photograph of the complete window and tube assembly installed on the sight port of the pilot-scale gasifier is shown in Figure 4.22 and a photograph of the complete unit mounted onto the gasifier is shown in Figure 4.23. Figure 4.19. Steel window holders, sapphire pieces, and the high-temperature silicone sealant. A fine, 1.25 inch thread was used for the holders. To prevent the parts from sticking together from the stress of the thermal cycles, the threads were coated with a nickel-based anti-seize compound, which can also be seen in the photo. 76 Figure 4.20. Assembled alumina tube diversion device. Figure 4.21. The retaining ring of tube diversion device. The ring compresses ceramic rope (not pictured) to hold the tube in place and to direct the purge flow down the interior of the tube. 77 Figure 4.22. Window assembly of tube diversion device. The small white ring seen through the sapphire is the Teflon seal that sits between the base and the screw down window holding piece. Also pictured is the purge inlet that directs a stream across the inner face of the sapphire. 78 Figure 4.23. Mounted assembly of tube diversion device. Photograph of the complete slag diversion assembly mounted on the gasifier sight port. The orange tint seen in the window is from the gasifier flame. 79 4.4.2 Coal Samples For the jet diversion device tests, the gasifier was run on a bituminous coal from the Skyline mine, located near Helper in Carbon County, Utah. The tube diversion tests used Sufco coal, from the Sufco mine near Salina in Sevier County, Utah. An ultimate analysis of the coals, as reported by Huffman Laboratories, is given in Table 4.5. The proximate analysis and the ash composition analysis for each coal are given in Tables 4.6 and 4.7. Table 4.5, Ultimate analysis of Skyline and Sufco coals. Element Skyline Sufco Carbon 70.60 67.87 Hydrogen 5.41 5.45 Nitrogen 1.42 1.09 Sulfur 0.53 0.36 Oxygen 13.21 16.87 Table 4.6, Proximate analysis of Skyline and Sufco coals. Skyline Sufco Volatile Matter 38.6 38.49 Fixed Carbon 49.39 47.04 Ash Content 8.83 8.36 HHV (BTU/lb) 12600 11900 HHB (kJ/kg) 29260 27620 80 Table 4.7, Ash composition analysis of Skyline and Sufco coals. Oxides Skyline Sufco Aluminum 14.52 8.34 Calcium 6.11 18.21 Iron 5.09 5.25 Magnesium 1.39 2.84 Manganese 0.02 0.05 Phosphorus 0.59 0.01 Potassium 0.57 0.33 Silicon 60.89 48.85 Sodium 1.41 3.09 Sulfur 2.33 5.96 Titanium 0.88 0.64 The NBO/T ratio of the coals is -0.477 for the Skyline sample and 0.395 for the Sufco sample. The disparity in the non-bridging oxygen ratio is primarily due to the high amount of calcium present in the Sufco coal. This difference indicates that the slags will have very different viscosity behaviors when heated. Each coal was received as a fine pulverized powder (200 mesh). The coal was mixed with water to create a thick slurry that was approximately 60 wt%. CHAPTER 5 RESULTS AND DISCUSSION In this chapter, the results of the physical model, the computational model, and the gasifier tests are presented and analyzed. The method of data analysis for the physical model is also developed. The results are presented in a logical, categorical fashion. However, they were not performed in the order in which they are presented. The physical model was developed first. Then, the gasifier test for the jet diversion device was performed. Next, the computational models were developed and finally the tube diversion device was applied to the gasifier. 5.1 Silicone Oil Experiments 5.1.1 Jet Diversion Experiments 5.1.1.1 General Observations The goal of the jet diversion experiments was to determine the effect of film mass flow rate, film viscosity, jet flow rate, and jet impingement angle on the behavior of the system. Figures 5.1-4 show a series of photographs of the film being impacted by the gas jet. When the jet was applied to the falling oil film at low jet flow rates, it was found that 82 Figure 5.1. Low flow rate jet impacting the silicone oil film. For this picture, the film was created with 12,500 cSt oil at a flow rate of 1.5 g/cm/s. The jet flow rate was about 0.2 g/s and the angle was set at 45o. 83 Figure 5.2. Higher flow gas jet with increased bubble size. This photo is of the same 12,500 cSt film as Figure 5.1 with an increased jet flow rate of 0.25 g/s. The bubble has pulled even farther away from the film. 84 Figure 5.3. Increased jet flow rate producing a fluctuating bubble. This photograph used an applied jet flow rate of 0.32 g/s. All other variables are consistent with the previous photographs. At these conditions, the tip of the bubble would begin to fluctuate up and down rapidly. 85 Figure 5.4. Silicone oil hood formation. For this photo, the 12,500 cSt silicone oil was used and the jet was set at 0.24 g/s with a 45o impingement angle. At these conditions, the hood was stable, but lowering the jet flow rate would cause the hood to fall over the jet stream and reform the film. 86 the jet would cause a small bubble to form in the face of the film, as shown in Figure 5.1. The force of the jet impacting the film would cause the film to bulge slightly in the direction of the jet. The surface of the bubble was stable and did not move unless the jet flow rate or the film flow rate changed. Increasing the jet flow rate would cause this bulge to predictably increase in size and the separation between the film and the bulging shape would become more distinct, stretching perpendicularly away from the surface of the film. As shown in photograph in Figure 5.2, while the overall size of the bubble increases, the size of the opening in the film stays the same. Further increasing the flow rate of the gas jet would cause the bubble to begin to fluctuate. The tip of the bubble would drop down and rise up again rapidly. Initially, it was thought that some of the gas was escaping through the bubble during these fluctuations and thus some of the pressure on the surface of the bubble was momentarily relieved. To test this theory, the experiments were repeated with thin strips of paper placed around the external surface of the bubble to act as a rudimentary method to detect air currents. It was discovered that no gas passed through the film, despite the erratic behavior of the bubble. The cause of the bubble's behavior at this gas flow rate was not determined, although the effect was repeatable and consistently occurred just before the bubble would burst. Figure 5.3 shows a photograph of the fluctuating bubble surface. Increasing the flow rate of the jet any further would cause the bubble to burst open. Once open, the film would continue to flow around the gas stream, leaving an open line of sight through the film, even as the jet flow rate was reduced. 87 At the top of the film opening, a lip would be formed. This lip would be stable at jet flow rates close to that which burst the film initially. As the flow rate was decreased, it was observed that the lip would destabilize and begin to oscillate away from the wall. At lower jet flow rates, the lip would stabilize again and form a hood. This hood formation is shown in Figure 5.4. Soon after this formation was observed, as the jet flow continued to decrease, the hood would fall over the weakened jet stream and reform a film. The reformed film is shown in Figure 5.5. 5.1.1.2 Method of Data Interpretation The precise jet flow rate at which the film burst or closed was recorded for each of the four viscosities of silicone oil across a range of jet impingement angles. Further tests using only the 12,500 cSt silicone oil were carried out in which film thickness was varied by adjusting the film mass flow rate. It was desired to develop a method of quantifying the effect of each variable (film viscosity, jet mass flow rate, jet impingement angle, and distance from jet to the film). To do this, a dimensionless number was developed, similar to the Weber number discussed in Chapter 2. This dimensionless number, called a Lotte number and noted as Lo, was intended to provide a ratio of the kinetic energy of the jet to the surface energy of the film and take each of the previously discussed variables into account. The number was formulated as ̇ 88 Figure 5.5. The reformed oil bubble. In this photograph is 12,500 cSt silicone oil film, after the gas jet flow rate was decreased to 0.22 g/s. At this flow rate, the hood formation has dropped across the jet stream and reformed the film. 89 where ̇ is the mass flow rate of the gas jet, j is the density of the jet, x is the distance from the outlet of the jet to the film surface, Af is the impact area of the jet onto the film surface, and f is the surface tension of the film. Information on the development of this equation and the calculation of the area of jet impact on the film follows. The dimensionless number used to analyze the data from the silicone oil experiments was based on the idea of a Weber number. The Weber number is commonly used to describe the behavior of droplets on impact with a surface. It is formulated as the ratio of the kinetic energy of the droplet and the surface energy: where ρ is the droplet density, D is the droplet diameter, V is the velocity, and γ is the surface tension. The Weber number can also be used to measure fluid interactions in other cases where kinetic forces are important versus surface forces. The concept has been applied to describe liquid droplet formation in a jet by Senecal (1999) and (Tambe 2004), fluid slag behavior at a solid interface by Jonsson and Jonsson (1996), and for slag droplet sticking predictions inside gasifiers by Yu (2010). For this study, it was found that a formulation similar to the Weber number could be used to quantify the conditions at which the film burst. The concept of a Weber number was modified so that, rather than give a ratio of the kinetic energy and surface 90 energy of a single droplet, the new dimensionless number (called a Lotte number, Lo) would give a measure of the ratio of kinetic energy in the gas jet and the surface energy of the oil film. Initially t