Thermal and Mechanical Analysis of Carbon Foam A dissertation presented to the faculty of the Russ College of Engineering and Technology of Ohio University In partial fulfillment of the requirements for the degree Doctor of Philosophy Mihnea S. Anghelescu March 2009 © 2009 Mihnea S. Anghelescu. All Rights Reserved. 2 This dissertation titled Thermal and Mechanical Analysis of Carbon Foam by MIHNEA S. ANGHELESCU has been approved for the Department of Mechanical Engineering and the Russ College of Engineering and Technology by M. Khairul Alam Moss Professor of Mechanical Engineering Dennis Irwin Dean, Russ College of Engineering and Technology 3 ABSTRACT ANGHELESCU, MIHNEA S., Ph.D., March 2009, Integrated Engineering Thermal and Mechanical Analysis of Carbon Foam (122 pp.) Director of Dissertation: M. Khairul Alam Carbon foams are porous materials which are attractive for many engineering applications because their thermal and mechanical properties can be customized by varying manufacturing process parameters. However, a highly random geometry at pore level makes it very difficult to analyze the properties and the behavior of this material in an application. Published research work on the analysis of foams has employed various ideal geometries to approximate the pore microstructure. However, these models are unable to predict accurately the foam properties and behavior in engineering applications. The objective of this research work is to determine thermal and mechanical properties of carbon foam on the basis of its true microstructure. A new approach is proposed by creating a three dimensional (3D) solid model based on an accurate representation of the real geometry of carbon foam. Finite element models are then developed to investigate the bulk thermal and mechanical properties of carbon foam using the three dimensional solid model. On the basis of the true 3D model of carbon foam, a study is undertaken to examine the effect of the unique microstructure on the flow field within the foam pores and the resultant convective heat transfer. A finite volume model is developed using the accurate representation of carbon foam microstructure inside a flow channel. The fluid 4 flow and heat transfer is simulated to evaluate pressure drop and heat transfer capabilities. The carbon foam permeability, inertial coefficient and friction coefficient are determined and found to be in good agreement with experimental and semi-empirical models. The results also show a large enhancement in the heat transfer due to the presence of carbon foam in the channel. These results are comparable to the experimental results available in published literature. Another application that has been analyzed in this study is the use of carbon foam as tooling material for manufacturing advanced composite materials. Finite element simulations are carried out to predict the process induced residual stresses and deformations when a composite part is manufactured on conventional tooling versus carbon foam tooling. The results show that both the lower coefficient of thermal expansion and the elastic modulus of carbon foam contribute to the reduction of residual stress and deformation of the composite part. Approved: _____________________________________________________________ M. Khairul Alam Moss Professor of Mechanical Engineering 5 ACKNOWLEDGMENTS I would like to express my deepest gratitude to my academic advisor, Professor Khairul Alam for his support and outstanding guidance during my PhD program. I am extremely indebted to him for giving me the opportunity to work on such an interesting research project. I would like to thank Drs. Dusan Sormaz, Daniel Gulino, David Ingram and Liwei Chen, for serving on my dissertation committee. I would like to thank the Air Force Research Laboratory (AFRL, Dayton, OH) for support and for providing the 3D rendering of the carbon foam microstructure. I am also thankful to Drs. Adriana and Calin Druma for their helpful suggestions during the first stage of this research project. I would like to acknowledge the support provided by GrafTech International Ltd., and the allocation of computing time and the software products Hyperworks, Abaqus and Fluent from the Ohio Supercomputer Center. I would also like to acknowledge the use of the software product Geomagic Studio from Geomagic, Inc.. I am extremely grateful to my family for their continuous support and encouragement in pursuing my career. I am especially thankful to my parents for supporting and helping me during various stages of my life. 6 TABLE OF CONTENTS Page Abstract ............................................................................................................................... 3 Acknowledgments............................................................................................................... 5 List of Tables ...................................................................................................................... 9 List of Figures ................................................................................................................... 10 Chapter 1: INTRODUCTION TO CARBON FOAM ...................................................... 13 Chapter 2: SOLID MODELING OF CARBON FOAM MICROSTRUCTURE ............. 18 Chapter 3: THERMAL AND MECHANICAL CHARACTERIZATION OF CARBON FOAM MICROSTRUCTURE.......................................................................................... 26 3.1 Thermal analysis ..................................................................................................... 26 3.1.1 Introduction and objective ............................................................................... 26 3.1.2 Methodology and simulations .......................................................................... 29 3.1.3 Results and conclusions ................................................................................... 32 3.2 Mechanical analysis ................................................................................................ 35 3.2.1 Introduction and objective ............................................................................... 35 3.2.2 Methodology and simulations .......................................................................... 36 3.2.3 Results and conclusions ................................................................................... 39 Chapter 4: FLUID FLOW AND CONVECTION HEAT TRANSFER IN CARBON FOAM ............................................................................................................................... 43 4.1 Introduction ............................................................................................................. 43 4.2 Volume averaged model ......................................................................................... 45 7 4.2.1 Fluid flow model .............................................................................................. 46 4.2.2 Heat transfer model ......................................................................................... 49 4.3 Direct simulation of fluid flow and heat transfer .................................................... 52 4.3.1 Ideal pore geometry ......................................................................................... 52 4.3.2 Real pore geometry .......................................................................................... 53 4.3.3 Fluid flow computational model ...................................................................... 54 4.3.4 Convection heat transfer computational model ............................................... 56 4.4 Results and conclusions .......................................................................................... 57 4.4.1 Fluid flow analysis ........................................................................................... 57 4.4.2 Convection heat transfer analysis .................................................................... 65 Chapter 5: CARBON FOAM TOOLING FOR ADVANCED COMPOSITE MANUFACTURING ....................................................................................................... 73 5.1 Introduction ............................................................................................................. 73 5.2 Thermo-chemical model ......................................................................................... 80 5.3 Stress-displacement model ..................................................................................... 85 5.4 Simulations and results ........................................................................................... 90 5.4.1 Results for curved tooling geometry (convex and concave) ............................ 94 5.4.2 Results for flat tooling geometry ...................................................................... 98 5.4.3 Effects of thermo-mechanical properties of tooling....................................... 100 5.5 Summary of analyses of carbon foam tooling ...................................................... 103 Chapter 6: CONCLUSIONS AND RECOMMENDATIONS FOR FUTURE WORK . 105 References ....................................................................................................................... 108 8 Appendix A: RELATIONS FOR THERMAL AND MECHANICAL PROPERTIES OF COMPOSITE MATERIALS .......................................................................................... 116 Appendix B: MATHEMATICAL MODEL FOR TURBULENT FLUID FLOW AND HEAT TRANSFER ........................................................................................................ 120 9 LIST OF TABLES Page Table 3.1: Non-dimensional bulk thermal conductivity of 90% porosity carbon foam .............................................................................34 Table 3.2: Non-dimensional bulk Young’s modulus of 90% porosity carbon foam .............................................................................41 Table 4.1: Permeability and inertial coefficient of various porous materials .................62 Table 4.2: Increase in effective heat transfer coefficient of porous channel relative to clear channel ..................................................................................70 Table 5.1: Thermal properties of AS4/3501-6 prepreg and its components (Lee et al., 1982; Loos & Springer 1983).......................................................84 Table 5.2: Thermal properties of tooling materials (“GrafoamTM Carbon Foam Solutions”, n.d.; Wiersma et al., 1998) .............84 Table 5.3: Weight factors and relaxation times for 3501-6 epoxy resin (Kim & White, 1996) .....................................................................................88 Table 5.4: Mechanical properties of AS4/3501-6 prepreg components (Bogetti & Gillespie, 1992; Kim & White, 1996; White & Kim, 1998) ........88 Table 5.5: Mechanical properties of various tooling materials (“GrafoamTM Carbon Foam Solutions”, n.d.; Wiersma et al., 1998) .............89 Table 5.6: Bent composite part spring-in angle ..............................................................93 Table 5.7: Flat composite part curvature ........................................................................94 10 LIST OF FIGURES Page Figure 1.1: SEM picture of graphitic carbon foam (Source: AFRL) ..............................14 Figure 2.1 SEM picture of (a) graphitic carbon foam (Source: AFRL) and (b) Duocel® aluminum foam (Source: “Metal foam” n.d.) ..........................19 Figure 2.2: Various ideal geometry models used to approximate carbon foam microstructure: (a) tetrahedron (Source: Sihn & Roy, 2004), (b) tetrakaidecahedron (Source: Li et al., 2005), (c) centered cube (Source: Yu et al., 2006), (d) BCC type cube (Source: Druma et al., 2004), (e) BCC type ellipse, vertical and horizontal (Source: Druma et al., 2004) .........................................................................20 Figure 2.3: 3D rendering of graphitic carbon foam by serial sectioning technique (Source: AFRL) ............................................................................21 Figure 2.4: 3D solid model of (a) carbon foam and (b) carbon foam saturated with fluid .......................................................................................23 Figure 3.1: Finite element discretization of 3D carbon foam microstructure .................31 Figure 3.2: (a) Temperature and (b) heat flux distributions in carbon foam microstructure when applying the heat flux in the x-direction .....................33 Figure 3.3: (a) Displacement and (b) stress distributions in carbon foam microstructure when applying the compressive load in the x-direction .......40 Figure 4.1: 3D solid model of porous channel................................................................45 Figure 4.2: Boundary conditions applied on the porous channel....................................54 Figure 4.3: Computational mesh for porous channel ......................................................58 Figure 4.4: Fluid flow pathlines in the porous channel colored by velocity magnitude for a free stream velocity of 0.5 m/s ...........................................60 Figure 4.5: Pressure drop per unit length as a function of free stream velocity .............62 Figure 4.6: Friction coefficient as a function of modified Reynolds number .................65 11 Figure 4.7: Fluid temperature distribution in (a) clear channel (no foam) and in (b) porous channel for a free stream velocity of 0.5 m/s ..............................67 Figure 4.8: Effective heat transfer coefficient as a function of free stream velocity and foam solid phase thermal conductivity.....................................69 Figure 4.9: Pumping power per unit volume of fluid as a function of thermal resistance of the channel ..................................................................72 Figure 5.1: CTE of different tooling materials and carbon fiber (Burke, 2003; Burden, 1989; “GrafoamTM Carbon Foam Solutions”, n.d.).........................75 Figure 5.2: Carbon foam tooling for a simple part (Source: GrafTech International Ltd.) ..........................................................77 Figure 5.3: Cure cycle for AS4/3501-6 (Kim & Hahn, 1989) ........................................83 Figure 5.4: (a) 3D solid model and (b) 2D finite element discretization of the composite part on convex and concave tooling ............................................91 Figure 5.5: The deflections and angles due to deformations from processing: (a) spring-in for a 90 degree angle and (b) warpage of a flat composite part. The values are determined after removing the part from the tooling ...............................................................................92 Figure 5.6: Stress distribution along fiber direction at the end of cure cycle before removing the composite part from convex tooling made of (a) carbon foam and (b) steel ........................................................................95 Figure 5.7: Stress distribution along fiber direction at the end of cure cycle after removing the composite part from convex tooling made of (a) carbon foam and (b) steel. A deformation scale factor of 10 is used for displacement ...................................................................................96 Figure 5.8: Stress distribution along fiber direction at the end of cure cycle before removing the composite part from concave tooling made of (a) carbon foam and (b) steel. .......................................................................97 Figure 5.9: Stress distribution along fiber direction at the end of cure cycle after removing the composite part from concave tooling made of (a) carbon foam and (b) steel. A deformation scale factor of 10 is used for displacement ...................................................................................97 12 Figure 5.10: Stress distribution along fiber direction at the end of cure cycle before removing the composite part from flat tooling made of (a) carbon foam and (b) steel ........................................................................99 Figure 5.11: Stress distribution along fiber direction at the end of cure cycle after removing the composite part from flat tooling made of (a) carbon foam and (b) steel. A deformation scale factor of 10 is used for displacement ...................................................................................99 Figure 5.12: Curing process of the composite part on convex tooling made of different materials. The stress distribution along fiber at point (A) is shown for different tooling materials. The curves for the degree of cure at points (A) and (B) overlap, even though the point (B) is insulated by the carbon foam tooling ..........................................................101 13 CHAPTER 1: INTRODUCTION TO CARBON FOAM Carbon foams are rigid, porous materials consisting of an interconnected network of ligaments. They can have open cell structure where pores are interconnected to one another or closed cell structure where pores are isolated from one another. There are several types of carbon foams, depending on the raw material (precursor) and the manufacturing process employed to produce the foam. The major categories are: reticulated vitreous carbon foam (RVC), graphitic carbon foam and non-graphitic carbon foam. In general, porous materials can be considered a subclass of cellular materials. By definition, a cellular material is a mechanical structure made of interconnected struts or plates which form the cells (Gibson & Ashby, 1997). All cellular materials have a certain degree of porosity. The first carbon foams were developed by W. Ford in the 1960s as reticulated vitreous carbon foams by carbonizing thermosetting polymer foams (Gallego & Klett, 2003). In the 1990’s, scientists at Air Force Research Laboratory (AFRL, Dayton, OH) developed graphitic carbon foams by blowing a melted mesophase pitch precursor. After the foaming process, the carbon foam is usually stabilized at 170°C and a heat treatment consisting of carbonization at 1000°C and graphitization at 2700°C is applied (Brow et al., 2003). Thermal conductivity of graphitic carbon foams generally ranges between 1 W/m°C and 250 W/m°C depending on the microstructure, porosity and process parameters (Druma, 2005). An SEM picture of a graphitic carbon foam produced by 14 AFRL is shown in Figure 1.1. The open cell structure due to the windows between adjacent pores is seen in this figure. In this particular case the pores are spherical or elliptical with diameter ranging between 100 and 350 μm. Scientists at Oak Ridge National Laboratory developed an alternative process to manufacture graphitic carbon foams and obtained bulk thermal conductivities as high as 180 W/m°C (Gallego & Klett, 2003). Potential utilizations of highly thermal conductive graphitic carbon foams with open cell structure include thermal management applications such as heat sink and heat exchanger cores. Figure 1.1. SEM picture of graphitic carbon foam (Source: AFRL). A method to manufacture non-graphitic carbon foams was developed by researchers at West Virginia University (Chen et al., 2006). They used inexpensive precursors such as coal, petroleum pitch, coal tar pitch and the result is a carbon foam 15 that can be either very strong mechanically or thermally conductive, depending on the manufacturing process parameters. In addition, the foam is nearly isotropic and it can have open cell structure or closed cell structure (Chen et al., 2006). Potential utilizations of non-graphitic carbon foams include structural applications such as tooling for composite materials manufacturing, stiffener inserts and core materials for composite sandwich structures. Non-graphitic carbon foams are also attractive as thermal protection materials, as they can be produced with very low thermal conductivity (Spradling & Guth, 2003). Overall, carbon foam is a very versatile material whose properties can be tailored by controlling three major factors: the precursor, the foaming process and the heat treatment conditions (Rowe et al., 2005). Moreover, carbon foams can have very low density and coefficient of thermal expansion (CTE), and can be produced with a wide range of porosities. These properties make carbon foams suitable for utilization in many engineering applications. However, designing with this material requires a good understanding of its thermal and mechanical properties, and the characteristics of fluid flow and convection heat transfer through the foam. Because this foam was developed within the last 10 – 15 years, the complex tridimensional (3D) geometry of its microstructure and associated fluid flow phenomena have not been studied in detail. Bulk thermal and mechanical properties of carbon foam, as well as its fluid flow and convection heat transfer characteristics are strong functions of the pore shape, dimensions and distribution in the solid matrix. They also depend upon the thermal and mechanical properties of the solid phase. 16 The objective of this research work is to develop an understanding of the effect of the microstructure and properties of carbon foam so that the feasibility of applications such as thermal management, composite tooling, etc. can be evaluated. This will be done by first modeling the properties of the carbon foam and then applying the models to the specific applications. Therefore, the objectives in this research are to investigate thermal and mechanical behavior of carbon foam by the finite element method (FEM). Based on the model of the foam, the fluid flow phenomenon through the foam is then investigated by finite volume method. Finally, the application of the foam as a tooling material for manufacturing composite materials is studied. Chapter 2 presents the development of the 3D solid model of carbon foam microstructure and calculation of its geometry parameters (solid volume, porosity and surface area). Chapter 3 presents the thermal and mechanical characterization of carbon foam microstructure. Finite element simulations to evaluate bulk thermal conductivity and bulk Young’s (elastic) modulus of carbon foam microstructure are carried out using the 3D solid model developed in Chapter 2. The finite element results are compared with experimental results from literature in order to assess their accuracy. Fluid flow and convection heat transfer in carbon foam are investigated in Chapter 4. In this chapter, finite volume simulations are used to evaluate pressure drop and effective heat transfer capabilities when fluid flows through the carbon foam. Fluid flow results are compared with experimental results based on various porous materials available in literature in order to assess their accuracy. The convection heat transfer capabilities are evaluated by comparing the effective heat transfer coefficient obtained in a porous channel with that 17 obtained in a clear channel (no carbon foam). The utilization of carbon foam as tooling material for manufacturing carbon fiber reinforced epoxy composites is investigated in Chapter 5. In this chapter, the process induced residual stresses and geometric deformations of composite parts manufactured on traditional tooling materials and carbon foam are compared in terms of warpage and spring-in. A unique aspect of this research is the development of an accurate solid model representation of 3D carbon foam microstructure, which is used to investigate the bulk (effective) thermal and mechanical properties of carbon foam as well as the fluid flow and heat transfer phenomena through the open pores of the foam. 18 CHAPTER 2: SOLID MODELING OF CARBON FOAM MICROSTRUCTURE One of the potential applications of carbon foam is its use in thermal management, including convection heat transfer through its pores. However, because of the manufacturing process, the microstructure of carbon foam is quite complex. Therefore, the study of the relationship between its microstructure and bulk properties is a difficult problem. Carbon foam obtains its microstructure through the processing method in which a pitch precursor is heated to the melting point while being pressurized in an inert atmosphere and then is blown by releasing the gas pressure (Druma, 2005; Brow et al., 2003). The blowing process produces pores that are bubbles which have grown either to form closed cell or open cell structure; in the latter case the walls between the cells have openings that can allow a flow through the structure (see Figure 2.1(a)). Metal foams used for heat exchanger, such as Duocel® produced by ERG Aerospace (see Figure 2.1(b)) have a more open reticulated structure and the pore size and the pore window are virtually identical. Reticulated structures are often modeled as a set of ligaments with a geometric relationship (Calmidi, 1998). 19 Figure 2.1. SEM picture of (a) graphitic carbon foam (Source: AFRL) and (b) Duocel® aluminum foam (Source: “Metal foam” n.d.). Due to its complexity and randomness in the pore shape, dimensions and distribution in the solid matrix, it is difficult to obtain an accurate representation of 3D microstructure of carbon foam in solid modeling software. This is the reason why the carbon foam microstructure is often approximated by idealized geometry. Sihn and Roy (2004) approximated the carbon foam with a unit cell obtained by subtracting four identical spheres from a regular tetrahedron (see Figure 2.2(a)). The spheres are located at the corners of the tetrahedron. By varying the diameter of the spheres, the porosity of the unit cell is varied. Li et al. (2005) extended this concept and proposed a tetrakaidecahedral unit cell for carbon foam (see Figure 2.2(b)). Yu et al. (2006) proposed the unit cell obtained by subtracting a sphere from a cube for the representative elementary volume of carbon foam (see Figure 2.2(c)). The sphere is located at the center of the cube. Druma et al. (2004) proposed a body centered cubic (BCC) type structure for the unit cell of carbon foam where nine spheres of equal volumes are subtracted from a cube (see Figure 2.2(d)). The spheres are located at the corners and the center of the cube. They also used ellipses (horizontal and vertical) to create the pores (see Figure 2.2(e)). 20 (a) (b) (c) (d) (e) Figure 2.2. Various ideal geometry models used to approximate carbon foam microstructure: (a) tetrahedron (Source: Sihn & Roy, 2004), (b) tetrakaidecahedron (Source: Li et al., 2005), (c) centered cube (Source: Yu et al., 2006), (d) BCC type cube (Source: Druma et al., 2004), (e) BCC type ellipse, vertical and horizontal (Source: Druma et al., 2004). 21 In this research work, an accurate representation of the 3D microstructure of carbon foam will be used. The foam was manufactured at AFRL and it had a bulk density of 0.24 g/cc (Alam & Maruyama, 2004) which correspond to approximately 89% porosity. The digital image of the foam was obtained by automated serial sectioning technique with Robo-Met.3D, a process presented in detail by Maruyama et al. (2006) and briefly described as follows: (i) the pore space in the carbon foam sample is filled with epoxy resin to give structural support and serial layers of approximately 3.5μm thickness are removed; (ii) a digital image of the cross-section of the carbon foam is taken after each cut; (iii) the collection of digital images obtained is assembled using a custom software. The result is a file containing a point cloud that resembles the geometry of the carbon foam, as shown in Figure 2.3. The AFRL supplied the 3D rendering of the carbon foam structure for this research work. Figure 2.3. 3D rendering of graphitic carbon foam by serial sectioning technique (Source: AFRL). 22 The computational effort required to set up and run a model that reflects accurate details at the pore level of the microstructure requires large computing resources. The ideal model should be exactly representative of the bulk material, i.e. it should be large enough to have the same properties (density, microstructure, thermal and mechanical properties, etc.) as the bulk material. Since the thermal and mechanical properties of the bulk foam material itself tend to be non-homogeneous and anisotropic, it is quite difficult to select a reasonably small volume of the foam that can reflect all of the bulk properties. To keep the computational effort feasible and within the constraints of the computational resources, only a quarter of the 3D rendering of carbon foam shown in Figure 2.3 was converted to a solid model. Surface reconstruction using reverse engineering software (Geomagic Studio) starts from the point cloud representation of the carbon foam (see Figure 2.3). The final result is a file containing a closed surface that resembles the geometry of the carbon foam microstructure. The file is imported into a solid modeling software and the 3D solid model of carbon foam microstructure is obtained, as shown in Figure 2.4(a). It is estimated that the model contains approximately 20 to 40 pores. The faces of the carbon foam solid model are trimmed to obtain flat surfaces. For the convection model with fluid flow, a boolean operation is carried out to subtract the shape of the foam from a parallelepiped representing the fluid. The foam is subsequently fitted into this empty space and aligned to the sides. The assembly consisting of carbon foam saturated with the fluid is shown in Figure 2.4(b). 23 (a) (b) Figure 2.4. 3D solid model of (a) carbon foam and (b) carbon foam saturated with fluid. 24 The geometric parameters of the carbon foam can be calculated using the 3D solid model representation of the microstructure. The dimensions of the carbon foam microstructure bounding parallelepiped in Figure 2.4(a) are 1.502 x 1.482 x 1.540 mm 3 . The accuracy of this measurement is determined by the scanning process in which 3.5 μm slices were used to build the digital point cloud representation. Therefore, the accuracy of the length scale in the solid model is of the order of 1 μm. The calculations in the model will have this scale of accuracy; however experimental results are expected to have much lower accuracy in the length scale. The porosity (void content) of a porous material is defined as φ (% ) = Vf V ⋅ 100 = Vf V f + Vs ⋅ 100 (2.1) where V f is the volume of the pores and Vs is the volume of the solid phase. The porosity of the carbon foam calculated using the solid model in Figure 2.4(a) is 90.31%. This is very close to the experimental porosity, which was measured to be approximately 89%. Therefore this sample was considered to be representative of the foam. Experimental determination of porosity can be done by using the following relation, which can be derived from Equation (2.1): ⎛ φ (% ) = ⎜⎜1 − ⎝ ρb ⎞ ⎟ ⋅ 100 ρ s ⎟⎠ (2.2) where ρ b is the bulk (effective) density and ρ s is the solid phase density (Druma, 2005). The surface area per unit volume of a porous material is defined as 25 a sf = Asf V = Asf V f + Vs (2.3) where Asf is the total pore surface area. The surface area per unit volume of the carbon foam calculated using the solid model in Figure 2.4(a) is 4339.5 m 2 m 3 . Since the length scale accuracy is of the order of 1 μm the surface area calculation has high accuracy for a given solid model that is used in the calculation. However it should be recognized that there are inaccuracies that are inherent in the transformation of the digital cloud representation to the surfaces in the solid model. It should be mentioned that experimental determination of surface area per unit volume of a porous material is a complex measurement with significant errors. 26 CHAPTER 3: THERMAL AND MECHANICAL CHARACTERIZATION OF CARBON FOAM MICROSTRUCTURE 3.1 Thermal analysis 3.1.1 Introduction and objective A significant amount of research on thermal analysis of porous materials has been done using highly porous aluminum foams. Calmidi and Mahajan (1999) developed a correlation for bulk thermal conductivity of aluminum foam saturated with air and water using an analytical model validated with experimental results. The foam was approximated with a 2D periodic hexagonal structure. Boomsma and Poulikakos (2001) extended this procedure and developed a bulk thermal conductivity model for aluminum foam saturated with fluid by approximating the aluminum foam with a 3D structure, the tetrakaidecahedron. Thermal conductivity analysis of idealized structure of carbon foam was carried out by Druma et al. (2004). They developed finite element models and calculated bulk thermal conductivity of carbon foam as a function of porosity, however they did not account for the effect of convection and radiation within the pores. The pore shapes were taken to be spherical and elliptical (horizontal and vertical orientation). Yu et al. (2006) developed a representative elementary volume model for heat transfer and fluid flow in carbon foam. An analytical correlation for the bulk (stagnant) thermal conductivity of carbon foam saturated with fluid is developed by taking the 27 porous structure to be a cube with a spherical centered pore and equating it to a square bar structure with the same porosity. Klett et al. (2004) proposed the following correlation for bulk thermal conductivity of graphitic carbon foam by fitting a curve through a set of experimental data points: ⎛ρ k e = 0.734⎜⎜ b ⎝ ρs ⎞ ⎟⎟ ⎠ 1.427 ks (3.1) where ρ b is the bulk density of carbon foam, ρ s is the density of the solid phase of the carbon foam and k s is the thermal conductivity of the solid phase of the carbon foam. The thermal conductivity of the solid phase depends on the degree of graphitization and ranges from ~1 to 2000 W/m°C (Alam & Maruyama, 2004). As mentioned earlier, the bulk thermal conductivity of graphitic carbon foams generally ranges between 1 W/m°C and 250 W/m°C depending on the microstructure, porosity and process parameters (Druma, 2005). Bauer (1993) developed an analytical model for thermal conductivity of porous media by applying a perturbation to a uniform temperature gradient. The perturbation is given by pores distributed in a homogeneous and isotropic material. For the particular case of spherical pores and thermal conductivity of gas in the pores much smaller than that of solid phase, the bulk thermal conductivity of the porous material is: φ ⎞ ⎛ k e = ⎜1 − ⎟ ⎝ 100 ⎠ 1.295 ks (3.2) 28 Alam and Maruyama (2004) tested the bulk thermal conductivity of carbon foam samples made by various manufacturers using laser flash and hot plate techniques. Their results showed a wide range of thermal conductivity of foams made by different manufacturers. The differences were due to differences in starting materials and foaming processes. It is known that graphitic foams made by a foaming process are thermally anisotropic and this was confirmed by bulk thermal conductivity measurements (Alam & Maruyama, 2004). The objective of the thermal analysis in the present work is to determine the bulk thermal conductivity of the carbon foam microstructure. The 3D solid model of carbon foam microstructure developed in Chapter 2 (see Figure 2.4(a)) is used to carry out the analysis. A study by Druma et al. (2005) investigated the bulk thermal conductivity of carbon foam microstructure using a smaller solid model. In the current analysis, the solid phase of the carbon foam is assumed homogeneous and isotropic. It is important to note that this is a simplification of the micro-level properties of graphitic carbon foam. In graphitic low density foams the cell walls can be highly anisotropic along the ligaments because they are composed of aligned graphene planes (Druma, 2005). These ligaments meet at thicker nodes where the structure is much less graphitic (Druma, 2005). In the current analysis, the ligaments and nodes will be assumed to be isotropic and homogeneous. Most of the studies on the modeling of carbon foam have used this assumption because of the computational complexity of a model that incorporates the variation of properties along the ligaments and nodes. 29 3.1.2 Methodology and simulations The carbon foam microstructure subjected to a thermal load is analyzed by finite element method. The results of the analysis allow for calculation of bulk thermal conductivity of carbon foam microstructure. The 3D solid model of carbon foam is sandwiched between two solid plates which will be used for applying the boundary conditions. A perfect thermal contact (zero thermal contact resistance) is assumed between plates and carbon foam. The pores are assumed empty and convection and radiation heat transfer are not considered in this analysis. Constant thermal conductivity of solid phase of the carbon foam and solid plates is assumed throughout the analysis. The temperature field in the system (carbon foam microstructure and plates) is governed by the steady state conduction heat transfer equation without internal heat generation ∂ ∂xi ⎛ ∂T ⎜⎜ k ⎝ ∂xi ⎞ ⎟⎟ = 0 ⎠ i = 1, 2, 3 (3.3) where T is the temperature and k is the thermal conductivity of the domain analyzed. The following boundary conditions are applied for thermal analysis: - a uniform heat flux on the top surface of the upper plate - a uniform temperature on the bottom surface of lower plate - temperature and heat flux continuity at the plate – carbon foam microstructure interface - all of the other surfaces are thermally insulated The thermal conductivity of the two solid plates is chosen to be three orders of magnitude higher than that of the solid phase of carbon foam in order to ensure constant temperature boundary conditions on the carbon foam top and bottom surfaces. 30 The assembly consisting of the 3D solid model of carbon foam microstructure and the two solid plates was meshed using 10-node quadratic tetrahedral elements. The general purpose finite element code ABAQUS is used to solve Equation (3.3) using Galerkin weighted residual method (Abaqus Inc., 2006). The solution of finite element analysis yields the temperature and heat flux distributions in the carbon foam microstructure and solid plates. Mesh independence of the finite element solution was established by solving the problem using two different mesh densities: 519,397 elements (see Figure 3.1) and 1,357,568 elements. The difference between the two meshes in terms of the bulk thermal conductivity is less than 0.1%. The results presented in the next section are obtained using the coarser mesh. 31 Figure 3.1. Finite element discretization of 3D carbon foam microstructure. The bulk thermal conductivity of carbon foam can be calculated by applying Fourier’s law for 1D conduction heat transfer: q x = −k e dT ΔT = ke dx Δx (3.4) where qx is the heat flux applied on the top surface of the upper plate, ΔT is the temperature difference between the two solid plates and Δx is the thickness of the carbon foam in the heat flux direction. 32 The non-dimensional bulk thermal conductivity of carbon foam is calculated as follows k eff = ke ks (3.5) to facilitate comparison with results reported in literature. Independent analyses are carried out in the x- y- and z- directions of the 3D solid model of carbon foam to account for randomness in pore shape, dimensions and distribution in the solid matrix. The thermal conductivity of the solid phase has been varied to study its influence on the bulk thermal conductivity of the microstructure. 3.1.3 Results and conclusions Figure 3.2(a) and (b) show the temperature and heat flux distributions in carbon foam microstructure when the heat flux is applied in the x- direction. It can be noticed that the temperature gradient is predominantly one dimensional in the heat flux direction. The heat flux distribution shows higher values in the thin ligaments of the carbon foam microstructure. Similar temperature and heat flux distributions are obtained when applying the heat flux in the y- and z- directions. 33 (a) (b) Figure 3.2. (a) Temperature and (b) heat flux distributions in carbon foam microstructure when applying the heat flux in the x-direction. 34 The thermal analysis results are summarized in Table 3.1 for comparison with similar results reported in literature. The bulk thermal conductivity values obtained by FEM analysis on real geometry of carbon foam microstructure are in very good agreement with the experimental results in Klett et al. (2004). It should be noted that these experimental values were obtained for a graphitic foam which is similar but not identical to the foam modeled in this study. The analytical solution proposed by Bauer (1993) and FEM analysis on ideal geometry developed by Druma et al. (2004) overpredict the experimental value in Klett et al. (2004) except for the case of ellipsoidal pores perpendicular to the heat flux. Table 3.1 Non-dimensional bulk thermal conductivity of 90% porosity carbon foam Type of k eff (% of solid phase) Source analysis Klett et al. (2004) Experimental 2.6 (Equation 3.1) 6.25 (spherical pores) FEM on 2.2 (ellipsoidal pores – perpendicular to heat flux) Druma et al. (2004) ideal 9.2 (ellipsoidal pores – aligned with heat flux) geometry Bauer (1993) Analytical 4.9 (spherical pores) (Equation 3.2) 2.78 (x-direction) FEM on real 2.93 (y-direction) 2.72 (average) Present study geometry 2.44 (z-direction) It can be concluded that, for the same porosity, the bulk thermal conductivity is a strong function of the particular geometry of the microstructure. Ideal geometries used by researchers do not seem to have the capability of predicting the bulk thermal conductivity 35 accurately. The lower value of the non-dimensional thermal conductivity in the true geometry is probably due to the higher degree of tortuosity caused by the randomness in pore shape, dimensions and distribution in the solid matrix. The differences between the bulk thermal conductivity values obtained when analyzing the foam in the three directions (see Table 3.1) indicate that the particular solid model of carbon foam microstructure analyzed is not thermally isotropic. This raises the question whether the model volume is representative of the bulk foam. It is well known that most carbon foams do not exhibit bulk isotropy in thermal conductivity (Alam & Maruyama, 2004); however the origin of the anisotropy could be a combination of the anisotropy of the geometry and the anisotropy of the thermal properties in the ligaments and nodes. Therefore, it is quite difficult to select a reasonably small volume of the foam that can reflect all of the bulk properties. The FEM analysis also exhibited a linear relationship between the bulk thermal conductivity and the thermal conductivity of the solid phase of the carbon foam. 3.2 Mechanical analysis 3.2.1 Introduction and objective Stress-strain analysis of idealized structure of carbon foam was carried out by Sihn and Roy (2004). They developed a finite element model and calculated bulk Young’s modulus and Poisson’s ratio of carbon foam as a function of porosity. The carbon foam unit cell was approximated with tetrahedral type geometry. Li et al. (2005) approximated the carbon foam unit cell with a tetrakaidecahedron and calculated bulk 36 Young’s modulus, Poisson’s ratio and shear modulus as a function of porosity for various cross-sectional shapes of the cell struts. Sihn and Rice (2003) tested the bulk Young’s modulus of carbon foam samples made by various manufacturers under compressive loading. For a 90% porosity foam they reported a bulk value that ranged from 0.4% to 1% of the intrinsic solid phase modulus. The objective of the mechanical analysis in the present work is to determine the bulk Young’s modulus of carbon foam microstructure. The analysis is carried out by using the solid model of carbon foam microstructure developed in Chapter 2 (see Figure 2.4(a)). A study by Druma et al. (2005) investigated the bulk Young’s modulus of carbon foam microstructure using a smaller solid model. The solid phase is assumed to have homogeneous isotropic modulus value. This is a simplification of the non-homogeneous properties of the solid phase. This assumption has been made by Sihn and Roy (2004), Li et al. (2005) and several other researchers. 3.2.2 Methodology and simulations The carbon foam microstructure subjected to a mechanical load is analyzed by finite element method. A linear elastic model with constant material properties and accounting only for small deformations (no large deformations and geometric nonlinearities) is employed, which is consistent with the brittle nature of graphitic carbon foams. The results of the analysis allow for calculation of bulk Young’s (elastic) modulus of carbon foam microstructure. Two solid plates are used to apply the boundary 37 conditions on the carbon foam microstructure sandwiched between them. A rigid mechanical contact (no penetration) is assumed between the solid plates and carbon foam in the normal direction. In the tangential direction, the mechanical interaction between the carbon foam and plates is frictionless in order not to introduce restrictions in deformation. The displacement field in the system (carbon foam microstructure and plates) is governed by the force equilibrium equation ∂ (σ ij ) = 0 ∂x j i = 1, 2, 3 (3.6) where σ ij is the stress tensor (symmetric). The body forces are neglected. The constitutive behavior of the solid phase of the carbon foam relates stress and strain and is modeled using linear elasticity: σ ij = Cijkl ε kl i, j , k , l = 1,2,3 (3.7) where C ijkl are the stiffness matrix components and ε kl are the displacement vector components. The C ijkl values are calculated using the relations in Appendix A with isotropic material properties. The following boundary conditions are applied for stress-strain analysis: - the bottom surface of lower plate is fully constrained - two adjacent sides of the carbon foam are constrained in their normal directions in order to avoid translation - two adjacent sides of the upper plate are constrained in their normal directions in order to avoid translation - a normal uniform compressive load on the top surface of the upper plate 38 In order to reduce the influence of the plates on the results, the Young’s modulus of the plates material was chosen to be three orders of magnitude higher than that of solid phase of the foam. Therefore, the strain of the plates is negligible in comparison with the strain of the carbon foam. The mesh created for thermal analysis (see Figure 3.1) is also used for mechanical analysis. The general purpose finite element software ABAQUS is used to solve Equation (3.6) using the virtual work principle (Abaqus Inc., 2006). The solution of the finite element analysis yields the displacement and stress distributions in the carbon foam microstructure and solid plates. Mesh independence of the finite element solution was established in a manner similar to the one used for thermal analysis. The difference between the two meshes (519,397 elements vs. 1,357,568 elements) in terms of the bulk Young’s modulus is less than 0.4%. The results in the next section were obtained by performing the analysis with the coarser mesh. The Young’s modulus of bulk foam can be obtained by applying Hook’s law for 1D stress situation: σ x = Eeε x = Ee ΔL L0 (3.8) where σ x is the uniform normal compressive load on the top surface of the upper plate, ε x is the strain in the x- direction, L0 is the initial (undeformed) thickness of the carbon foam and ΔL is the displacement of the top plate, both in the x- direction. The non-dimensional bulk value for Young’s modulus of carbon foam is obtained as 39 E eff = Ee Es (3.9) where E s is the Young’s modulus of the solid phase of carbon foam. The nondimensional value is used to facilitate comparison with the results reported in literature. By following the same approach as in the thermal analysis, independent simulations are carried out in the x- y- and z- directions of the 3D solid model of carbon foam. The Young’s modulus and Poisson’s ratio of the solid phase have been varied to study their influence on the bulk Young’s modulus of the microstructure. The two values used for solid phase Poisson’s ratio are ν s = 0.2 and 0.33 (Sihn & Roy, 2004). The values used for Young’s modulus of the solid phase are chosen between 10 and 20 GPa, which is consistent with the values in Sihn and Roy (2004) and Li et al. (2005). The magnitude of the compressive load applied to the model was selected such that the maximum resulted strain is smaller than 0.2%. 3.2.3 Results and conclusions Figure 3.3(a) and (b) show the x- displacement and stress distributions in carbon foam microstructure when the compressive load is applied in the x- direction. It can be noticed that the x-displacement gradient is mostly uniform in the compressive stress direction. The stress distribution shows higher values in the thin ligaments of carbon foam microstructure, which act as stress concentrators. Similar displacement and stress distributions are obtained when applying the compressive stress in the y- and zdirections. 40 (a) (b) Figure 3.3. (a) Displacement and (b) stress distributions in carbon foam microstructure when applying the compressive load in the x-direction. 41 The stress-strain analysis results are summarized in Table 3.2 for comparison with similar results reported in literature. It needs to be mentioned that a range is specified in Table 3.2 for the analysis by Li et al. (2005) because different values are calculated for various cross-sectional shapes of the ligament. It can be seen that the simple regular tetrahedral type geometry used by Sihn and Roy (2004) produces a very different result for Young’s modulus when compared with the experimental values for a variety of foams (Sihn & Rice, 2003). The results are different by a factor of about 6. It is probably due to the complex deformation process that takes place at the pore level when a mechanical load is applied. However, the tetrakaidecahedron, a much more complex geometrical model used by Li et al. (2005), produces much better agreement with experimental values; therefore this particular idealized geometry appears to reflect the stresses and strains in the foam microstructure. Table 3.2 Non-dimensional bulk Young’s modulus of 90% porosity carbon foam E eff (% of solid phase) Source Type of analysis Sihn & Rice (2003) Experimental 0.4 – 1.0 Sihn & Roy (2004) FEM on ideal geometry 3.4 Li et al. (2005) Analytical 0.45 – 0.77 0.596 (x-direction) Present study FEM on real geometry 0.76 (y-direction) 0.623 (average) 0.513 (z-direction) The bulk Young’s modulus values for the true geometry in the present study are in good agreement with both the experimental results in Sihn and Rice (2003) and 42 solution of analytical model by Li et al. (2005). The results of this analysis on the true geometry produced different values of bulk Young’s modulus in the three directions (see Table 3.2). The differences for the three directions could be due to the inherent geometric anisotropy of the foam or due to the small sample size of the foam used in analysis. As discussed earlier, the foam sample has the same porosity that is obtained by experimental measurement of bulk foam samples, so that the foam sample was a taken to be a reasonable representation of the bulk material. As expected, the bulk Young’s modulus for the true geometry showed a linear variation with Young’s modulus of the solid phase; so the non-dimensional values in Table 3.2 did not change. The change of Poisson’s ratio of solid phase from ν = 0.2 to 0.33 showed negligible variation in Young’s modulus of the bulk foam. This has also been reported by Sihn and Roy (2004). 43 CHAPTER 4: FLUID FLOW AND CONVECTION HEAT TRANSFER IN CARBON FOAM 4.1 Introduction In this chapter, the application of carbon foam in thermal management is studied by considering a coolant flow through the pores of the foam. This study is motivated by the continuous increase in thermal power dissipated by electronic devices, which requires more efficient cooling solutions. Traditionally, the high density electronic circuits use heat sinks made of a metallic finned structure (copper or aluminum) cooled by natural or forced convection. However, the requirement for increasingly powerful cooling devices has boosted the research and development of advanced heat sinks such as micro-channels and heat pipes. The interconnected pore structure of porous materials allows for fluid flow and also offers a significant increase in surface area available for convective heat transfer. Highly porous aluminum foams have been investigated as possible solutions for thermal management of electronics (Bhattacharya & Mahajan, 2002; Boomsma et al., 2003). Because of its reduced weight and high thermal conductivity, carbon foam is considered as potential candidate for heat sinks and heat exchangers core. Typical porous media heat exchange system consists of a channel filled with an open cell porous material and saturated with a coolant that flows through the pores. In order to be able to design cooling systems based on a porous material, its characteristics must be known in terms of pressure drop and convective heat transfer coefficient when a fluid flows through it. 44 This chapter is focused on the analysis of fluid flow and convection heat transfer in 3D porous channel using the finite volume method as implemented in the FLUENT software (Fluent Inc., 2006). Carbon foam permeability, inertial coefficient and friction coefficient will be calculated from the simulation results of pressure drop and free stream fluid velocity. Comparison with experimental results available in literature will be used to validate the results of our simulations. The effective heat transfer coefficient will also be calculated from the results of heat transfer analysis in the porous channel. It will be compared with the convection heat transfer coefficient obtained for the case of clear channel (no foam) in order to quantify the enhancement in heat transfer due to the presence of the carbon foam in the channel. The assembly consisting of carbon foam saturated with fluid discussed in Chapter 2 (see Figure 4.1) is used to carry out this analysis. The fluid part of the model is extended beyond the carbon foam on two opposite sides in order to create clear inlet and outlet for the porous channel. 45 Carbon foam Interstitial fluid Flow direction Figure 4.1. 3D solid model of porous channel. 4.2 Volume averaged model Analytical and numerical modeling of transport phenomena (heat transfer and fluid flow) in porous media is difficult due to its totally irregular structure. Not only is the typical porous media non-homogeneous, but the pores are irregular in geometry and dimension, and randomly distributed in the solid matrix. The traditional approach in modeling transport phenomena in porous media employs local volume averaging in formulation of mass, momentum and energy conservation equations (Vafai & Tien, 1981; Calmidi, 1998). Under this approach, the 46 quantities associated with the fluid flow are averaged over a representative elementary volume consisting of both interstitial fluid and porous material (Vafai & Tien, 1981). When modeling the heat transfer between porous material and interstitial fluid, the two energy equation approach must be used (one for each phase) and the temperatures are averaged separately for solid phase and fluid phase (Calmidi, 1998). This averaging approach reduces the complexity of the general problem, but information about the transport phenomena at pore level and the influence on the overall transport phenomena are lost (Vafai & Tien, 1981; Calmidi, 1998). 4.2.1 Fluid flow model The steady state mass and momentum conservation equations for incompressible flow through porous media are (Hunt & Tien, 1988) ∇⋅ u = 0 ρ u ⋅ ∇u = −∇ p φ2 where (4.1) f + ρ cf μ 2 μ ∇ u− u − u u φ K K (4.2) is the volume average symbol, u is the fluid velocity vector, p is the fluid pressure, ρ is the fluid density, μ is the fluid viscosity, φ is the porous media porosity, K is the porous media permeability and c f is the porous media inertial coefficient. The mass and momentum conservation equations can be solved numerically for pressure and velocity fields provided that φ , K and c f are known. It needs to be mentioned that in comparison to Navier-Stokes equations, the momentum conservation equation for flow through porous media contains two extra 47 terms, which are the last two terms on the right hand side of Equation (4.2). They account for additional pressure loss due to the presence of the porous media. The second term on the right hand side of Equation (4.2) accounts for pressure loss due to the presence of the solid wall boundary. For steady state fully developed flow through porous media without a solid wall boundary, the momentum conservation equation can be expressed in a simpler 1D form (Darcy-Forchheimer equation) which gives the pressure drop when fluid flows through porous media ρ cf 2 Δp μ = uD + uD Δx K K (4.3) where the free stream velocity in the clear channel before (or after) the porous region is given by uD = m& ρAch and Δx is the length of the porous region in the direction of the flow, m& is the mass flow rate and Ach is the cross sectional area of the channel filled with porous material. The fluid flow regime in clear pipes and channels with constant flow area is established as either laminar or turbulent using the Reynolds number based on the equivalent hydraulic diameter which is the length scale of the flow. In porous channels it is difficult to define an equivalent hydraulic diameter because the flow area is continuously changing. The square root of the permeability ( K ) represents the length scale of flow through porous media and has been used in defining the “modified” 48 Reynolds number (Beavers & Sparrow, 1969; Paek et al., 2000; Boomsma & Poulikakos, 2002) Re K = ρu D K μ (4.4) The fluid flow in porous media is in the “Darcy regime” when the velocities are small enough so that the inertial effects are insignificant and the quadratic term in DarcyForchheimer equation is negligible. At high velocities, the quadratic term in DarcyForchheimer equation cannot be neglected and the flow regime is “non-Darcy”. Equation (4.3) has been used to determine K and c f from experimental measurements of pressure drop and free stream velocity for a large variety of porous materials: aluminum, nickel and carbon foams (Hunt & Tien, 1988), compressed aluminum foam (Antohe et al., 1997) and carbon foam (Straatman et al., 2007). Experimental testing on aluminum foams showed that permeability of a porous material is a strong function of porosity and pore size, and inertial coefficient is influenced by the solid phase shape and pore structure (Paek et al., 2000). In an attempt to unify the pressure loss characteristics for various porous media, a dimensionless friction coefficient has been used (Beavers & Sparrow, 1969; Paek et al., 2000) Δp K Δ x f = ρu D2 (4.5) 49 Equation (4.5) was used to determine f from experimental measurements of pressure drop and free stream velocity for aluminum foams (Paek et al., 2000) and Foametal (Vafai & Tien, 1982). Beavers and Sparrow (1969) combined together Equations (4.3), (4.4) and (4.5) and obtained the following relation for friction coefficient: f = cf + 1 Re K (4.6) It must be noticed that in the Darcy flow regime, the quadratic term in Equation (4.3) can be neglected, so that the friction coefficient becomes equal to the inverse of Reynolds number. Based on Equation (4.6), semi-empirical correlations for friction coefficient were developed for aluminum foams by Paek et al. (2000) f = 0.105 + 1 Re K (4.7) and for Foametal by Vafai and Tien (1982) f = 0.057 + 1 Re K (4.8) 4.2.2 Heat transfer model Two distinct approaches have been used in studying convection heat transfer in porous media: (i) thermal equilibrium of the solid and fluid phases when only one equation can be used to model the thermal transport (Vafai & Tien, 1981) and (ii) thermal non-equilibrium, when the two phases have different temperatures and one equation is 50 necessary for each phase (Calmidi, 1998; Calmidi & Mahajan, 2000). In the first approach, there is only one temperature in the system, which is averaged over a representative volume consisting of both solid and fluid, so there is no thermal transport between phases. This approach is not accurate when the thermal conductivities of the two phases are very different, such as aluminum foam and air (Calmidi, 1998). The steady state energy equations for porous material and fluid flowing through it in the thermal non-equilibrium approach are (Calmidi, 1998) 0 = k se∇ 2Ts − hsf Asf (Ts − T f ) (4.9) ρc p u ⋅ ∇T f = (k fe + kd )∇ 2T f + hsf Asf (Ts − T f ) (4.10) where Ts and T f are the solid and fluid temperatures, c p is the fluid specific heat, k se and k fe are the solid and fluid bulk (stagnant) thermal conductivities, kd is the fluid dispersion thermal conductivity, hsf is the interfacial heat transfer coefficient and Asf is the interfacial surface area available for heat transfer in the porous channel. The energy conservations equations can be solved for solid and fluid temperature fields provided that k se , k fe , kd , hsf and Asf are known. Calmidi and Mahajan (2000) developed semi-empirical correlations for dispersion thermal conductivity and interfacial heat transfer coefficient. The empirical parameters were determined by curve fitting through experimental results for aluminum foam. Straatman et al. (2007) used the same type of correlation for interfacial heat transfer coefficient as Calmidi and Mahajan (2000) but they determined the empirical parameter by curve fitting through experimental results for carbon foam. 51 An overall energy balance is generally employed in order to calculate the effective heat transfer coefficient from the heated surface area in a porous channel: Q = m& c p (Tout − Tin ) = ∫ h (T sf s − T f )dA = heff Ab ΔTLM (4.11) Atot where the log mean temperature difference between solid and fluid is ΔTLM = Tout − Tin T − Tin ln 0 T0 − Tout and Q is the heat flux transferred to the fluid in the porous channel, Tin and Tout are the channel inlet and outlet fluid temperatures and T0 is the heat source constant temperature applied to the porous channel. The local interfacial heat transfer coefficient hsf is highly dependent on fluid velocity and porous material geometry and varies from point to point on the porous material surface. It is calculated taking into account the entire surface area available for heat transfer in the porous channel, which is: Atot = Asf + φAb where Asf is the porous material surface area and Ab is the upper (base) surface area of the channel. The effective heat transfer coefficient heff is calculated by taking into account the base surface area of the porous channel Ab rather than the entire surface area available for heat transfer. This reflects the enhancement in heat transfer relative to the base area due to the presence of the porous material in the channel. 52 In this research we will use the direct simulation model because the 3D microstructure of the porous material can have significant effect on the thermal and mechanical properties and also affects the heat transfer and fluid flow through the foam. 4.3 Direct simulation of fluid flow and heat transfer 4.3.1 Ideal pore geometry The pore level simulation of fluid flow and heat transfer in porous media accounts for the microscopic phenomena by modeling the geometry of the porous material microstructure. The geometry of the porous media is often approximated with an ideal geometry. For example, Krishnan et al. (2006) calculated permeability and friction factor from direct simulation of fluid flow through metal foam with periodic boundary conditions. Their representative elementary volume is created by subtracting nine spheres of equal volumes from a cube. The spheres are located at the corners and the center of the cube. This geometry is the same as one of the several geometries studied by Druma et al. (2004). Krishnan et al. (2006) also calculated the heat transfer in terms of Nusselt number by applying a constant heat flux boundary condition on the foam surface, neglecting the conduction heat transfer through the solid phase. Karimian and Straatman (2008) studied the hydraulic and thermal characteristics of porous material assuming spherical pores and developed correlations for pressure drop and heat transfer based on simulations results. The most important simplification of these two models is that they use ideal periodic geometries to represent the porous material microstructure. 53 4.3.2 Real pore geometry The porous material – fluid system used in the present analysis (see Figure 4.1) consists of a rectangular channel filled with a real 3D representation of carbon foam microstructure and saturated with coolant (air) flowing through the foam. The channel walls can be used as virtual walls in the simulations by using symmetry conditions at these boundaries. The effect of such a boundary condition is to extend the flow regime beyond the channel walls and to eliminate the effect of the channel walls on the flow. Symmetry conditions will be used on all surfaces of the channel (see Figure 4.2). For the heat transfer, a constant temperature heat source is applied on the upper surface of the channel to both carbon foam and fluid. The surfaces belonging to channel inlet and outlet are not heated. Heat is transferred by convection from the channel upper wall to the fluid and also by conduction through the solid phase of the foam and then convection from the foam to the fluid. The system is analyzed considering steady state fluid flow and heat transfer by direct simulation on the real scale model. Fluid velocity and pressure distributions, as well as temperature distribution in carbon foam and fluid at pore level are obtained from numerical simulation by commercial software FLUENT (Fluent Inc., 2006). The mathematical model used for fluid flow and heat transfer calculations, as implemented in FLUENT is summarized in the next section. The above approach was adopted in a preliminary study by Anghelescu and Alam (2006) to investigate the convection heat transfer in carbon foam microstructure by using a smaller solid model. 54 Surface not heated Surfaces heated (solid and fluid) Surface not heated Flow direction Symmetry (also on the bottom and on the back surfaces) Figure 4.2. Boundary conditions applied on the porous channel. 4.3.3 Fluid flow computational model The velocity and pressure fields of an incompressible, steady state, Newtonian fluid flow in laminar regime are governed by the instantaneous mass conservation equation ∂ui =0 ∂xi i = 1, 2, 3 and the Navier-Stokes (momentum conservation) equations (4.12) 55 ρu j ∂ui ∂p ∂ ⎛⎜ ∂ui ⎞⎟ =− +μ ∂x j ∂xi ∂x j ⎜⎝ ∂x j ⎟⎠ i, j = 1, 2, 3 (4.13) where ui are the fluid velocities, p is the fluid pressure, μ is the fluid viscosity and ρ is the fluid density. The fluid flow is assumed viscous and the body forces on the fluid are neglected. Due to the complexity of the geometry, a large number of finite volume elements have to be used. Therefore, to reduce computational effort, the thermophysical properties of the fluid are assumed constant with temperature. The following boundary conditions are applied for fluid flow analysis: - constant fluid velocity at the inlet of the porous channel - constant fluid pressure at the outlet of the porous channel - zero fluid velocity (no-slip) on carbon foam surfaces - symmetry on top, bottom and side surfaces of the porous channel Symmetric boundary conditions are applied on all outer surfaces of the porous channel except for inlet and outlet. As explained earlier, this approach extends the dimensions of the porous media and avoids solid wall boundaries which would increase the pressure drop in the porous channel (due to additional friction) and affect the accuracy of permeability, inertial coefficient and friction coefficient calculations for the porous microstructure. The fluid flow model yields the velocity and pressure distributions in the fluid as a function of space coordinates x, y and z . Simulations are carried out by varying the velocity at the inlet of the channel. The carbon foam permeability, inertial coefficient and the relation between the friction coefficient and the modified Reynolds number 56 (f − Re K ) will be determined from simulation results using Equations 4.3 and 4.5. The f − Re K relation will be compared with Equations (4.7) and (4.8). 4.3.4 Convection heat transfer computational model The steady state fluid temperature distribution in laminar flow regime is governed by the energy conservation equation ρc p ∂ (uiT f ) = ∂ ∂xi ∂x j ⎛ ∂T f ⎞ ⎜k f ⎟ ⎜ ∂x ⎟ j ⎝ ⎠ (4.14) where k f is the fluid thermal conductivity. The steady state temperature distribution in the carbon foam is governed by the energy conservation equation ∂ ⎛⎜ ∂Ts ⎞⎟ =0 ks ∂x j ⎜⎝ ∂x j ⎟⎠ (4.15) where k s is the thermal conductivity of the solid phase. The following boundary conditions are applied for heat transfer analysis: - constant temperature on the upper surface of the porous channel (carbon foam and fluid) - constant fluid temperature at the inlet of the channel - symmetry on the sides and lower surface of the channel - temperature and heat flux are continuous at the solid – fluid interface 57 Different from fluid flow computational model, a wall boundary condition is used on the top surface of the porous channel in order to apply the constant temperature heat source (see Figure 4.2). The two energy equations solved together yield the temperature distributions in carbon foam and fluid as a function of space coordinates x, y and z . The fluid flow computational model provides the velocity distributions necessary for solving fluid energy conservation equation. The thermophysical properties of the fluid are assumed constant with temperature during the convection heat transfer analysis so that the fluid flow governing equations and the energy conservation equations can be decoupled and solved independently. Simulations are carried out by varying the intrinsic thermal conductivity of the solid phase in the carbon foam to study its influence on the effective heat transfer coefficient in the porous channel. The effective heat transfer coefficient at the base area of the porous channel will be calculated from simulation results using Equation (4.11). 4.4 Results and conclusions 4.4.1 Fluid flow analysis The solid model assembly of porous channel (carbon foam microstructure and interstitial fluid) is discretized using unstructured tetrahedral mesh in commercial software HYPERMESH (Altair Engineering Inc., 2007) as showed in Figure 4.3. The mesh created is imported into computational fluid dynamics (CFD) software FLUENT for fluid flow and heat transfer calculations. The unstructured tetrahedral mesh is first 58 converted to polyhedral mesh in order to improve the mesh quality (eliminate bad elements) and reduce the computational time. In the polyhedral mesh 367,462 elements are generated for the solid phase and 1,148,766 elements for the fluid. The second order upwind scheme is used for discretizing the momentum and fluid energy conservation equations. The SIMPLE algorithm is used for pressure-velocity coupling. Mesh independence of the finite volume solution was established by solving the problem using a different mesh density: 355,088 elements for solid and 756,387 for fluid. The difference between the two discretizations in terms of permeability and inertial coefficient of carbon foam microstructure is 1.6%, and 1.1%, respectively. The results presented in this chapter are obtained using the finer mesh. Figure 4.3. Computational mesh for porous channel. 59 It must be mentioned that the fluid flow and heat transfer simulation results presented in this chapter were obtained using the mathematical models for laminar flow regime as described in sections 4.3.3 and 4.3.4. The reason behind this is that the Reynolds number calculated based on the equivalent hydraulic diameter of the flow channel is smaller than Re cr = 2320 for the entire range of velocities used in simulations. However, it can be argued that, even at low velocities, the flow in the porous channel might have a certain degree of turbulence at pore level due to the presence of ligaments and pore edges which obstruct part of the flow area and creates tortuous pathlines. In order to account for this possibility the simulations were also run by including turbulence in the flow model. The fluid flow turbulence was modeled using the Reynolds-Averaged Navier-Stokes (RANS) method. The turbulent viscosity and turbulence kinetic energy were modeled by employing the k − ε method as implemented in the commercial code FLUENT (Fluent Inc., 2006). The mathematical model for turbulence is presented in Appendix B. The difference between the laminar and turbulent flow simulation results in terms of permeability and inertial coefficient of carbon foam microstructure is 0.7% and 5.3%, respectively. This is an indication that for higher velocities, the turbulent model needs to be used for accurate simulation results. Fluid flow simulations for inlet (free stream) velocities ranging between 0.01 m/s and 1.5 m/s are run in order to determine the carbon foam microstructure permeability, inertial coefficient and friction coefficient. Fluid thermophysical properties at 20°C are used in the simulations. 60 Figure 4.4 shows the fluid flow pathlines colored by velocity magnitude for a free stream fluid velocity of 0.5 m/s. The fluid flows along the x- axis in the positive direction. The presence of the carbon foam in the channel reduces the cross-sectional area available for fluid flow and increases the fluid velocity. It can be noticed that the maximum fluid velocity in the porous channel is about 3.6 times higher than free stream velocity. This, in turn increases the pressure drop per unit length. The tortuosity of the fluid flow pathlines around the foam ligaments is observed in the simulations as shown in Figure 4.4 Flow direction Figure 4.4. Fluid flow pathlines in the porous channel colored by velocity magnitude for a free stream velocity of 0.5 m/s. 61 The simulation results of pressure drop per unit length across the carbon foam as a function of free stream fluid velocity are shown in Figure 4.5 as discrete points. Pressure drop per unit length for the clear channel is plotted for comparison. The pressure drop obtained in the porous channel filled with ideal model of porous material with the same porosity has also been plotted in this figure. It consists of a matrix of 20 by 5 BCC type spherical pores (see Figure 2.2(d)) with the same boundary conditions as porous channel filled with the real model. Quadratic curves are fitted through the data points using the least square method and their equations are also shown in Figure 4.5. It can be noticed that the simulation results of pressure drop per unit length can be represented by a quadratic dependence on free stream velocity, which is in agreement with the DarcyForchheimer equation. The quadratic behavior also shows that, for this particular geometry of carbon foam microstructure, the fluid flow makes the transition from Darcy to non-Darcy flow regime in the velocity range of the simulations. By comparing the curve equations in Figure 4.5 with the Darcy-Forchheimer equation (Equation (4.3)) the values of permeability and inertial coefficient for carbon foam can be calculated. The values of these two quantities are listed in Table 4.1 along with similar values reported in literature for comparison. 62 30000 Porous channel - real geometry Clear channel Pressure drop per unit length (Pa/m) 25000 Porous channel - ideal geometry 20000 y = 6160.7x2 + 7527.1x 15000 10000 y = 2749.4x2 + 4010.4x 5000 0 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 Free stream fluid velocity (m/s) Figure 4.5. Pressure drop per unit length as a function of free stream velocity. Table 4.1 Permeability and inertial coefficient of various porous materials K ( x 10 8 m 2 ) Source Porous material Type of analysis Calmidi & Aluminum Mahajan 90.05% Experimental 9.0 (2000) 20 PPI CFD on real 0.4462 geometry Carbon Present study 90.31% CFD on ideal 0.2377 geometry cf 0.088 0.15 0.245 63 It can be noticed from Table 4.1 that there are significant differences in permeability and inertial coefficient between carbon foam and aluminum foam. The permeability of aluminum foam is generally much higher than that of carbon foam for the same porosity because of the significant differences in pore structure. In aluminum foams the pore opening is of the same dimension as the pore (cell) size. On the other hand, the carbon foam pores tend to have the shape of bubbles and the pore openings are smaller than the bubble (cell) size. The aluminum foams tend to have larger pores than carbon foams for the same porosity. For instance, the 90.05% porous aluminum foam studied by Calmidi and Mahajan (2000) has an average pore diameter of 1.27 mm (20 pores per inch) in comparison with carbon foam which usually ranges between 0.1 and 0.5 mm. This makes the aluminum foam much more permeable to fluid than carbon foam at similar porosity. The inertial coefficient of carbon foam is higher than that of aluminum foam because of the differences in pore structure: the windows between pores are much smaller in carbon foams and the ligaments of carbon foam have much more complex geometry. It can also be noticed from Table 4.1 that the ideal geometry gives about half the permeability and a 63% higher inertial coefficient in comparison with the real geometry of carbon foam. It has already been shown that the thermal conductivity of the true foam is lower than that of the idealized geometry, which can be explained by higher tortuosity of the true foam. Therefore, the permeability of the true foam could be expected to be lower. However, the idealized geometry has much sharper edges than the real geometry, and this is reflected in the lower permeability and higher inertial coefficient. 64 Figure 4.6 shows the dependence between the friction coefficient calculated from Equation (4.5) and the modified Reynolds number, as discrete points. For the purpose of comparison, the experimental correlations from Paek et al. (2000) (Equation (4.7)) and Vafai and Tien (1982) (Equation (4.8)) are also plotted. It can be seen from Figure 4.6 that for Re K < 1 (when the fluid flow is in the Darcy regime) the friction coefficient calculated from simulation results follows very accurately as the inverse of the modified Reynolds number. Our results are also in very good agreement with experimental results reported by Paek et al. (2000) and Vafai and Tien (1982). It should be noticed that for Re K > 1 the friction coefficient starts to deviate from 1 Re K because as the higher velocity changes the flow to non-Darcy regime, the inertial effects can no longer be neglected. 65 100 Present study - real geometry Paek et al., 2000 Vafai & Tien, 1982 Friction coefficient 10 1 0.1 0.01 0.1 1 10 Modified Reynolds number Figure 4.6. Friction coefficient as a function of modified Reynolds number. 4.4.2 Convection heat transfer analysis In order to reduce the computational effort for the convection heat transfer analysis, the fluid flow and heat transfer problems are decoupled and solved independently. The fluid (air) and solid (carbon) thermophysical properties are assumed constant with temperature during both fluid flow and heat transfer simulations. The fluid flow analysis is first carried out with fluid properties calculated at 20°C and the fluid pressure and velocity distributions are obtained for inlet fluid velocities ranging between 0.1 m/s and 1.5 m/s. The heat transfer analysis is then carried out using the fluid velocity field obtained from the flow simulations and the temperature distributions in fluid and 66 solid are then obtained. The fluid temperature at the channel inlet is 20°C. The constant temperature heat source applied on the upper surface of the porous channel is at 50°C. Figure 4.7 shows the temperature distribution in the fluid in a clear channel (no carbon foam) and porous channel (solid phase thermal conductivity of 50 W/m°C) for an inlet fluid velocity of 0.5 m/s. It can be noticed that the fluid average temperature at the channel outlet is higher for the porous channel in comparison with the clear channel. This is due to the fact that a higher heat flux can be dissipated into the fluid when the porous material is present in the channel. The porous material acts like highly efficient fins, therefore the heat is conducted through the solid phase and then dispersed deep into the fluid. Overall, the effective heat transfer in the porous channel is increased by two means: higher velocity due to restrictions in flow area and extended surface area available for convection. At higher flow velocity, the effect of turbulence also enhances the heat transfer. 67 Flow direction Flow direction (a) (b) Figure 4.7. Fluid temperature distribution in (a) clear channel (no foam) and in (b) porous channel for a free stream velocity of 0.5 m/s. 68 The plots in Figure 4.8 are drawn to show the effective heat transfer coefficient calculated at the base area of the porous channel as a function of flow velocity and the intrinsic solid phase thermal conductivity of the foam. Three different values for thermal conductivity of the solid phase in carbon foam are used in simulations, which are 10, 50 and 100W/m°C. The results, as shown in Figure 4.8, demonstrate the influence on the effective heat transfer coefficient due to the presence of the foam. A simulation of convection heat transfer in clear channel (no foam) was also performed in order to provide a baseline that would show the enhancement in the effective heat transfer coefficient due to the presence of carbon foam in the channel. There is a significant increase in the effective heat transfer coefficient obtained in the porous channel in comparison with the clear channel (no foam). As expected, the effective heat transfer coefficient increases with the solid phase thermal conductivity. However, there is a saturation effect in heat transfer enhancement when the solid phase thermal conductivity rises from 50 W/m°C to 100 W/m°C. The saturation limit is due to the convection heat transfer limitation and is therefore dependent on the flow velocity. At higher flow rates the convection heat transfer increases and the effect of higher solid phase thermal conductivity shows up in the enhancement of the thermal transport. 69 1200.0 no foam Effective heat transfer coefficient (W/m2°C) 1000.0 k=10 W/m°C k=50 W/m°C 800.0 k=100 W/m°C 600.0 400.0 200.0 0.0 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 Free stream fluid velocity (m/s) Figure 4.8. Effective heat transfer coefficient as a function of free stream velocity and foam solid phase thermal conductivity. Table 4.2 lists the ratios between the effective heat transfer coefficients obtained with carbon foam and without carbon foam for various free stream velocity and solid phase thermal conductivity. It can be noticed that for high values of thermal conductivity of the solid phase and velocity, the heat transfer due to the presence of carbon foam in the channel is about 4.5 times higher than the value in the clear channel. This result compares well with experimental results reported by Ultramet (Pacoima, CA) on their website. They tested a cooling system made of a tungsten tube filled with tungsten foam which was cooled by helium and noticed that “the heat flux performance of the tungsten 70 foam/tungsten tube structure was approximately five times better than that of a tungsten tube with no foam” (“Refractory open-cell,” n.d.). Table 4.2 Increase in effective heat transfer coefficient for porous channel relative to clear channel Solid phase thermal conductivity (W/m°C) 10 50 100 0.1 2.81 2.92 2.96 Free stream 0.5 3.65 4.79 5.01 fluid velocity 1 3.03 4.51 4.65 (m/s) 1.5 2.96 4.46 4.65 As it can be noticed from the fluid flow and heat transfer simulation results presented above, the presence of carbon foam in the channel increases the heat transfer, but also causes a much higher pressure drop. When designing a heat sink, the energy required to flow the coolant needs to be compared against the thermal performance in order to ensure a certain level of efficiency. The energy required to flow the coolant through a heat sink can be quantified by pumping power per unit volume: Δp W& = uD Δx (4.16) The thermal performance of a heat sink can be quantified by its thermal resistance Rθ = ΔTLM Q (4.17) and a lower thermal resistance is desirable for a cooling system as it indicates good heat transfer capability. 71 Figure 4.9 shows the dependence between the thermal resistance and pumping power per unit volume of fluid for clear channel and porous channel with various solid phase thermal conductivities. It can be noticed that for a pumping power in the range of 40 to 600 W/m3 the porous channel gives a lower thermal resistance, which means better heat transfer performance. As expected, a higher thermal conductivity of the solid phase of carbon foam reduces the thermal resistance of the system. For instance, in order to obtain a thermal resistance of 2000 °C /W, the pumping power needed for the clear channel is about 8 times higher than that needed for the porous channel with a solid phase thermal conductivity of 100 W/m°C. It can be concluded that by adding carbon foam into channel, the heat transfer performance of the system can be increased in an efficient manner. 72 8000 7000 k=10 W/m°C k=50 W/m°C Thermal resistance (°C/W) 6000 k=100 W/m°C no foam 5000 4000 3000 2000 1000 0 1 10 100 1000 10000 3 Pumping power per unit volume (W/m ) Figure 4.9. Pumping power per unit volume of fluid as a function of thermal resistance of the channel. 100000 73 CHAPTER 5: CARBON FOAM TOOLING FOR ADVANCED COMPOSITE MANUFACTURING 5.1 Introduction A typical advanced composite material for aerospace application is made of carbon fiber reinforcement embedded in epoxy matrix. For example, clothes of woven carbon fiber impregnated with partially cured resin (prepreg) are commercially available for making composite parts. The composite part can be manufactured by manually laying up the prepreg on a tooling material and curing it under a pressure-temperature cycle in an autoclave. Alternatively, the automatic lay up of the carbon fiber yarn can be used when large composite parts are to be made. A critical problem for manufacture of composite parts is the geometrical and dimensional stability associated with process induced residual stresses. The severity of residual stresses and geometrical deformation is strongly influenced by the mismatch of thermal expansion of the composite part and the tool. A recent development in composites processing is the tendency towards manufacture of increasingly larger components, including fuselage sections, as a onepiece composite part. This requires larger tools and the handling of the massive tool can become an issue. The heating and cooling of the large thermal mass of a dense tool can also increase production costs. Consequently, lightweight materials would be preferred in composite tooling. Metals have been traditionally used as tooling materials for composite processing; in particular, steel and aluminum are quite popular. Steel is a reasonably cheap material, 74 with very good durability. Unfortunately, it has a coefficient of thermal expansion (CTE) higher than composite materials and the high density results in a high thermal mass. Aluminum is lighter than steel but has higher thermal conductivity and a higher CTE. Therefore, it exhibits a good heat up rate but the high CTE is detrimental to the dimensional stability (see Figure 5.1). In general, materials with high CTE are not used as tooling material for composite structures when geometrical and dimensional stability is a critical issue. To overcome this drawback, nickel base alloys, such as Invar, have been introduced as tooling for advanced composite materials. For example, Invar 36 closely matches the CTE of carbon fiber reinforced composites (see Figure 5.1). However, it is very expensive and dense. Due to its higher cost, Invar 36 is used primarily in special applications of advanced composites to aerospace structures such as wing skin (Campbell, 2004). Because of good CTE, carbon fiber reinforced epoxy laminates have also been used as composite tooling. A master tooling is first prepared, and then composite laminates are laid on the master tool and cured in an autoclave. By laying up many laminae at different orientation angles, the composite laminate tooling can be tailored to match closely the CTE of carbon fiber reinforced composite parts. Such composite toolings have the advantage of being light, but their use is limited by the fact that the matrix can crack after repeated thermal cycles in autoclave (Campbell, 2004). A tooling made of a layer of syntactic epoxy resin molded over an aluminum honeycomb core has been described by Cloud and Norton (2001) as a low cost tooling alternative. A similar tooling system, based on aluminum honeycomb core and Ren 75 patties is described by Burke (2003). Both of these types of tooling system have the disadvantage of having a CTE almost as high as aluminum (Figure 5.1). 25 20 CTE (10-6/°C) 15 10 5 0 -5 Carbon Fiber Carbon Fiber (longitudinal) (transverse) Invar 36 Carbon Foam Monolithic Graphite Steel Ren Patties Aluminum Figure 5.1. CTE of different tooling materials and carbon fiber (Burke, 2003; Burden, 1989; GrafoamTM Carbon Foam Solutions). Carbon as a tooling material is attractive because it can match the CTE of carbon based composites. Burden (1989) has discussed the advantages of monolithic graphite in composite tooling due to its CTE match with carbon fiber composites, high thermal conductivity, low thermal mass and a good dimensional stability at high temperatures. However, monolithic graphite is dense enough to create weight problems for large tooling. 76 Non-graphitic carbon foam is a good candidate for a tooling material by virtue of its wide range of properties. Unlike most traditional tooling materials, it is quite easy to tailor its thermal and mechanical properties to specific application requirements (Rowe et al., 2005). By decreasing the porosity of the foam the elastic modulus and mechanical strength will increase. On the other hand, by increasing the porosity, the bulk density decreases, which in turn reduces thermal mass and allows for better heat up rate. A major consideration in the application of carbon foam to tooling is its CTE which can match the CTE of carbon fiber reinforced composites. Rowe et al. (2005) provides a discussion of the advantages of carbon foam tooling. Experimental work on carbon foam tooling has been carried out by some carbon foam manufacturers and efforts are underway to commercialize the tooling application. A carbon foam tooling example is shown in Figure 5.2. This tool has two types of features (i) flat surfaces where curvature (warpage) of the composite part can occur due to stresses induced by mismatch between CTE of the tooling and the composite part, and (ii) curved surfaces (convex and concave) where spring-in of the composite part can result from anisotropic nature of the composite as well CTE mismatch with the tooling (Twigg et al., 2004a; Twigg et al., 2004b). This chapter focuses on the comparative analysis of the carbon foam vs. traditional tooling on dimensional stability of the composite; caused by the interaction of tooling and composite part, and strongly influenced by the thermomechanical properties of the tooling (Twigg et al., 2004a; Twigg et al., 2004b). 77 Figure 5.2. Carbon foam tooling for a simple part (Source: GrafTech International Ltd.). Even though the CTE of carbon foam is well matched to aerospace composites, and the thermal and mechanical properties of carbon foam can span a wide range, it is important to note that these properties are dramatically different from traditional tooling materials. In general, the density, modulus, and strength of typical carbon foam are orders of magnitude lower than steel. For example, carbon foam can be made in a wide range of bulk densities, such as 0.03 g/cc to 0.56 g/cc (“GrafoamTM Carbon Foam Solutions”, n.d.), which is much lighter than steel (~7.8 g/cc). The tensile strength of carbon foam can vary from 0.37 MPa to 6.3 MPa (“GrafoamTM Carbon Foam Solutions”, n.d.), which is much lower than steel (~710 MPa). The typical modulus of carbon foam is two orders of magnitude lower than steel. The bulk thermal conductivity of carbon foam can range from 1 W/m°C to 250 W/m°C, and is strongly dependent on degree of graphitization (Druma, 2005). The effect of all these properties on the performance or feasibility of 78 carbon foam tooling is not well documented since there are very few published studies of carbon foam tooling. The lack of process simulation on carbon foam tooling is in contrast to several analytical studies of composite processing using traditional tooling materials that have been published. For the carbon foam tooling design, it is important that process analysis must be available to identify the foam properties that would ensure the best geometrical and dimensional stability of the composite part at the lowest cost. Such an analysis should include the effect of tooling material properties on process induced stress development and deformation in fiber reinforced composites during autoclave curing. In particular, the effect of low values of CTE, thermal conductivity, mechanical strength, modulus and density of the carbon foam tooling on the product dimensional changes and process feasibility is of great interest. A preliminary study by Anghelescu and Alam (2008) studied non-graphitic carbon foam tooling for a simple shape. This chapter extends these results to predict the deformation associated with the different features shown in the tooling in Figure 5.2. The stresses in the carbon foam tooling are examined to evaluate the limitations due to lower mechanical strength of the foam. The effect of the distinct carbon foam properties on the stresses and deformations of the composite part are determined and compared with process simulations that use Invar and steel tooling. The numerical simulation in this study is based on the studies of thermal and mechanical behavior of composite materials that have been developed over several decades (Springer & Tsai, 1967; Bogetti & Gillespie, 1992). Over the last several years, experimental and numerical work has been carried out in order to understand the causes 79 of process induced residual stresses and their effect on the final geometry and dimensions of the composite part. Mechanical behavior of these materials is complex, therefore constitutive models have been developed with different simplifying assumptions and degree of detail. An important aspect of the model is to predict the behavior of thermosetting resins that changes from a highly viscous liquid to an elastic solid. Johnston et al. (2001) proposed a “cure-hardening instantaneously linear elastic” model for the development of the elastic modulus of the resin. Kim and White (1996) developed a stress relaxation model for an epoxy resin (3501-6) during cure, assuming a thermorheologically simple behavior and considering both time and cure-dependent effects. White and Kim (1998) developed a combined model in which thermal history is predicted by finite difference method and the stresses and deformations of composite part are simulated by finite element method. This approach was extended to a 3D finite element model by Zhu et al. (2001) in which process induced residual stresses accumulated during matrix curing are determined, and the deformations that result when the cured composite part is removed from tooling are then calculated. A transversely isotropic linear viscoelastic constitutive equation is used to model the composite material. It is shown that the mismatch between coefficients of thermal expansion (CTE) of composite part and tooling is an important contributor to process induced residual stresses developed during autoclave curing (Zhu et al., 2001). In the present research work, the approach developed by Zhu et al. (2001) is followed and the focus is on the effect of carbon foam tooling properties on the cured composite part. The numerical model is developed and validated by comparing with prior 80 results on traditional tooling. Then the model is used to examine the residual stresses and deformations of different features of composite parts. The stresses and deformations for the different thermo-mechanical properties of carbon foam, steel, and invar tooling are then evaluated. The composite material system used in this study is an AS4/3501-6 prepreg which has been widely used for making aerospace structures. Typically, this prepreg has AS4 carbon fiber with volume fraction ν = 0.6 . AS4 is a continuous, high strength, PAN based carbon fiber (“HexTowTM AS4 Carbon Fiber”, n.d.) and 3501-6 is a typical aerospace grade epoxy resin. This prepreg is selected because it has been used in prior studies, so the current numerical model could be validated by comparison with prior results. In the curing process, the composite material is assumed to be homogeneous and transversely isotropic. The carbon foam tooling material GRAFOAMTM FPA-35 (“GrafoamTM Carbon Foam Solutions”, n.d.) is taken to be homogeneous and isotropic. The thermal, chemical and mechanical processes for the processing of the composite-tooling system in this study have been described by Anghelescu and Alam (2008), which follows the approach of Zhu et al. (2001). The models are summarized below. 5.2 Thermo-chemical model The temperature distribution in the composite part during autoclave curing can be found by solving transient conduction heat transfer equation with internal heat generation dc ∂ 2T ∂ 2T ∂T ∂ 2T ρc p = k x 2 + k y 2 + k z 2 + ρ (1 − ν ) H R dt ∂t ∂z ∂y ∂x (5.1) 81 where ρ and c p are the composite material density and specific heat, T is the composite material temperature, k x , k y and k z are the composite material thermal conductivities in x, y and z directions, v is the fiber volume fraction in the composite material, c is the degree of cure of epoxy resin and H R is the ultimate heat of reaction of epoxy resin. The degree of cure of a thermosetting polymer at time t is defined as the ratio of the heat of reaction released up to time t to ultimate heat of reaction c= H (t ) HR and lies between 0 for a completely uncured polymer and 1.0 for a fully cured polymer. The heat released by a thermosetting polymer during curing is described by the cure kinetics as a function of time, temperature and degree of cure: dc = f (T , c ) dt The rate of degree of cure for 3501-6 epoxy resin used in this research work is (Lee et al., 1982) dc = (K 1 + K 2 c )(1 − c )(0.47 − c ) dt dc = K 3 (1 − c ) dt c ≤ 0.3 (5.2) c > 0.3 where K i = Ai e − ΔEi RT i = 1, 2, 3 and R = 8.31 J/mol K is the gas constant and Ai and ΔEi are pre-exponential factors and activation energies, respectively: 82 A1 = 2.101 ⋅ 10 9 min −1 A2 = −2.014 ⋅ 10 9 min −1 A3 = 1.96 ⋅ 10 5 min −1 ΔE1 = 8.07 ⋅ 10 4 J / mol ΔE 2 = 7.78 ⋅ 10 4 J / mol ΔE3 = 5.66 ⋅ 10 4 J / mol The initial conditions at each point in the composite T = T0 and c = 0 at t = 0 and the convective boundary condition on the surface of the composite exposed to autoclave environment − ki ∂T = h(T − T∞ ) ∂xi i = 1, 2, 3 must be imposed when solving Equations (5.1) and (5.2) for temperature and degree of cure distributions, where T∞ is the autoclave environment temperature and h is convection heat transfer coefficient from autoclave environment to composite surface. The autoclave environment is assumed to follow the typical pressure-temperature cure cycle described by Kim and Hahn (1989) which is shown in Figure 5.3. The value of the heat transfer coefficient associated with the convective boundary condition is 140 W/m2°C (“Autoclave and tooling effects”, 2003). A perfect thermal contact between composite part and tooling material was assumed during autoclave curing. The temperature distribution in the tooling during autoclave curing can be found by solving Equation (5.1) with two simplifications: (a) there is no degree of cure and 83 internal heat generation in tooling and, (b) tooling material is assumed to be isotropic so k x = k y = k z . The boundary conditions are changed accordingly. 200 1200 180 1100 160 1000 Temperature (°C) 140 900 120 800 100 700 80 600 60 500 40 400 20 300 0 0 50 100 150 200 250 Pressure (KPa) Temperature (°C) Pressure (KPa) 200 300 Time (min) Figure 5.3. Cure cycle for AS4/3501-6 (Kim & Hahn, 1989). Thermal properties of the composite material are shown in Table 5.1. These are calculated using rule of mixture and models analogous to electrical circuits as introduced by Springer and Tsai (1967) (see Appendix A). Thermal properties of tooling materials are shown in Table 5.2. Thermal properties of carbon fiber, epoxy resin and tooling material have been assumed to be constant during the curing cycle. 84 Table 5.1 Thermal properties of AS4/3501-6 prepreg and its components (Lee et al., 1982; Loos & Springer, 1983) AS4 Carbon 3501-6 Epoxy AS4/3501-6 Property Fiber Resin Composite Density, 1790 1260 1578 ρ (kg/m3) Specific heat, 712 1260 887 c p (J/kg°C) Thermal conductivity, k (W/m°C) Heat of reaction, H R (J/kg) 26 0.167 15.7 (longitudinal) 0.687 (transverse) - 473600 - Table 5.2 Thermal properties of tooling materials (“GrafoamTM Carbon Foam Solutions”, n.d.; Wiersma et al., 1998) GRAFOAMTM FPA-35 Property Steel Invar 36 carbon foam Density, 7833 8055 560 ρ (kg/m3) Specific heat, 434 515 710 c p (J/kg°C) Thermal conductivity, k (W/m°C) Thermal diffusivity, α t (m2/s) 60.5 10.5 0.3 1.8·10-5 2.5·10-6 7.5·10-7 A 2D version of the thermo-chemical model was used to study the stresses and deformations. A 2D version is preferred because the causes and effects can be readily identified in the results. For the same reason, the composite part is taken to be made of a simple unidirectional laminate. The model was implemented in the finite element 85 computer software ABAQUS (Abaqus Inc., 2006). A user subroutine was written to calculate the internal heat generation and the degree of cure of composite material for the ABAQUS model. The thermo-chemical model yields the temperature and degree of cure distributions in composite as well as the temperature distribution in tooling as a function of time t and space coordinates x and y . These results are used as input data for the stress-displacement model. 5.3 Stress-displacement model The time-dependent mechanical behavior of epoxy resin is modeled using linear viscoelasticity while the mechanical behavior of carbon fiber is modeled using linear elasticity. The constitutive equation in hereditary integral form is therefore based on the model of a linear anisotropic viscoelastic material undergoing changes in temperature and degree of cure (White & Kim, 1998; Zhu et al., 2001): t σ ij (t ) = ∫ C ijkl (c, T , t − t ') −∞ d (ε kl (t ') − ε kl (t ')) dt ' dt ' i, j , k , l = 1, 2, 3 (5.3) where σ ij are the stress components, Cijkl are the composite material stiffness matrix components, ε kl are the total strain components, t is the current time and t ' is the past time. The material stiffness Cijkl has a strong variation with time, temperature and degree of cure. Assuming that the linear anisotropic viscoelastic material exhibits thermorheologically simple behavior at constant degree of cure (White & Kim, 1998; Zhu et al., 2001), reduced time ‘ ξ ’, can be introduced: 86 t ξ =∫ 0 dt" aT (c, T ) (5.4) t' dt" ξ '= ∫ a (c, T ) 0 T where the shift factor aT (c, T ) allows for time-temperature superposition. For the 3501-6 epoxy resin the shift factor is (Kim & White, 1996) 1 ⎞ ⎛ aT (c, T ) = ⎜⎜ − 1.4e c −1 − 0.0712 ⎟⎟ ⋅ (T − 30) ⎠ ⎝ Assuming that the linear anisotropic viscoelastic material has been free of strain before t = 0 , the constitutive equation ca be written in a simpler form (Zhu et al., 2001): t σ ij (t ) = ∫ C ijkl (ξ − ξ ') 0 d (ε kl (t ') − ε kl (t ')) dt ' dt ' i, j , k , l = 1, 2, 3 (5.5) The thermo-chemical strain is the sum of the effects of thermal and chemical expansion/shrinkage of linear viscoelastic material, and is expressed as: ε kl = α kl ΔT + β kl Δc where α kl and β kl are thermal and respectively, chemical expansion/shrinkage coefficients. The relaxation modulus of a thermoreologically simple viscoelastic material is modeled using an assembly of n Maxwell elements in parallel, which gives a Prony series with n terms (Kim & White, 1996) [ ] n E (c, ξ ) = E (c ) + E (c ) − E (c ) ∑ wi (c )e ∞ 0 ∞ −ξ (c ,T ) τ i (c ) (5.6) i =1 where E 0 and E ∞ are the unrelaxed and the relaxed elastic modulus of viscoelastic material, respectively; and wi and τ i are the weight factor and the relaxation time of the 87 i th Maxwell element. According to experimental results reported by Kim and White (1996), E 0 , E ∞ and wi do not depend significantly on degree of cure so the relaxation modulus can be written in the simpler form ( E (ξ ) = E + E − E ∞ 0 ∞ )∑ w e n i =1 − ξ ( c ,T ) τ i (c ) (5.7) i where the wi and τ i values for 3501-6 epoxy resin are given in Table 5.3. Following the approach of White and Kim (1998) it has been assumed that the relaxation behavior of the composite material has the same thermorheologically simple behavior as its polymer matrix material in pure state. Under this assumption, the relaxation behavior of the 0 composite material can be written in terms of its unrelaxed and relaxed stiffnesses Cijkl ∞ and Cijkl C ijkl (ξ ) = C ∞ ijkl ( + C 0 ijkl −C ∞ ijkl )∑ w e n i =1 −ξ ( c ,T ) τ i (c ) i (5.8) The unrelaxed and relaxed composite material stiffnesses are calculated using the values of mechanical properties shown in Table 5.4 and the micromechanics model introduced by Bogetti and Gillespie (1992) (see Appendix A). The mechanical properties of carbon fiber are assumed to be constant during the autoclave temperature-pressure cure cycle. Poisson’s ratio and thermal and chemical expansion/shrinkage coefficients of 3501-6 epoxy resin are also assumed constant during cure (Kim & White, 1996). 88 Table 5.3 Weight factors and relaxation times for 3501-6 epoxy resin (Kim & White, 1996) τi (min) wi i 1 1 2.92·10 0.059 3 2 2.92·10 0.066 3 1.82·105 0.083 7 4 1.10·10 0.112 8 5 2.83·10 0.154 6 7.94·109 0.262 11 7 1.95·10 0.184 8 3.32·1012 0.049 14 9 4.92·10 0.025 Table 5.4 Mechanical properties of AS4/3501-6 prepreg components (Bogetti & Gillespie, 1992; Kim & White, 1996; White & Kim, 1998) AS4 3501-6 Epoxy Resin 3501-6 Epoxy Resin Property Carbon Unrelaxed Relaxed Fiber Longitudinal elastic 206.8 modulus, E1 (GPa) 3.2 0.032 Transverse elastic 20.68 modulus, E 2 = E3 (GPa) In-plane Poisson’s ratio, 0.2 ν 12 = ν 13 0.35 Transverse Poisson’s 0.3 ratio, ν 23 In-plane shear modulus, 27.58 G12 = G13 (GPa) 1.185 0.01185 Transverse shear modulus, 6.894 G23 (GPa) Longitudinal CTE, -9·10-7 α 1 (1/°C) 5.76·10-5 Transverse CTE, 7.2·10-6 α 2 = α 3 (1/°C) Chemical shrinkage 0 -0.01695 coefficient, β1 = β 2 = β 3 89 The mechanical behavior of tooling material can be described using linear elasticity. The constitutive equation of a linear isotropic elastic material undergoing changes in temperature is: σ ij = Cijkl (ε kl − αΔT ) i, j , k , l = 1, 2, 3 (5.9) where α is thermal expansion coefficient of tooling material. The tooling material stiffnesses are calculated using mechanical properties in Table 5.5 and relations in Appendix A. Mechanical properties of tooling material are assumed to be constant during the autoclave temperature-pressure curing cycle. Table 5.5 Mechanical properties of various tooling materials (“GrafoamTM Carbon Foam Solutions”, n.d.; Wiersma et al., 1998) GRAFOAMTM FPA-35 Property Steel Invar 36 carbon foam Elastic modulus, 210 145 3.5 E (GPa) Poisson’s ratio 0.3 0.3 0.3 ν Coefficient of thermal expansion 12·10-6 1.6·10-6 2.3·10-6 α (1/°C) 60 (compressive) Strength 710 450 6.3 (tensile) (MPa) A 2D generalized plane strain version of the stress-displacement model was implemented in the general purpose finite element computer code ABAQUS. As the z dimension of the composite part is much longer than the other two dimensions and there are no mechanical constrains in this direction a generalized plane strain behavior can be assumed. A user subroutine was written to implement the constitutive mechanical 90 behavior of transversely isotropic linear viscoelastic material. The stress-displacement model yields the stress distribution in the composite part throughout the curing process and the deformation of the composite part after removing from the tooling. This model was first validated by simulating several test cases from prior studies (Zhu et al., 2001; Wiersma et al., 1998). The simulations for the carbon foam tooling were then carried out with the model, as described below. 5.4 Simulations and results The feasibility and carbon foam tooling-composite part interaction properties during composite processing is investigated by analyzing the deformation and residual stress generated in three cases of different geometric features. The composite part is a 2 mm thick [016]T laminate of the following geometry: (i) laminate bent at 90° on a convex tooling, (ii) laminate bent at 90° on a concave tooling and, (iii) flat laminate on a flat tooling. For the first two cases of 90° bend, the inner radius is 4 mm and the length of straight section is 15 mm (Figure 5.4) and carbon fibers are oriented along the tooling surface parallel to the length direction. The flat composite part has a length of 30 mm, with fibers oriented along the tooling surface parallel to the length direction. The tooling is 8 mm thick in all cases. Because of symmetry, just half of the geometry is analyzed by applying appropriate boundary conditions on the symmetry line. A schematic representation of geometries, finite element meshes, spring-in and warpage for all three cases (i) - (iii) are shown in Figures 5.4 and 5.5. 91 Symmetry line Convex tooling Convex tooling Composite part Composite part Symmetry line Composite part Composite part Concave tooling Concave tooling (a) (b) Figure 5.4. (a) 3D solid model and (b) 2D finite element discretization of the composite part on convex and concave tooling. The spring-in angle was calculated from the displacements of the composite part side facing the tooling, which means the angle Δθ1 for convex tooling and Δθ 2 for concave tooling. These angles are defined by the lines through the corners of the straight section of the composite part, as shown in Figure 5.5(a), so they account for the spring-in of the curved section along with the warpage of the flat section. The warpage of the flat composite part was calculated by taking the displacement δ at the bottom of the composite. After removing from the tooling, the flat composite part is assumed to be an arc of a circle, where the curvature is calculated as Kim and Hahn (1989) 92 κ= 1 2δ = R ⎛ L ⎞2 2 ⎜ ⎟ +δ 2 ⎝ ⎠ (5.10) where R is the radius of the curved composite part and δ and L are the dimensions given in Figure 5.5(b). Symmetry line Δθ1 Δθ2 Symmetry line (a) δ L 2 (b) Figure 5.5. The deflections and angles due to deformations from processing: (a) spring-in for a 90 degree angle and (b) warpage of a flat composite part. The values are determined after removing the part from the tooling. Two limiting cases of tooling-composite interactions were investigated: (a) in the first case the composite part is assumed perfectly bonded to the tooling throughout the cure cycle, and (b) the composite and the tooling are assumed to have frictionless sliding contact with no separation. For the first case, the composite part is separated from the tooling at the end of the cure cycle to determine the residual stresses and the deformation 93 of the free part. In the model this is simulated by deactivating the mechanical interaction between the two. The spring-in angles and curvatures predicted by the model for both convex and concave cases are listed in Tables 5.6 and 5.7, respectively. As expected, the tooling material does not affect the final spring-in of the composite when there is no bonding between the tool and the composite. However, even for this “ideal” case, there is stress accumulation in the composite part that produces a spring-in for the case of curved geometry due to anisotropic nature of the composite material. The remainder of the results discussed below is based on perfect bonding between the tooling and the composite, so that the effect of tooling material properties can be evaluated. At the end of the cure cycle, after removing from the tooling, the curved section of the composite part will spring-in and the flat section will become curved (warpage) as shown in Figure 5.5(a) and (b). Table 5.6 Bent composite part spring-in angle Tooling material/geometry Carbon Foam Convex 0.27 Bonding Concave 0.24 Convex 0.27 No Bonding Concave 0.27 Spring-in angle (deg.) Invar 36 0.26 0.2 0.27 0.27 Steel 0.37 0.05 0.26 0.27 94 Table 5.7 Flat composite part curvature Tooling material/geometry Carbon Foam Bonding 2.4 ⋅ 10−5 No Bonding 0 Curvature ( mm −1 ) Invar 36 5.6 ⋅ 10−5 0 Steel 29.2 ⋅ 10−5 0 5.4.1 Results for curved tooling geometry (convex and concave) As it can be seen from Figure 5.6(a), a good CTE match between composite part and tooling, which is the case for carbon foam, will generate almost no stress accumulation in the straight section of the composite laminate; while the stress distribution in the curved section of the composite laminate is mostly due to the composite material anisotropy. It can be noticed from Figure 5.6(b) that, curing on steel tooling generates compressive stresses up to 167 MPa in the mid-section of the composite part. There are also tensile stresses up to 9 MPa generated in the composite part at the interface with tooling. This high stress variation across the thickness increases the risk of composite part delamination. After removing from the tooling, part of this complex state of stress will relax generating a permanent deformation, namely the curved section of the composite part will spring-in and the straight section will warp as shown in Figure 5.7(b). The total spring-in angle, as given in Table 5.6 represents the superposition of these two effects. The two effects add-up, as both of them tend to deform the composite part towards the tooling. The angle is smaller for carbon foam because there is almost no warpage when curing the composite part on carbon foam tooling as shown in Figure 95 5.7(a). Simulations have also been carried out which show that an increased warpage will be obtained when the composite part is thinner or the straight section is longer. 1 1 2 2 (a) (b) Figure 5.6. Stress distribution along fiber direction at the end of cure cycle before removing the composite part from convex tooling made of (a) carbon foam and (b) steel. 96 1 1 2 (a) 2 (b) Figure 5.7. Stress distribution along fiber direction at the end of cure cycle after removing the composite part from convex tooling made of (a) carbon foam and (b) steel. A deformation scale factor of 10 is used for displacement. The stress accumulation process in the case of concave tooling is similar to the case presented for convex tooling. The stress distribution is shown in Figure 5.8(a) and (b) for curing the composite part on carbon foam tooling and steel tooling, respectively. The only relevant difference from the convex tooling case is the following: during the stress relaxation process after removing the composite part from tooling, the curved section spring-in will deform the composite part away from the tooling and the straight section will warp towards the tooling. The two effects tend to deform the composite part in opposite directions. For this reason, a higher CTE tooling material (steel) generates a smaller spring-in angle after removing from tooling (see Table 5.6), but also a compressive residual stress up to 42.3 MPa, which is 5.5 times higher than when using carbon foam (see Figure 5.9(a) and (b)). 97 1 1 2 2 (a) (b) Figure 5.8. Stress distribution along fiber direction at the end of cure cycle before removing the composite part from concave tooling made of (a) carbon foam and (b) steel. 1 1 2 (a) 2 (b) Figure 5.9. Stress distribution along fiber direction at the end of cure cycle after removing the composite part from concave tooling made of (a) carbon foam and (b) steel. A deformation scale factor of 10 is used for displacement. 98 The history of stresses in the carbon foam tooling was also studied for these cases of perfect bonding, and the stresses were observed to be much lower (<2.5 MPa) as seen in Figure 5.6(a) because of the lower CTE and lower elastic modulus values of carbon foam. Since the stresses in the carbon foam tooling are low, the tooling stresses have been omitted in the rest of the figures. This helps to clearly delineate the partition line between the tooling and the part. The following discussion focuses first on the stresses and deformation in the flat composite part; the issues of the lower modulus and tensile strength of the carbon foam tooling will be discussed later. 5.4.2 Results for flat tooling geometry When curing flat composite laminates, the tooling material has a very strong influence on the amount of stress accumulated at the end of the autoclave curing cycle as it can be seen in Figure 5.10(a) and (b). The stress patterns are similar but the maximum compressive stress in case of steel tooling (147 MPa) is much higher than in case of carbon foam tooling (4.6 MPa). This results in a much higher deformation after removing the composite part from the steel tooling as shown in Figure 5.11(a) and (b). The results obtained for curvature are presented in Table 5.7. It can be noticed that both Invar 36 and carbon foam tooling give a smaller curvature of composite part in comparison with steel tooling because their CTEs are much smaller than steel. 99 1 2 (a) (b) Figure 5.10. Stress distribution along fiber direction at the end of cure cycle before removing the composite part from flat tooling made of (a) carbon foam and (b) steel. 1 2 (a) (b) Figure 5.11. Stress distribution along fiber direction at the end of cure cycle after removing the composite part from flat tooling made of (a) carbon foam and (b) steel. A deformation scale factor of 10 is used for displacement. 100 5.4.3 Effects of thermo-mechanical properties of tooling The history of the curing process, as shown by the degree of cure at points (A) and (B) for carbon foam tooling in Figure 5.12, demonstrates that the lower thermal conductivity of the foam does not affect the composite part significantly. This is due to the fact that the thermal transport can occur through the composite part, which has conductive carbon fiber reinforcement. It has been shown that thermal transport issues can become significant for thicker laminates (Bogetti & Gillespie, 1992), even when the tooling material is a conductive metal. The stress variation along the fiber direction throughout the cure cycle for the three different tooling is also quite similar at the specific point (A) with the three different tooling materials until the cool down process begins, when stresses start to develop due to CTE mismatch and material anisotropy. 101 20 1.00 0 0.89 -20 0.78 -40 0.67 -60 0.56 Degree of cure at points (A) and (B) -80 0.44 Degree of cure Stress (MPa) Carbon Foam Invar -100 0.33 -120 0.22 B -140 A 0.11 Steel -160 0 50 100 150 200 250 300 0.00 350 Time (min) Figure 5.12. Curing process of the composite part on convex tooling made of different materials. The stress distribution along fiber at point (A) is shown for different tooling materials. The curves for the degree of cure at points (A) and (B) overlap, even though the point (B) is insulated by the carbon foam tooling. It can be seen in Figure 5.12 that only a small part of the total process induced stress is accumulated in the composite part before the final cool down stage because of viscoelastic stress relaxation. When the composite part is fully cured and the system cools down, it will cause an increase in compressive stress along fiber. At this point the CTE mismatch between the fully cured composite part and tooling becomes important. As the system is cooling down, the tooling shrinks due its positive CTE, while the composite part tends to expand since it has negative CTE. This generates a high compressive stress in composite part at the end of cure cycle before the composite part is removed from 102 tooling, as shown in Figure 5.12. The stress generated for the case of steel tooling is much higher in comparison with the case of carbon foam or Invar 36 tooling because the CTE mismatch is higher with steel. The stress distribution in carbon foam tooling for all three tooling configurations was also analyzed. The maximum value, as discussed earlier, is about 2.5 MPa and it was obtained for convex configuration. This value is low because of (i) better CTE match with the composite and, (ii) lower modulus of the carbon foam. In other words, the low CTE produces a better match with the CTE of the composite part resulting in smaller strains in the carbon foam tooling, and this, in combination with the lower modulus of the foam gives very low stress in the foam tooling. This is an important issue because the stresses of the same order as in the composite part (~90 MPa) can not be sustained in the foam tooling. The effect of the lower modulus of carbon foam tooling is evident in Table 5.6, where the curvature produced by carbon foam tooling is less than what is produced by Invar tooling, even though the carbon foam has a higher CTE than Invar 36. This is due to the lower bulk modulus of carbon foam, which reduces the stress and the strain on the composite part. While the lower modulus is seen to be helpful in reducing residual stresses, it can be an issue in terms of deformation of the tooling-composite part system. For example, the composite part in Figure 5.6 exerts a moment on the tooling; resulting in a curvature that is a “spring-in” of the total tooling-composite part system. This curvature was determined to be 0.017 degree, which is more than an order of magnitude less than the final spring-in of 0.27 degree of the composite part. Therefore it can be 103 expected that the deformation of the tooling will not have a significant effect as long as the tooling is properly supported. In practice, the carbon foam tooling may have an adhesive seal and/or composite face sheet on the surface which can also reduce deformation of the tooling (Roy, 2008). The tensile strength of this carbon foam is about 6.3 MPa (“GrafoamTM Carbon Foam Solutions”, n.d.). This is higher than the highest stress (~2.5 MPa) in the simulation; but they are of similar order of magnitude. Therefore, even though the mechanical integrity of the tooling in this particular case should not be affected during autoclave curing of the composite, the strength of the foam is a possible limitation for designing complex, large tooling with carbon foams of low strength. It should be noted that the compressive strength of the same foam is 60 MPa (“GrafoamTM Carbon Foam Solutions”, n.d.), which would provide adequate strength in compression mode for most tooling applications. 5.5 Summary of analyses of carbon foam tooling A detailed analysis has been carried out for a small tooling-composite part system to compare the residual stresses and deformations when processing composite laminates using carbon foam versus traditional tooling materials. The results have shown that the residual stresses can be reduced significantly by using carbon foam tooling. The stresses are reduced due to lower CTE and lower modulus of the carbon foam. Because of lower residual stresses, carbon foam tooling material generated smaller deformation of composite parts. An exception was seen when large spring-in with steel tooling was 104 compensated by high warpage in a concave system. As regards the spring-in angles of curved composites, the values obtained for carbon foam tooling are similar to those obtained for Invar tooling. These results are obtained for the simple geometries and small parts that were analyzed. Additional studies need to be done to determine the thermomechanical effects in curing thicker and/or larger parts with complex geometry. Care must be taken to evaluate the mechanical integrity of the tooling and in selection of a carbon foam with adequate strength. Experimental research also has to be carried out to validate these results and develop a reliable carbon foam tooling design for composite processing. Even though the dimensional stability of the composite part is a major factor in the selection of the tooling material, it is just one of the many requirements for composite processing. Additional requirements include a long-term resistance to high temperature and heat-up/cool-down cycles in autoclave, ability to maintain vacuum integrity and suitable heat-up rate (Campbell, 2004; Rowe et al., 2005). All these factors play a key role in production of geometrical and dimensionally accurate composite structures; therefore, these issues must also be considered in making a tooling material selection. 105 CHAPTER 6: CONCLUSIONS AND RECOMMENDATIONS FOR FUTURE WORK The objective of this research work was two fold: (i) to use an accurate three dimensional solid model of carbon foam microstructure in developing finite element models to calculate bulk thermal and mechanical properties; and (ii) to study the application of carbon foam in convection heat transfer and composite tooling. The results from the model compare favorably with experimental and analytical results available in literature. One of the limiting factors of the modeling effort is the computational effort required to set up and run a model that reflects accurate details at the pore level of the microstructure. In order to have a model that is representative of the bulk material, the model should to be large enough to have the same properties (density, thermal and mechanical properties, etc.) as the bulk material. Since the thermal and mechanical properties of the bulk material tend to be non-homogeneous, it is quite difficult (if not impossible) to select a reasonably small volume of the foam that can reflect all of the bulk properties. To keep the computational effort reasonable, the small volume of the model in this study was selected only on the basis of similarity to the bulk density. The validity of the model is checked by comparing with experimental values from the literature. A high performance CAD workstation was used for solid modeling of carbon foam. The number of mesh elements in the model is typically 1 million or more, especially when the fluid flow was included. Because of the large number of mesh 106 elements, a 64-bit Linux machine was used to run the FEA and CFD simulations With the continuous improvement in the capabilities of computational systems, it is expected that larger volume models would be handled more easily in future work and this would result in greater confidence in the results of the numerical model. The results of the solid model and the fluid flow simulations compared well with the experimental values. This is particularly significant considering that the idealized geometry models used in prior research produced results that are typically very different from experimental values. For example, the thermal conductivity of carbon foam is higher in the idealized model by a factor of 2 to 3 when compared to experimental values. The elastic modulus has a higher discrepancy, a factor of about 6. But the model developed in this current study predicts values that are in the range of experimental values. Similar differences in the flow model of idealized vs. true three dimensional carbon foam geometry have significant implications for the design of heat exchangers and heat sinks using carbon foam. The solid model developed in Chapter 2 was then used in Chapter 4 to develop a fluid flow and convection heat transfer model to study the behavior of thermal transport of carbon foams in presence of a fluid flow. For future work, the method and the results of this research should be used to model a heat sink or heat exchanger core which is typically much larger than the solid model in this study. In general, a heat sink or heat exchanger core is a periodic structure that can be modeled by identifying a repeating cell and applying periodic boundary conditions. In this approach, the solid model of carbon 107 foam used in this study would be mirrored in different directions, so that it represents a periodic cell and periodic boundary conditions would be applied on each of its faces. For a solid model which is at least several millimeters in dimension, it will be more accurate to include a bulk variation in conductivity that is often the case in graphitic foams. The complex task of incorporating the anisotropic and inhomogeneous material properties of the ligaments and nodes can be modeled when greater computational capability becomes available. 108 REFERENCES Abaqus Inc. (2006). ABAQUS 6.6 Theory Manual. Alam, M. K., & Maruyama, B. (2004). Thermal conductivity of graphitic carbon foams. Experimental Heat Transfer, 17(3), 227-241. Altair Engineering Inc. (2007). Hypermesh 8.0. Anghelescu, M. S., & Alam, M. K. (2008). Carbon foam tooling for aerospace composites manufacturing. SAMPE Journal, 44(1), 6-13. Anghelescu, M. S., & Alam, M. K. (2006). Finite element modeling of forced convection heat transfer in carbon foams. Proceedings of 2006 ASME International Mechanical Engineering Congress and Exposition, Chicago, Illinois. Antohe, B. V., Lage, J. L., Price, D. C., & Weber, R. M. (1997). Experimental determination of permeability and inertia coefficients of mechanically compressed aluminum porous matrices. Journal of Fluids Engineering, Transactions of the ASME, 119(2), 404-412. Autoclave and Tooling Effects. (2003). Retrieved April 25, 2008 from http://www.convergent.ca/Objects/Case%20Studies/Autoclave%20and%20To oling%20Effects/2004-02-12%20Autoclave_and_Tooling_Effects.pdf Bauer, T. H. (1993). A general analytical approach toward the thermal conductivity of porous media. International Journal of Heat and Mass Transfer, 36(17), 4181-4191. 109 Beavers, G. S., & Sparrow, M. (1969). Non-Darcy flow through fibrous porous media. Transactions of the ASME. Series E, Journal of Applied Mechanics, 36(4), 711-714. Bhattacharya, A., & Mahajan, L. (2002). Finned metal foam heat sinks for electronics cooling in forced convection. Transactions of the ASME. Journal of Electronic Packaging, 124(3), 155-163. Bogetti, T. A., & Gillespie, J. W. (1992). Process-Induced Stress and Deformation in Thick-Section Thermoset Composite Laminates. Journal of Composite Materials, 26(5), 626-660. Boomsma, K., & Poulikakos, D. (2001). On the effective thermal conductivity of a three-dimensionally structured fluid-saturated metal foam. International Journal of Heat and Mass Transfer, 44(4), 827-836. Boomsma, K., & Poulikakos, D. (2002). The effect of compression and pore size variations on the liquid flow characteristics in metal foams. Journal of Fluids Engineering (Transactions of the ASME), 124(1), 263-272. Boomsma, K., Poulikakos, D., & Zwick, F. (2003). Metal foams as compact high performance heat exchangers. Mechanics of Materials, 35(12), 1161-1176. Brow, M., Watts, R., Alam, M.K., Koch, R., & Lafdi, K. (2003). Characterization requirements for aerospace thermal management applications. Proceedings of SAMPE 2003, Dayton, Ohio. Burden, J. H. (1989). Overview of monolithic graphite tooling. 34th International SAMPE Symposium and Exhibition, Reno, NV, May 8-11. 110 Burke, S. (2003). Low cost tooling for composites: a real world analysis of tooling. SAMPE Journal, 39(1), 46-50. Calmidi, V. V. (1998). Transport phenomena in high porosity metal foams (Doctoral Dissertation, University of Colorado, Boulder, CO). Calmidi, V. V., & Mahajan, R. L. (1999). The effective thermal conductivity of high porosity fibrous metal foams. Transactions of the ASME. Journal of Heat Transfer, 121(2), 466-471. Calmidi, V. V., & Mahajan, R. L. (2000). Forced convection in high porosity metal foams. Journal of Heat Transfer, Transactions ASME, 122(3), 557-565. Campbell, F. C. (2004). Manufacturing processes for advanced composites. UK: Elsevier Advanced Technology. Chen, C., Kennel, E., Stiller, A., Stansberry, P., & Zondlo, J. (2006). Carbon foam derived from various precursors. Carbon, 44(8), 1535-1543. Cloud, D., & Norton, J. (2001). Low-cost tooling for composite parts: the LCTC process. Assembly Automation, 21(4), 310-316. Druma, A. M. Alam, M. K., Anghelescu, M. S., Druma, C., (2005). Three dimensional modeling of carbon foams. Proceedings of 2005 ASME International Mechanical Engineering Congress and Exposition, Orlando, Florida. Druma, A. (2005). Analysis of carbon foams by finite element method. (Doctoral Dissertation, Ohio University, Athens, OH). 111 Druma, C., Alam, M. K. & Druma, A. M. (2004). Finite Element Model of Thermal Transport in Carbon Foams. Journal of Sandwich Structures and Materials, 6(6), 527-540. Fluent Inc. (2006). FLUENT 6.3 User’s Guide. Gallego, N. C., & Klett, J. W. (2003). Carbon foams for thermal management. Carbon, 41(7), 1461-1466. Gibson, L. J., & Ashby, M. F. (1997). Cellular solids: Structure and properties (2nd ed.). UK: Cambridge University Press. GrafoamTM Carbon Foam Solutions. (n.d.). Retrieved April 28, 2008 from http://www.graftech.com/getdoc/c8d02e1c-504a-4104-8fac8184c4704696/CompositeTooling_2008_FIN.aspx HexTowTM AS4 Carbon Fiber (n.d.). Retrieved November 17, 2008 from http://www.hexcel.com/NR/rdonlyres/5659C134-6C31-463F-B86B4B62DA0930EB/0/HexTow_AS4.pdf Hinze, J. O. (1975). Turbulence (2nd ed.). NY: McGraw-Hill Publishing Co. Hunt, M. L., & Tien, C. L. (1988). Effects of thermal dispersion on forced convection in fibrous media. International Journal of Heat and Mass Transfer, 31(2), 301-309. Johnston, A., Vaziri, R., & Poursartip, A. (2001). A plane strain model for process-induced deformation of laminated composite structures. Journal of Composite Materials, 35(16), 1435-1469. 112 Karimian, S. A. M., & Straatman, A. G. (2008). CFD study of the hydraulic and thermal behavior of spherical-void-phase porous materials. International Journal of Heat and Fluid Flow, 29(1), 292-305. Kim, K. S., & Hahn, T. (1989). Residual stress development during processing of graphite/epoxy composites. Composites Science and Technology, 36(2), 121132. Kim, Y. K., & White, S. R. (1996). Stress relaxation behavior of 3501-6 epoxy resin during cure. Polymer Engineering, 36(23), 2852-2862. Klett, J., McMillan, A., Gallego, N., & Walls, C. (2004). The role of structure on the thermal properties of graphitic foams. Journal of Materials Science, 39(11), 3659-3676. Krishnan, S., Murthy, J. Y., & Garimella, S. V. (2006). Direct Simulation of Transport in Open-Cell Metal Foam. Journal of Heat Transfer, 128(8), 793799. Lee, W. I., Loos, A. C., & Springer, G. S. (1982). Heat of Reaction, Degree of Cure, and Viscosity of Hercules 3501-6 Resin. Journal of Composite Materials, 16(6), 510-520. Li, K., Gao, X. - L., & Roy, A. K. (2005). Micromechanical modeling of threedimensional open-cell foams using the matrix method for spatial frames. Composites Part B: Engineering, 36(3), 249-262. Loos, A. C., & Springer, G. S. (1983). Curing of Epoxy Matrix Composites. Journal of Composite Materials, 17(2), 135-169. 113 Maruyama, B., Spowart, J. E., Hooper, D. J., Mullens, H. M., Druma, A. M., Druma, C., & Alam, M. K. (2006). A new technique for obtaining threedimensional structures in pitch-based carbon foams. Scripta Materialia, 54(9), 1709-1713. Metal foam. (n.d.). Retrieved November 10, 2008 from http://en.wikipedia.org/wiki/Metal_foam Paek, J. W., Kang, B. H., Kim, S. Y., & Hyun J. M. (2000). Effective thermal conductivity and permeability of aluminum foam materials. International Journal of Thermophysics, 21(2), 453-464. Refractory open-cell foams: carbon, ceramic and metal. (n.d.). Retrieved October 24, 2008 from http://www.ultramet.com/refractoryopencells_thermalmanagement.html Rowe, M. M., Guth, R. A., & Merriman, D. J. (2005) Case studies of carbon foam tooling. SAMPE 2005, Long Beach, CA, May 1-5. Roy, A. (2008). Personal Communication, Air Force Research Laboratory. Sihn, S., & Roy, A. K. (2004). Modeling and prediction of bulk properties of open-cell carbon foam. Journal of the Mechanics and Physics of Solids, 52(1), 167-191. Sihn, S., & Rice, B. P. (2003). Sandwich construction with carbon foam core materials. Journal of Composite Materials, 37(15), 1319-1336. Spradling, D., & Guth, R. (2003). Carbon foams. Advanced Materials & Processes, 161(11), 29-31. 114 Springer, G. S., & Tsai, S. W. (1967). Thermal Conductivities of Unidirectional Materials. Journal of Composite Materials, 1(2), 166-173. Straatman, A. G., Gallego, N. C., Yu, Q., Betchen, L., & Thompson, B. E. (2007). Forced convection heat transfer and hydraulic losses in graphitic foam. Transactions of the ASME. Journal of Heat Transfer, 129(9), 1237-1245. Twigg, G., Poursartip, A., & Fernlund, G. (2004a). Tool–part interaction in composites processing. Part I: experimental investigation and analytical model. Composites Part A, 35(1), 121-133. Twigg, G., Poursartip, A., & Fernlund, G. (2004b). Tool–part interaction in composites processing. Part II: numerical modeling. Composites Part A 35(1), 135-141. Vafai, K., & Tien, C. L. (1981). Boundary and inertia effects on flow and heat transfer in porous media. International Journal of Heat and Mass Transfer, 24(2), 195-203. Vafai, K., & Tien, C. L. (1982). Boundary and inertia effects on convective mass transfer in porous media. International Journal of Heat and Mass Transfer, 25(8), 1183-1190. White, S. R. & Kim, Y. K. (1998). Process-induced residual stress analysis of AS4/3501-6 composite material. Mechanics of Advanced Materials and Structures, 5(2), 153-186. 115 Wiersma, H. W., Peeters, J. B., & Akkerman, R. (1998). Prediction of springforward in continuous-fiber/polymer L-shaped parts. Composite Part A, 29A, 1333-1342. Yu, Q., Thompson, B. E., & Straatman, A. G. (2006). A unit cube-based model for heat transfer and fluid flow in porous carbon foam. Transactions of the ASME. Journal of Heat Transfer, 128(4), 352-360. Zhu, Q., Geubelle, P. H., Li, M., & Tucker, C. L., III. (2001). Dimensional accuracy of thermoset composites: simulation of process-induced residual stresses. Journal of Composite Materials, 35(24), 2171-2205. 116 APPENDIX A: RELATIONS FOR THERMAL AND MECHANICAL PROPERTIES OF COMPOSITE MATERIALS In the following relations, subscript “f” denotes a carbon fiber property and subscript “m” denotes an epoxy matrix property. Density and specific heat of the composite material can be calculated from constituents’ density and specific heat using the rule of mixture: ρ = vρ f + (1 − v )ρ m cp = vρ f c pf + (1 − v )ρ m c pm ρ Longitudinal and transverse thermal conductivities of the composite material can be calculated from constituents’ thermal conductivity using the rule of mixture and the relations introduced by Springer and Tsai (1967): k l = vk f + (1 − v )k m ⎛ D 2 v ⎞⎟ ⎜ 1− ⎛ v ⎞ km ⎜ 4 π ⎟ ⎟+ k t = k m ⎜⎜1 − 2 tan −1 π− ⎜ ⎟ ⎟ π⎠ D v ⎟ D 2v ⎝ ⎜ 1+ D 1− ⎜ π ⎟⎠ π ⎝ ⎛k ⎞ D = 2⎜ m − 1⎟ ⎜k ⎟ ⎝ f ⎠ Lamina composite mechanical properties can be calculated from constituents’ mechanical properties using the following micromechanics model introduced by Bogetti and Gillespie (1992). The relations are written for a composite material that is transversely isotropic in the 2-3 plane (the fibers oriented in the 1-direction). 117 The longitudinal elastic modulus: E1 = E1 f v + E m (1 − v ) + ( ) 4 ν m − ν 122 f K f K m Gm (1 − v )v (K f + Gm )K r + (K f − K m )Gm v The in-plane Poisson’s ratio: ν 12 = ν 13 = ν 12 f v + ν m (1 − v ) + (ν (K − ν 12 f )(K m − K f )Gm (1 − v )v m f + Gm )K m + (K f − K m )Gm v The in-plane shear modulus: G12 = G13 = Gm G12 f + Gm + (G12 f − Gm )v G12 f + Gm − (G12 f − Gm )v The transverse elastic modulus: E 2 = E3 = 1 ν 122 1 1 + + 4 K T 4G23 E1 The transverse Poisson’s ratio: ν 23 = 2 E1 K T − E1 E 2 − 4ν 122 K T E 2 2 E1 K T The transverse shear modulus: G23 = Gm K m (G23 f + Gm ) + 2G23 f Gm + K m (G23 f − Gm )v K m (G23 f + Gm ) + 2G23 f Gm − (K m + 2Gm )(G23 f − Gm )v The longitudinal coefficient of thermal expansion α1 = vα 1 f E1 f + (1 − v )α m E m (ξ ) α 1 f E1 f + α m E m (ξ ) 118 The transverse coefficient of thermal expansion α 2 = α 3 = (α 2 f + ν 12 f α 1 f )v + α m (1 + ν m )(1 − v ) − [ν 12 f v + ν m (1 − v )]α 1 The longitudinal and transverse coefficients of chemical expansion are obtained by replacing fiber and matrix coefficients of thermal expansion by their respective coefficients of chemical expansion. In the above relations, the isotropic plane strain bulk modulus and the effective plane strain bulk modulus of composite are (Bogetti & Gillespie, 1992): K= E 2 1 − ν − 2ν 2 ( KT = (K f ) + Gm )K m + (K f − K m )Gm v (K f + Gm ) − (K f − K m )v The stiffness matrix of the transversely isotropic composite material is: C ijkl ⎡C1111 ⎢C ⎢ 1122 ⎢C = ⎢ 1122 ⎢ 0 ⎢ 0 ⎢ ⎣⎢ 0 C1122 C1122 0 0 C 2222 C 2233 0 0 C 2233 C 2222 0 0 0 0 0 0 C1212 0 0 C1212 0 0 0 0 where the matrix elements are: C1111 = E1 1 − ν 23ν 32 Λ C 2222 = E 2 1 − ν 13ν 31 Λ C1122 = E1 ν 21 + ν 31ν 23 Λ 0 ⎤ 0 ⎥⎥ 0 ⎥ ⎥ 0 ⎥ 0 ⎥ ⎥ C 2323 ⎦⎥ 119 C 2233 = E 2 ν 32 + ν 12ν 31 Λ C1212 = G12 C 2323 = G23 Λ = 1 − ν 12ν 21 − ν 23ν 32 − ν 31ν 13 − 2ν 21ν 32ν 13 ν ij = Ei ν ji Ej i, j = 1, 2, 3 120 APPENDIX B: MATHEMATICAL MODEL FOR TURBULENT FLUID FLOW AND HEAT TRANSFER The mathematical model for turbulent fluid flow and heat transfer as implemented in commercial CFD software FLUENT (Fluent Inc., 2006) is used for performing the simulations in this research. This model is summarized below. In the Reynolds-Averaged Navier-Stokes (RANS) method for modeling fluid flow turbulence, the instantaneous fluid velocities and pressure are decomposed into the time-averaged components (overbar) and the fluctuating components (prime) ui = ui + ui ' p = p + p' and then substituted into Equations (4.12) and (4.13). After taking the time-average and omitting the overbar on ui and p , the fluid flow governing equations are ∂u i =0 ∂xi ρu j ∂u i ∂p ∂ =− +μ ∂x j ∂xi ∂x j ⎛ ∂u i ⎜ ⎜ ∂x ⎝ j ⎞ ∂ ⎟+ − ρu i ' u j ' ⎟ ∂x j ⎠ ( ) The Boussinesq hypothesis gives the Reynolds stresses as a function of the average velocity gradients (Hinze, 1975) ⎛ ∂u ∂u j − ρ ui ' u j ' = μ t ⎜ i + ⎜ ∂x ⎝ j ∂xi ⎞ 2 ⎟ − ρkδ ij ⎟ 3 ⎠ 121 where μ t is the turbulent viscosity, k is the turbulence kinetic energy and δ ij is the Kroneker delta. The turbulent viscosity and turbulence kinetic energy are modeled by employing the k − ε method. This method introduces two additional equations: ⎡ ∂ (ρku j ) = ∂ ⎢⎛⎜⎜ μ + μ t ∂x j ∂x j ⎢⎣⎝ σk ⎞ ∂k ⎤ ⎟⎟ ⎥ + Gk − ρε ⎠ ∂x j ⎥⎦ ⎡ ∂ (ρεu j ) = ∂ ⎢⎛⎜⎜ μ + μ t ∂x j ∂x j ⎣⎢⎝ σε ⎞ ∂ε ⎤ ε2 ⎟⎟ ⎥ + ρC1 Sε − ρC 2 k + νε ⎠ ∂x j ⎦⎥ where the constants are given by ⎡ η ⎤ C1 = max ⎢0.43, η + 5 ⎥⎦ ⎣ σ k = 1.0 C 2 = 1.9 η=S k ε σ e = 1.2 and ε is the turbulence dissipation rate. The generation of turbulence kinetic energy is given by Gk = μ t S 2 S = 2S ij S ij where S ij = 1 ⎛⎜ ∂u j ∂u i + 2 ⎜⎝ ∂xi ∂x j ⎞ ⎟ is the rate of strain tensor of the fluid. ⎟ ⎠ The turbulent viscosity is given by μ t = ρC μ k2 ε 122 Cμ = 1 A0 + As ~ ~ U = S ij S ij + Ω ij Ω ij kU ε ~ Ω ij = Ω ij − 2ε ijk ω k Ω ij = Ω ij − ε ijk ω k where Ω ij is the is the mean rate of rotation tensor viewed in a rotating reference frame with the angular velocity ω k . The model constants are computed from As = 6 cos ϕ A0 = 4.04 1 3 ( ϕ = cos −1 6W ) W = S ij S jk S ki ~ S3 ~ S = S ij S ij The fluid temperature distribution in turbulent fluid flow is governed by the energy conservation equation (4.14) where the fluid thermal conductivity k f is replaced by fluid effective thermal conductivity computed from keff = k f + Prt = 0.85 c p μt Prt
© Copyright 2024 Paperzz