Buckling Of Isogrid Plates by Jeffrey Lavin An Engineering Project Submitted to the Graduate Faculty of Rensselaer Polytechnic Institute in Partial Fulfillment of the Requirements for the degree of MASTER OF ENGINEERING IN MECHANICAL ENGINEERING Approved: _________________________________________ Ernesto Gutierrez-Miravete, Project Adviser Rensselaer Polytechnic Institute Hartford, CT June 2010 i © Copyright 2010 by Jeffrey Lavin All Rights Reserved ii CONTENTS List of Figures ................................................................................................................... iv List of Tables ..................................................................................................................... v List of Symbols ................................................................................................................. vi Abstract ............................................................................................................................ vii 1. Introduction/Background ............................................................................................. 1 1.1 Problem Description ........................................................................................... 1 1.2 Methodology ....................................................................................................... 2 1.3 Expected Outcome .............................................................................................. 3 2. Buckling Theory And Analytical Solutions ................................................................ 4 2.1 Isogrid Simplification ......................................................................................... 5 2.2 Plate Buckling Analytical Solutions ................................................................... 8 3. Numerical Results and Discussion ............................................................................ 13 3.1 Model Creation and Explanation ...................................................................... 13 3.2 Model Parameters and Boundary Conditions ................................................... 14 3.3 Plate Critical Load Comparison........................................................................ 19 3.4 Isogrid Critical Load Comparison .................................................................... 20 3.5 Mode Shape Visualization ................................................................................ 21 3.6 Mode Shape Comparison .................................................................................. 23 3.7 Rib Geometry Variation ................................................................................... 26 4. Conclusions................................................................................................................ 29 References........................................................................................................................ 32 Appendixes ...................................................................................................................... 33 iii List of Figures Figure 1: Free body diagram............................................................................................. 2 Figure 2: Off design condition free body diagram ........................................................... 2 Figure 3: Fixed-free beam ................................................................................................ 4 Figure 4: Column buckling example ................................................................................ 5 Figure 5: Typical isogrid pattern ...................................................................................... 6 Figure 6: Transformed isogrid structure ........................................................................... 7 Figure 7: Isogrid geometry ............................................................................................... 9 Figure 8: FBD load case 1 and 2 ...................................................................................... 9 Figure 9: FBD load case 3 .............................................................................................. 11 Figure 10: Loading applied to edge a and b ................................................................... 12 Figure 11: Cross section geometry comparison.............................................................. 14 Figure 12: Isogrid edge conditions ................................................................................. 15 Figure 13: Isogrid point constraints ................................................................................ 15 Figure 14: Initial isogrid finite element model ............................................................... 16 Figure 15: Final isogrid geometry .................................................................................. 17 Figure 16: Final isogrid boundary conditions ................................................................. 18 Figure 17: Final isogrid load case 1 ................................................................................ 18 Figure 18: Element count vs. Isogrid 1st Critical Buckling Load ................................... 18 Figure 19: Plate model showing first buckling mode shape and load ............................ 22 Figure 20: Plate model showing second buckling mode shape and load (m=2, n=1) .... 22 Figure 21: Second mode shape at section A-A (mid span) m=2 .................................... 22 Figure 22: Second mode shape at section B-B (max displacement) n=1 ....................... 22 Figure 23: Isogrid and plate model mode shape 1-5 comparison loaded on edge b ....... 24 Figure 24: Isogrid and plate model mode shape 1-5 comparison loaded on edge a ....... 25 Figure 25: Isogrid and plate model mode shape 1-5 comparison loaded on both edges 26 Figure 26: Example of rib buckling mode ...................................................................... 27 Figure 27: 1st Critical buckling load as a function of .................................................. 28 iv List of Tables Table 1: Transformed section properties using parallel axis theorem .............................. 7 Table 2: Material properties for 1” x 1” plate................................................................. 10 Table 3: 1” x 1” Critical buckling loads .......................................................................... 10 Table 4: Buckling load/mode change with edge loading................................................ 11 Table 5: Critical values for loading on both edges 1.000” x 1.1547” plate .................... 12 Table 6: Material property comparison .......................................................................... 15 Table 7: Model Element Count Comparison .................................................................. 17 Table 8: 4.000” x 4.618” Critical load (lb) on edge b .................................................... 19 Table 9: 4.000” x 4.618” Critical load (lb) on edge a and b ........................................... 19 Table 10: Isogrid model critical loads (lb) comparison with load applied to edge b ..... 20 Table 11: Isogrid model critical loads (lb) comparison load with applied to edge a...... 20 Table 12: Isogrid model critical loads (lb) comparison, load applied to both edges ...... 21 Table 13: Rib study parameter variation ........................................................................ 27 v List of Symbols Symbol Description Units E0 Elastic Modulus psi E* Equivalent Elastic Modulus psi t Skin thickness in t* Equivalent plate thickness in D Bending Stiffness lb/in K Tensile Stiffness lb/in P, F Critical Buckling Load lb a Horizontal Edge Length in b Vertical Edge Length in l Column Length in Dimensionless Rib Height - Dimensionless Cap Parameter - Dimensionless Cap Height - Dimensionless Rib Parameter - h Isogrid height in Poisson Ratio - A Area in2 I Cross Sectional Inertia in4 Centroid Distance in vi Abstract The focus of this report will be the buckling of an isogrid structures. The simplification reduces the isogrid structure to a single sheet with an equivalent bending and tensile stiffness by creating and equivalent modulus of elasticity (E*) and thickness (t*). This simplification is often used in structural analysis models to reduce engineering lead-time. However, the results from these models must still produce accurate results. A simply supported isogrid and equivalent stiffness plate model will both be subjected to varying load orientations. The orientation of the load relative to the isogrid pattern will vary to ensure the simplification is applicable to parts subject to varying load conditions. The accuracy of the critical buckling loads and predicted mode shape for both the isogrid and the plate models will be compared to an analytical solution with similar boundary conditions and loads. Finally, a recommendation for when the simplification process is no longer valid due to a change in failure mode will also be discussed. vii 1. Introduction/Background Isogrid structures are used in a variety of aerospace components. The design of the isogrid structure allows the part to maintain isotropic properties even though material has been removed for weight reduction. Due to their wide use in applications they are subjected to a variety of boundary constraints and loading conditions all of which can vary during operation. Initial designs are completed with the best available predicted loads that must be validated during engine test. Even with advances in technology the predicted temperature and pressures are not exact and these changes vary the component-to-component load direction and orientation. In order to maintain structural integrity the isogrid must continue to function properly if load orientation is changed during operation. Thus the simplification process must also capture the change in critical buckling load and mode shape if load orientation is changed. 1.1 Problem Description The problem is formulated from an existing isogrid part experiencing deformation attributed to a change in boundary conditions and loading. To reduce design lead-time the isogrid part was simplified to a single sheet. This simplification was developed for NASA and is known as the E*t* method. The simplification process of the isogrid is noted as a potential cause of failure and is being reviewed to ensure this is not the cause. The part is designed with specific load directions and boundary conditions. The change in load direction is also a potential cause of failure. A simplified free body diagram of the component with the correct loading condition is seen in figure 1. Figure 2 shows a possible change in the boundary conditions that could change the input loads to the part. These changing load directions are also a potential cause of the part deformation. For this reason three different load orientations are used in the report to ensure the load orientation relative to the isogrid is not a reason for failure. 1 y x a b Figure 1: Free body diagram y x a b Figure 2: Off design condition free body diagram 1.2 Methodology A closed form solution is solved for each loading condition for comparison to the numerical analysis. The E*t* method is used to simplify the isogrid to a plate. This simplification will be used as an input to the analytical solution as well as used in a numerical analysis of the plate. These two solutions will be compared to ensure proper boundary condition modeling technique as well as mesh density. The plate model will also be used for comparison to the modeled isogrid geometry. Each numerical model is created using the Comsol finite element code. In Comsol both a static solution and a linear buckling solution is completed. The static solution is used to verify the input load and that the initial boundary conditions do not impart any additional constraint or load. Upon verification of the boundary conditions and input loads the linear buckling analysis is completed. This analysis is completed with a unit load so that the calculated eigen value is the critical buckling load. These loads are then compared to the analytical solution for verification of the E*t* 2 methodology. Analytical buckling mode shapes will also be compared to numerical mode shapes of the plate and isogrd to ensure the numerical model captures the correct buckled shape. 1.3 Expected Outcome It is expected that the change in plate size and load orientation will cause a change in the critical buckling loads and mode shapes. It is expected that the E*t* method can be used to accurately predict the buckling mode shapes and critical loads of isogrid plates to a certain extent. It is also expected that there will be additional modes not captured by the E*t* method due to the ribs and smaller panels created by the ribs. 3 2. Buckling Theory And Analytical Solutions The problem will begin with the basics of buckling and steadily progress in complexity. All of the problems considered below have a closed form solution. In order to verify the numerical results and modeling approach each model result will be compared to the closed form solution. This allows verification of each analysis step to ensure the outcome is accurate. The first problem focused on creating a fixed-free beam with a single load applied in the vertical direction. A diagram of this can be seen in figure 3. The critical load for this problem is shown in equation 1. P l Figure 3: Fixed-free beam Pcr 2 EI 4l 2 [1] This problem can be turned into the most common or fundamental case, which consists of a beam pinned at both ends. An example of this can be seen in figure 4. This problem has the same solution as the fixed-free case assuming the critical length is now l 2 and a symmetrical boundary condition at the center of the column. The solution to this fundamental case is seen in equation 2 in which the critical value occurs at n = 1. This fundamental case is the most often assumed condition in analysis [2] and will be the basis for the simply supported plate boundary condition. 4 l/2 l/2 Figure 4: Column buckling example n 2 2 EI P l2 [2] The symmetry condition allows the finite element model to be simplified. A freefree beam will have a rigid body motion due to the lack of constraint in one direction. Thus this simplified symmetric model is better suited for numerical analysis when applicable. The problem under consideration is not a 2D structure but rather a complex 3D part that contains an isogrid pattern. As previously stated the isogrid is used to reduce the weight of the structure while still maintaining the isotropic material properties of a single sheet of material. 2.1 Isogrid Simplification A typical isogrid structure is used to reduce weight while maintaining structural efficiency. However it is difficult to model and often requires multiple iterations to obtain the correct stiffness required in the design. Thus it saves design iteration time if the structure can be turned into an equivalent single sheet with representative stiffness. A typical isogrid structure can be seen below in figure 5. The plate will not have the same geometric dimensions as the isogrid but it will have the same stiffness in both the tensile and bending directions. This simplification allows multiple design iterations to be completed by changing only the stiffness of the part and not the model geometry. The simplification also allows for a reduction in computational time. 5 QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. Figure 5: Typical isogrid pattern Assuming the isogrid plate is in a state of uniaxial stress it can be shown that the structure is equivalent to a single sheet in plane stress [1]. The process to simplify the sheet uses several non-dimensional parameters ( , , , , h ) for a unit width of isogrid. The procedure also uses the parallel axis theorem to reduce the isogrid to a single sheet. The non-dimensional parameters are defined in equation 3. The single sheet will have an equivalent tensile stiffness (K) and bending stiffness (D), where is the material Poisson’s ratio and E0 is the material elastic modulus. The different stiffness equations are shown in equation 4 and equation 5. 3 d c bd wc , , , ,h a t t th th 2 D 1 E0 I 1 2 [4] K 1 E0 A 1 2 [5] [3] The procedure reduces any isogrid geometry down to a unit width of isogrid, which is created from the variables in equation 3. The transformed isogrid is shown in figure 6. 6 QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. Figure 6: Transformed isogrid structure Ai i Aii Aii2 I0 1 t t (1 ) 2 t2 (1 ) 2 t3 (1 )2 4 t 2 t 12 2 t 0 0 0 t ( t)2 12 3 t t ( ) 2 t3 ( )2 4 t (t)2 12 Total t(1 ) t3 (1 )2 ( )2 4 t3 1 2 2 12 Part t2 ( ) 2 t2 (1 ) ( ) 2 Table 1: Transformed section properties using parallel axis theorem To complete the transformation the individual areas (Ai), centroids (i) and moment of inertia (I0) are calculated with the equations in table 1. This table contains all the required information to calculate the appropriate stiffness properties for the equivalent thickness plate. The final stiffness for the plate is calculated from the total section properties. These are seen in equations 6, 7 and 8. A Ai t [6] A i i A I Ai i2 I 0i A i2 7 [7] [8] The equations for both I and A are now in terms of variable t only. Thus with two equations for stiffness (K, D) and two unknowns (E, t) equation 4 and 5 are solved simultaneously to determine an equivalent thickness and stiffness. The solution of the two equations creates a sheet with an equivalent thickness (t*) and an equivalent elastic modulus (E*) that must be used together to create the required tensile and bending stiffness. By producing the correct stiffness with t* and E* these variables can be used together to predict loads but not stress. Thus the combination is used for the calculation of critical buckling loads. Equations for t* and E* are shown in equation 9 and equation 11. Equation 10 can be used to calculate the internal parameter , which can be used to calculate t* from only non-dimensional parameters. The parameters E* and t* will be used when comparing an equivalent single sheet to the isogrid structure. t* 12I t A 1 [9] 2 2 2 2 2 (1 ) 3(1 ) 3 ( ) 1 3(1 ) ( ) E * E0 A t* 2 [10] [11] With equation 9 and 11 it can be shown that equation 4 and equation 5 can be put in terms of both E* and t* where D and K are now represent the correct stiffness for the isogrid. These can be seen in equation 12 and equation 13. D E *t *3 12(1 v2 ) E *t * K 1 2 [12] [13] 2.2 Plate Buckling Analytical Solutions Similar to the initial 2D applications the initial problem 3D plate problems were compared to known analytical solutions to ensure completeness. The first case is a simply supported plate in uni-axial compression. The plate was 1” x 1” with a thickness 8 of .040”. The second case used was a 1.000” x 1.1547” plate with a thickness of .046”. This thickness is derived from the isogrid geometry shown in figure 7 and the equations from section 2.1. The isogrid is representative of the geometry found in the failing component. The face sheet thickness is .030” and the ribs are .050” tall by .040” wide with a triangle side length (a) of 1.154” and a triangle height (h) of 1.000”. The two cases will be used to verify the numerical results for a plate with varying height to width ratio as well as thickness. .040У .030У .050У 1.000У Figure 7: Isogrid geometry The plate has all edges simply supported. This free body diagram is similar to a section of the component in which the suspected change in boundary conditions is being observed. The part is free in the y direction and interference is predicted in the x direction. See figure 8 for a free body diagram of the structure. The buckling load for this problem is calculated using equation 14. The critical value will occur with n equal to 1, where both m and n are integers. This is similar to the beam model of figure 4 and equation 2. The value of m corresponds to the number of half waves parallel to the direction of loading while n determines the number of half waves perpendicular to the direction of loading. These equations can be further simplified for m n 1 in a square plate and are shown in equation 15. y a x b Figure 8: FBD load case 1 and 2 9 2 a 2 2 D m 2 n 2 Px b m 2 a 2 b 2 Px b 4 2 D a2 where D [14] Et 3 12(1 2 ) [15] For the 2 cases above the first 5 critical buckling load values are shown in table 3. The material properties used in the calculations are shown in table 2. The modulus of elasticity (E*) is calculated from the isogrid simplification of the previous section however for this comparison there is no requirement to use the value of (E*). The calculated values will be compared to the numerical results of the next section. E* 20.83e6 psi v .3 Table 2: Material properties for 1” x 1” plate 1.000” x 1.1547” Plate 1.000” x 1.000” Plate b = 1.000 Fcritical (lb) m n Fcritical (lb) m n 4818 1 1 7529 1 1 7529 2 1 9833 2 1 13385 3 1 16406 3 1 19274 2 2 25966 4 1 21758 4 1 30115 2 2 Table 3: 1” x 1” Critical buckling loads From the data in table 3 it is seen that the mode shapes of the plate change as a function of the ratio of a/b. This can be seen by the change in critical load and the values of m and n between the fourth and fifth values of the calculated examples. The third case used was a change in orientation of the loading on the plate. This loading is shown in figure 9. For the 1x1 plate there is no change in critical buckling 10 loads but for the rectangular plate of case 2 (1.000 x 1.1547) there is a change to the critical loads. Again this is caused by the ratio of a/b and the change in load orientation. The comparison of calculated loads can be seen in table 4. y x a b Figure 9: FBD load case 3 Load on edge a Load on edge b a = 1.000” Fcritical (lb) m b = 1.1547” n Fcritical (lb) m n 6520 1 1 7529 1 1 12008 2 1 9833 2 1 22487 3 1 16406 3 1 26080 2 2 25966 4 1 34064 3 2 30115 2 2 Table 4: Buckling load/mode change with edge loading Additionally the critical loads were calculated for loading on both of the edges of the plate. This load case is most similar to the loading of the failing part in the field. The additional load on the plate further reduces the load required to buckle the plate. The applied loads are seen in figure 10. 11 y x a b Figure 10: Loading applied to edge a and b This load case has a solution shown in equation 16. Again the changing values of m and n change the mode shape and the critical load required to produce buckling in the plate. As expected the combined load case has reduced the critical value below the previous two load cases. The calculation of the critical values can be seen in table 5. 2 (mb a)2 n 2 2D Px mb a 2 (Pyb Px a)n2 b Calculated (lb) 3494 9872 7631 13976 14541 Kcc 1.9 5.4 4.1 7.6 7.9 m 1 1 2 2 3 [16] n 1 2 1 2 1 Table 5: Critical values for loading on both edges 1.000” x 1.1547” plate Each of these load cases will be compared to the simplified isogrid structure to determine the applicability of the E*t* method. 12 3. Numerical Results and Discussion Comsol was used to solve the isogrid buckling and equivalent plate buckling numerical analysis. The results in this section are then compared to the analytical solutions from section 2 to determine the validity of both numerical models. The numerical results are also compared to each other to ensure the use of the equivalent plate method will provide similar numerical answers to the isogrid model. Each model created is required to compute a static solution prior to calculating a buckling solution. This is required so that Comsol can calculate the pre-stress in the model. The pre-stress is required for the calculation of the stiffness matrix needed in the eigen buckling value solution. 3.1 Model Creation and Explanation There were four different models created to verify the applicability of the E*t* method. Each model was created using the gravitational IPS units. The two main variables elastic modulus (E and E*) and thickness (t and t*) have units of psi and inches. The first model created was a 1.000” x 1.1547” x 0.046” plate. This plate model corresponds to a case completed in section 2.2 and was the beginning of the E*t* verification. The plate model was also used to calibrate the Comsol modeling technique to the analytical solution. The plate is a constant thickness (t* = 0.046”) and given modulus of elasticity (E* = 20.83e6 psi). To match the plate model described above a single isogrid panel was created for comparison. The geometry was created using a single block (1.000” x 1.1547” x 0.030”) and adding individual ribs. The ribs were created at the center of the plate and then rotated 60 about the center in either direction. Each rib was 1.500” x 0.040” x 0.050”. These ribs were then trimmed with separate blocks to create the proper length. The final rib was then created at the center of the plate and formed the last piece of the isogrid. The third model created was a 4.000” x 4.618” x 0.046” plate. The plate was again modeled from the simplification of the isogrid, which required a constant thickness (t*) of 0.046”. The modulus of elasticity (E* = 20.83e6 psi) used was also the same as the original 1.000” plate and calculated using the E*t* process from section 2. 13 The final model created was an isogrid geometry that maintained the same rib length (s) and height (h) of the initial 1.000” model but contained more isogrid cells. The model was created using a similar technique to the first model but after rotating the diagonal ribs each rib was arrayed in the x direction to create multiple entities from the original. The array process allowed for a reduced number of modeling steps. Once the rotated ribs were created the final horizontal ribs were added and united to the rotated ribs. Creating a composite object of just the ribs allowed for a reduction in the number of subtractions required to create diagonal ribs with the proper length. The model now had two components. The face sheet of constant thickness and the rib structure were left as separate entities so that a rib height variation could also be completed. The rib height study models were each created separately by scaling the ribs of the original isogrid model in the z direction. This allowed for the face sheet thickness to remain unchanged for each separate analysis so that only the change in rib geometry was evaluated. 3.2 Model Parameters and Boundary Conditions The isogrid dimensions seen in figure 11 were used along with the procedure described in section 2.1 to produce the appropriate thickness (t*) and modulus of elasticity (E*). These computations were completed in excel to simplify the calculation effort. A comparison of the geometries is shown in figure 11 and table 6 shows the comparison of thickness and elastic modulus. .046У 1.000У .040У .030У .050У 1.000У Figure 11: Cross section geometry comparison 14 Thickness Elastic Modulus (lb/in2) Plate .046” 20.83e6 (E*) 30.00e6 (E0) Isogrid As drawn Table 6: Material property comparison The boundary conditions for the isogrid were identical to the plate. The isogrid boundary conditions are shown in figure 12. Each side of the isogrid is simply supported to match the analytical solution constraints. This requires support in the vertical z direction with additional point constraints at specific points in order to prevent a rigid body motion. The point constraints are at the center of each edge and constrain movement in the direction parallel to the edge. The point constraints are seen in figure 13. Figure 12: Isogrid edge conditions Figure 13: Isogrid point constraints The geometry for both the equivalent sheet and the isogrid were meshed with tetrahedral 3D solid elements. The mesh geometry for the isogrid is shown in figure 14. The mesh contained 25423 elements and 126690 degrees of freedom. The analytical 15 solution of a plate with a single edge load produces a critical buckling load of 7529 lb. The critical buckling load for the isogrid finite element model was computed at 6735 lb and the finite element model of the plate calculated a critical buckling load of 7180 lb. This is within 10.5% and 4.5% respectively of the analytical solution for a flat plate of equivalent stiffness. Figure 14: Initial isogrid finite element model These predictions were not acceptable for verification of the equivalent stiffness method so a larger model was completed to remove the influence that the boundary conditions may have on the results. The final model was a 4.000” x 4.618” rectangular isogrid. The lengths a and h (figure 5) for the isogrid were kept the same so that the E*t* simplification did not change between the small and large models. The isogrid simplification reduces the system to a unit width, which remains applicable to the larger system, provided the geometry is produced properly. All final results presented will be from the larger model. The isogrid model mesh, boundary conditions and loads can be seen in figures 1517. The final mesh consisted of 16358 elements and 96660 degrees of freedom. The free mesh parameters were set to “coarser” to create the mesh. A study was completed using the “extremely coarse” and “extra coarse” free mesh parameter option. Element count versus percent error to the first critical buckling load is shown in table 7. As the element count increased the accuracy of the solution increased. A graph of the data from table 7 can be seen in figure 18. The slope of the graph shows little change in accuracy for the increase in element count once ~10000 elements are used. The final element count provided accurate results while allowing the model to solve in approximately 5 minutes. 16 A similar study was also completed for the plate model. The model was run with the “coarser”, “coarse”, “normal” and “fine” free mesh parameters. This study showed that the model converged to a solution and the “normal” free mesh parameter was used. This produced a model with 12203 elements. Unlike the isogrid model, the plate model under-predicts the first critical buckling load. Model Element Count Isogrid 6716 10076 16358 4408 7668 12203 17432 Plate Analytical Solution (lb) 1882 1882 1882 1882 1882 1882 1882 First Critical Buckling Mode (lb) 2049 1972 1946 1870 1862 1858 1857 Percent Error 8.86 4.77 3.39 -.65 -1.07 -1.28 -1.33 Table 7: Model Element Count Comparison The boundary conditions were again applied to approximate a simply supported plate and boundary loads were applied to the main face of the plate. The loads were applied to the boundary face and applied so that the total input load was 1 lb per each side. Additional models were created with loads applied to additional faces to simulate each of the load cases in section 2.2. This included loading on the long edge and both edges. Figure 15: Final isogrid geometry 17 Figure 16: Final isogrid boundary conditions Figure 17: Final isogrid load case 1 st Critical Buckling 2100 2000 Numerical Solution Anaytical Solution 1900 1 st Critical Buckling Value (lb) Element Count vs. 1 1800 6000 8500 11000 13500 16000 18500 Element Count Figure 18: Element count vs. Isogrid 1st Critical Buckling Load 18 3.3 Plate Critical Load Comparison A comparison of the buckling mode shapes and the corresponding loads are shown in table 8 for the 4.000” x 4.618” plate with loading on edge b. As the mode shapes increase in complexity the accuracy of the model does not reduce. This shows the model is capable of predicting accurate displacement with the mesh. For the loads applied to edge b the first 4 modes should correspond to m 1 4 with the value n remaining constant at n = 1. The fifth mode being the first mode were the value of n = 2. Calculated 1882 2458 7529 4102 6492 Percent Error -1.28 -1.11 -1.27 -0.94 -0.87 Comsol Plate 1858 2431 7433 4063 6435 m 1 2 2 3 4 n 1 1 2 1 1 Table 8: 4.000” x 4.618” Critical load (lb) on edge b The change in loading direction will change the critical buckling value as well as the mode shape ordering. This was seen in the calculations completed in section 2.2. No additional modeling was completed for loads applied to edge a due to the accuracy of the loads on edge b. The model was also evaluated for loading applied to both sides (a and b) of the plate and the solution was compared to the analytical value using equation 16. The results compared well to the calculated value. The results can be seen in table 9. Calculated 874 2468 5128 1908 3494 3635 5215 Percent Error -0.75 0.13 1.38 0.02 0.71 1.18 -2.07 Comsol Plate 867 2471 5199 1908 3519 3678 5107 Table 9: 4.000” x 4.618” Critical load (lb) on edge a and b 19 3.4 Isogrid Critical Load Comparison To ensure the E*t* isogrid simplification method produces accurate critical loads all three load cases were run with the 4.000” x 4.618” size isogrid model. This will also provide substantiation that load orientation into the isogrid can be neglected. The 4.000” x 4.618” isogrid results are compared to the E*t* Comsol plate and the analytical solution. The model for the isogrid accurately predicted the critical buckling loads as compared to both the analytical value and the Comsol plate model. The results for the case with loads applied to edge b (short edge) can be seen in table 10. Calculated 1882 2458 7529 4102 6492 Percent Error -1.28 -1.11 -1.27 -0.94 -0.87 Comsol Plate 1858 2431 7433 4063 6435 Isogrid 1946 2543 7620 4223 6568 Percent Error 3.39 3.44 1.21 2.96 1.18 m 1 2 2 3 4 n 1 1 2 1 1 Table 10: Isogrid model critical loads (lb) comparison with load applied to edge b To ensure the model will capture the change in geometry (i.e. non-square) and load orientation, the model was run with a load applied to edge a (long edge). The error results are similar to the model with load applied to edge b. This showed the isogrid model and the E*t* method accurately predicted loads with a changing length ratio. The results for loading on edge a can be seen in table 11. Calculated 1630 8516 3002 6520 5622 Percent Error -1.47 -1.10 -1.21 -1.47 -1.08 Comsol Plate 1606 8422 2966 6424 5561 Isogrid 1684 8839 3114 6592 5760 Percent Error 3.3 3.8 3.7 1.1 2.5 m 1 1 2 2 3 n 1 2 1 2 1 Table 11: Isogrid model critical loads (lb) comparison load with applied to edge a The final case used to verify the E*t* method accurately calculates critical load is the combined loading on both edge a and edge b. This load case also produces critical 20 buckling loads similar to both the plate numerical and analytical analysis. The result for the model with loads applied to both edges is seen in table 12. Calculated 874 2468 1908 3494 3635 Percent Error -0.75 0.13 0.02 0.71 1.18 Comsol Plate 867 2471 1908 3519 3678 Isogrid 902.00 2556 1971.00 3532.00 3737.00 Percent Error 3.26 3.57 3.32 1.08 2.80 m 1 1 2 2 3 n 1 2 1 2 1 Table 12: Isogrid model critical loads (lb) comparison, load applied to both edges 3.5 Mode Shape Visualization Equations 14 and 16 can be used to calculate multiple buckling loads for the part depending on boundary conditions, the lowest load being the most important. As described in the sections above, the mode shape changes based on the value of m and n in equation 14 and equation 16. Mode shapes are reviewed to verify that Comsol is computing not only critical load but also the correct mode shapes. This is done by comparing the calculated buckling load and mode shape based on the values of m and n to the shapes and loads produced by Comsol. An example of the first mode shape for the plate can be seen in figure 19. The second mode shape for the plate can be seen in figure 20. The second mode shape is used as an example to show how the shape varies with the value of m and n. The corresponding displacements in figure 21 and figure 22 help visualize the values of m and n. 21 Figure 19: Plate model showing first buckling mode shape and load B A A B Figure 20: Plate model showing second buckling mode shape and load (m=2, n=1) Figure 21: Second mode shape at section A-A (mid span) m=2 Figure 22: Second mode shape at section B-B (max displacement) n=1 22 3.6 Mode Shape Comparison Despite accurate critical loads being calculated the mode shape of the linear buckling solution must be verified in order to ensure there are no modeling issues present. These issues can arise due to imperfections in the model geometry, boundary conditions or from anti-symmetric loads. They can also arise from abrupt stiffness changes that may be seen in the transition between the skin thickness and the ribs. Thus the final step to verifying the E*t* method is to compare the predicted mode shapes for the isogrid geometry with the mode shapes produced by the plate model and the analytical solution. Each mode produced from the isogrid model is compared to the plate model to verify the shapes are correct. The plate models can be compared to the analytical solution with specific values of m and n to determine if they are accurate. A comparison of the first 5 mode shapes for loading in along edge b is seen in figure 23. A mode shape comparison for loading along edge a is seen in figure 24. The combined loading condition mode shape comparison is seen in figure 25. 23 Figure 23: Isogrid and plate model mode shape 1-5 comparison loaded on edge b 24 Figure 24: Isogrid and plate model mode shape 1-5 comparison loaded on edge a 25 Figure 25: Isogrid and plate model mode shape 1-5 comparison loaded on both edges 3.7 Rib Geometry Variation Changing rib geometry relative to the plate changes the critical buckling loads without changing the buckling mode shapes until the ribs begin to dominate the stiffness of the structure. An example of the change in buckling mode shape for varying rib geometry can be seen in figure 26. This change in failure mode signifies when the E*t* method is no longer applicable to use to determine the critical buckling load or mode shape of the part. To determine the change in mode shape with respect to rib geometry a single load case was completed with loads applied to both edge a and edge b. Four different rib geometries are modeled with an increasing height. The width of the rib, height of the isogrid triangle and the skin thickness is all held constant. The heights used for the study 26 can be seen in table 13. The table also contains the rib geometry and the calculated value for As stated previously, the ribs were created in the Comsol model as one composite object and the skin was created as a separate object, which allowed for rib scaling in the z direction. For each new model a new value of E* and t* are calculated for use in the analytical solution. Each model was then compared to the analytical solution for verification of the critical buckling load. Additionally the mode shapes for each geometry change are evaluated to ensure the failure mode did not change from plate buckling to rib buckling. The variable chosen to determine when the simplification could no longer be used due to the change in failure mode is . This variable takes into account the rib cross-section as well as the length of the rib and skin thickness. A plot of vs. critical buckling load is seen in figure 27. Rib Width .040" .040" .040" .040" Rib Height .050" .100" .250" .500" 0.07 0.13 0.33 0.67 Table 13: Rib study parameter variation Figure 26: Example of rib buckling mode 27 1st Critical Buckling Load (lb) vs. 1 st Critical Buckling 250000 200000 150000 Plate mode Isogrid Mode 100000 50000 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Figure 27: 1st Critical buckling load as a function of 28 4. Conclusions Comsol accurately predicts both the critical buckling loads and mode shapes for simply supported flat plates. Several models of varying size were compared to analytical solutions and the final results were within 2%. There is good correlation between the two solution techniques. The 4” x 4.618” plate numerical model underpredicts the buckling load relative to the analytical solution by approximately 1.3%, while the isogrid model over-predicts the critical buckling load relative to the analytical solution by approximately 3.5%. The plate model is converged and it is assumed that with additional computational resources the isogrid model error could be reduced. This is shown in the comparison between element count and percent error for the isogrid model. This comparison and correlation provides the basis for using Comsol to predict the validity of the E*t* method. The model was completed as a full symmetry model for all cases to ensure the proper boundary conditions were established. Initially the model was completed with a symmetry boundary condition on two sides to simplify the modeling constrain. Upon comparison to the analytical solutions and the full symmetry model the symmetric model did not produce answers that matched either the analytical solution or the full symmetry numerical solution. It was assumed that the symmetry modeling constraints were either incorrectly applied or calculated incorrect loads and mode shapes. No further investigation was completed into why the symmetric model did not compare favorably to the analytical or full symmetry solutions. An improvement to the Comsol modeling package would include the ability to use 2D plate or shell elements and calculate buckling load. Using tetrahedral elements required a larger number of elements and degrees of freedom when compared to a similar model completed in Ansys with shell elements. This added to the computational time required for this analysis. Without the ability to complete a buckling run with plate or shell elements most sheet metal structures can be better analyzed with another finite element code. The isogrid design manual report [1] contained an error that was found during the creation of the E*t* excel sheet in the appendix. This error was confirmed with unpublished work completed at Pratt and Whitney. The error was in the calculation of 29 the distance to the centroid of the transformed isogrid cap. The problems involved in this report did not contain a capped isogrid, however the correction was included in the report for completeness. Without a capped isogrid model the correction has not been verified numerically. The load orientation into the isogrid model does not change the result from the E*t* method. The change in load orientation completed in section 3.5 and 3.6 shows that the simplification process does not effect the calculation of critical load or mode shape between the isogrid model and plate model. This allows the simplification process to be used despite load orientation into the isogrid. For the specific isogrid geometry modeled in figure 11 the E*t* method can be used to accurately predict both critical buckling loads and mode shapes. The prediction for both the critical loads are within 4% of the analytical values and within 6% of the numerically calculated critical loads. Comsol correctly produced the first five mode shapes when compared to the analytical solution. Only when buckling of the ribs occurred were incorrect mode shapes produced. The incorrect mode shapes were used to determine the applicability of the E*t* method to the geometry. With the geometry presented in this report the structure is dominated by skin buckling and not rib buckling. Rib buckling can occur with larger rib geometry as shown in the rib buckling section. With larger and stiffer ribs the isogrid simplification process does not predict the correct buckling load or mode shapes and the E* t* method should not be used. Based on the information in this report the E*t* method should not be used for values of > 0.23. For values of < 0.23 rib buckling must be verified prior to design finalization however the E*t* method should provide sufficient data. If the value of > 0.23 the model will no longer predict plate buckling modes but will instead predict rib buckling modes. These modes are also of concern but are not applicable to the E*t* method of analysis. Additionally care should be taken when using eigenvalue buckling during component design. The predicted critical load is incorrectly predicts values that are higher than the actual buckling value. For this reason eigen buckling should not be used for component design without adding additional safety margin during the design phase. 30 31 References [1] McDonnell Douglas Astronautics Company. 1973. Isogrid Design Handbook. CA. McDonnell Douglas Astronautics Company. [2] Timoshenko and Greer. 1961. Theory of Elastic Stability. NY. McGraw Hill Inc. [3] Brush and Almroth. 1975. Buckling of Bars, Plates, and Shells. NY McGraw-Hill Inc. 32 Appendixes 33