Craig Collier James Ainsworth Phil Yarrington Ryan Lucking

Ares V Interstage Composite Panel Concept and Ringframe Spacing Trade Studies
Craig Collier James Ainsworth Phil Yarrington Ryan Lucking Collier Research Corporation Hampton, VA © 2010 Collier Research Corp.

Outline Intro into the Ares V Heavy Launch Vehicle Summary Results
Interstage Summary Results Weight predictions Weight trends as a function of ringframe spacing (per panel concept) Focus on the hat stiffened panel Compare to the honeycomb sandwich as a baseline Describe why hat is lighter than sandwich Analysis and Sizing process used on this project with HyperSizer FEA verifications © 2010 Collier Research Corporation.

Ares V Composite Structures
© 2010 Collier Research Corporation.

Ares V Payload Shroud Ares V Payload Shroud Length = 77’
Diameter = 33’ Lightly loaded by Aerodynamic pressure Panels sized to minimum gage thickness Four segments for separation Panel concepts evaluated: Honeycomb, hat, skin stringer, RCS © 2010 Collier Research Corporation.

Ares V Core Intertank Ares V Core Intertank Length = 27.5’
Diameter = 33’ Highly loaded Axial compression forces In-plane shear forces Tension/compression hoop forces Unitized and manufacturing segments Panel concepts evaluated: Honeycomb, hat, skin stringer, reinforced core sandwich, PRSEUS © 2010 Collier Research Corporation.

Ares V Interstage Ares V Interstage Length = 47.5’ Diameter = 33’
Moderately loaded Axial compression forces Tension/compression hoop forces Unitized and manufacturing segments Panel concepts evaluated: Honeycomb, hat, skin stringer, reinforced core sandwich, PRSEUS, © 2010 Collier Research Corporation.

Weight Summary: All Ares V Composite Structures
Interstage Core Intertank Payload Shroud PRSEUS Unit Weight Ogive Barrel Honeycomb Sandwich Panel and Ringframe Total Smeared Unit Weight Reinforced Core Sandwich Hat Stiffened Panel All Loading Normalized to Max Axial Compression Internal Axial Load (Includes 1.4 load factor) © 2010 Collier Research Corporation.

Ares V Interstage Panel Concepts
Honeycomb Sandwich Bonded Hat Integral Blade Stiffened Reinforced Core Sandwich PRSEUS Integral Blade Sandwich Many panel concepts are considered and each concept is optimized to find the lightest weight combination of cross sectional dimensions, materials and layups based on ringframe spacing © 2010 Collier Research Corporation.

Ares V Interstage Loads

Ares V Interstage Internal Load Derivation
Unit Axial Unfactored Loads For ALL Stations Effective line load determined from flight loads All Loads in report normalized to maximum load Nx Fx Nx(m) Mz Mx(max) Fy Stn A L = 570 in Max Axial line load Nx(max) Stn B Constrained at B © 2010 Collier Research Corporation.

Ares V Interstage Hoop Load
Load sharing between ringframes and skin is considered Internal pressure causes tension hoop load External pressure and ringframe pinch causes compressive hoop load Note: End Frames are = ½ EA of internal ringframes, to achieve uniform hoop load gradient © 2010 Collier Research Corporation.

Ares V Interstage Load Cases
Cylinder FEM Internal Loads Load Case Panel Axial Load Nx Panel Hoop Load Load Case 1 or 101 Max Axial Compression Tension Load Case 2 or 102 Compression 2 Main load cases control panel dimensions and materials for Interstage Load Case 1: Material strength critical Load Case 2: Buckling Critical © 2010 Collier Research Corporation.

Total Weight Summaries

NASA Advanced Composite Technology Team
HyperSizer® used by the national team represented by NASA centers Langley, Glenn, Marshall, Ames, Goddard Results reported here are those from Collier Research Corporation – makers of the HyperSizer software Our reported weights are different than those of the team but show similar trends and relative values Our company put > 9000 hrs into effort in last year Experience sizing with the software for industry (wings and fuselages) and in particular with hats © 2010 Collier Research Corporation.

WML We rate our results with a high WML (Weight Maturity Level)
Strictly following HyperSizer best practices use recommendations for apples-to-apples evaluation Standardized team HyperSizer databases, materials, load factors, etc. Quality of FEA internal loads - resolution of joint effects on acreage panels, design and analyses of joints, convergence of load sharing between ringframes and panels Fully exploring sizing variables and laminate generation, redesign based on manufacturing review input, redesign, implementing >10 repair rules in sizing Performing independent analysis verification at all steps along during the design and analysis maturation process FEA verification of barrel buckling with ringframes FEA verification of panels Nx/Nastran, Nei/Nastran, Abaqus Linear static, eigenvalue, and nonliner post buckling In all >100 individual FEMs generated and FEA soln’s performed © 2010 Collier Research Corporation.

Interstage Weight Summary All Panel Concepts
Ringframe Spacing Panel Unit Weight (lb/ft2) Ringframe Unit Weight Total Unit Weight Hat Stiffened Panel 57 1.38 0.15 1.53 Reinforced Core Sand. 71 1.56 0.16 1.72 Honeycomb Sand. 1.77 0.09 1.86 Blade Sand. 114 0.05 1.91 Integral Blade 21 1.45 0.46 PRSEUS 52 2.02 N/A Hat is 21% lighter than honeycomb sandwich © 2010 Collier Research Corporation.

Comparison of Load Carrying Panel Weight
Minimum Laminate Weight for Strength Hat 43.86“ Spacing Hat 57" Spacing Sandwich ” Spacing (36% Core Wt) Sandwich 57“ Spacing (40% Core Wt) Sandwich 142” Spacing (45% Core Wt) © 2009 Collier Research Corporation.

Two Hat Panel Designs: Local Buckling at Limit and Local Buckling at Ultimate Loads
RF spacing = 43.86” Local buckling at limit load hat panel = 1.38 psf (blue curve) Local buckling at ultimate load hat panel = 1.41 psf (green curve) Both designs have post buckling strengths > 8000 lb/in © 2009 Collier Research Corporation.

HyperSizer Trade Study Analysis Process
Pure Panel Trade studies Internal Loads are known No FEM required Workspace Analysis Reinforced Core Sandwich Integral Blade Sandwich Honeycomb Sandwich Integral Blade Stiffened PRSEUS Bonded Hat © 2010 Collier Research Corporation.

Weight Trends plotted as a function of ringframe spacing FEA: Verify Eigenvalue Optimum solutions determined from trend lines Plot Weight Trend for Panel & Ring Spacing Workspace Analysis © 2010 Collier Research Corporation.

FEA: Verify Eigenvalue Optimum solutions determined from trend lines Plot Weight Trend for Panel & Ring Spacing Workspace Analysis Full Cylinder FEMs created with automated HyperSizer utility ‘CylFEMgen’ Full Cylindrical FEM Input: Workspace sizing Interstage Geometry Number of Ring frames Loads and Boundary conditions ESTIMATED RING EI Optimum Hat Stiffened Panel 9 Ring Frames 57” Ringframe Spacing Optimum Honeycomb Sandwich 7 Ring Frames 71” Ringframe Spacing © 2010 Collier Research Corporation.

FEA: Verify Eigenvalue Optimum solutions determined from trend lines Plot Weight Trend for Panel & Ring Spacing Workspace Analysis Use full cylinder models to iterate with FEA and Size Ring Frames Full Cylindrical FEM EI Specified (lb-in2) Resulting EA (lb) Beam Unit Weight (lb/ft) Panel Buckling EigV Buckling EigV > 2.15? Buckling Across Ringframe? 1.60E+08 1.72E+07 0.73 2.14 No Yes 1.70E+08 1.75E+07 0.74 2.148 1.75E+08 0.75 2.15 1.80E+08 1.77E+07 2.16 Input: Workspace sizing Interstage Geometry Number of Ring frames Loads and Boundary conditions ESTIMATED RING EI Internal Ringframe Sizing EI Sizing In process FEA– Verify Eigen Value FEA – Determine Hoop Load (For Panel) © 2010 Collier Research Corporation.

FEA: Verify Eigenvalue Optimum solutions determined from trend lines Plot Weight Trend for Panel & Ring Spacing Workspace Analysis New panel concepts introduced New design criteria imposed Seamlessly integrated into trade study Full Cylindrical FEM Input: Workspace sizing Interstage Geometry Number of Ring frames Loads and Boundary conditions ESTIMATED RING EI PRSEUS Reinforced Core Sandwich Internal Ringframe Sizing EI Sizing In process FEA– Verify Eigen Value FEA – Determine Hoop Load (For Panel) © 2010 Collier Research Corporation.

Honeycomb Sandwich Concept Weight Trend
7 Ring Frames 71 inch Spacing © 2010 Collier Research Corporation.

Integral Blade Sandwich Concept Weight Trend
4 Ring Frames 114 Inch Spacing © 2010 Collier Research Corporation.

Integral Blade Concept Weight Trend

Hat Stiffened Concept Weight Trend

All Panel Concepts Weight Trends

Weight Summary: All Ares V Composite Structures
Interstage Core Intertank Payload Shroud PRSEUS Unit Weight Ogive Barrel Honeycomb Sandwich Panel and Ringframe Total Smeared Unit Weight Reinforced Core Sandwich Hat Stiffened Panel All Loading Normalized to Max Axial Compression Internal Axial Load (Includes 1.4 load factor) © 2010 Collier Research Corporation.

Conclusions These weight studies show hat stiffened panels are the most efficient panel concept to carry the axial compressive load experienced in the Ares V Heavy Launch Vehicle. From the trends we see the hat stiffened panel concept is 20% lighter than the honeycomb sandwich concept. Even though ringframes are required for buckling stability of stiffened panels the weight savings in the hat panels overcome the added weight of the ringframes, and associated joints and fasteners. All panel weight trends provided were generated using HyperSizer software and the local and global buckling margins of safety were verified with FEA. Each panel concept has been rated using a weight maturity level scale. Weight trend plots presented, per panel concept, are useful for future cylindrical space launch designs. Though future vehicles may have different loads, diameters, and lengths, the general trends and weight % deltas will hold true for composite designs. © 2010 Collier Research Corporation.

Conclusions As the project continues HyperSizer can be used to further reduce the stiffened panel weight and optimize design details like bonded/bolted joints and minimize manufacturing complexity by reducing the laminate ply drops across component boundaries. Stiffened panels are more complex than sandwich panels and are more difficult to optimize because there are more optimization variables. this provides more customization and weight savings opportunities. In the short term as the design criteria changes each concept will adapt to new criteria. It is expected that the stiffened panel weight will decrease and the sandwich weight will remain constant. To effectively optimize stiffened panels to the many complex failure analyses an automated analysis tool like HyperSizer is needed to fully explore the design space. © 2010 Collier Research Corporation.

Ring Frame EI, Global Buckling Study
EI Specified (lb-in2) Resulting EA (lb) Beam Unit Weight (lb/ft) Panel Buckling EigV Buckling EigV > 2.15? Buckling Across Ringframe? 1.60E+08 1.72E+07 0.73 2.14 No Yes 1.70E+08 1.75E+07 0.74 2.148 1.75E+08 0.75 2.15 1.80E+08 1.77E+07 2.16 © 2010 Collier Research Corporation.

Ring Frame EI, Global Buckling Study

Interstage Joint Analysis

Longitudinal Construction Joints
Segmented joints considered for use with smaller autoclaves and fabrication rates Load sharing found to be negligible and weight impact was small Honeycomb Sandwich Longitudinal Joints Hat Stiffened Panel Longitudinal Joints © 2010 Collier Research Corporation.

Circumferential Connection Joints
Circumferential joints required to connect interstage to adjacent components in vertical stack © 2010 Collier Research Corporation.

Circumferential Connection Joints
Trade studies performed to evaluate impact of circumferential joints If modeled properly, Sandwich and stiffened panels should see little difference in end frame weight Lip Curls in due to Mx Caused by Eccentric Load Path Improper modeling Causes Rolling Static Deformation Improper modeling Causes localized buckling Proper modeling Drives Buckling Into Acreage © 2010 Collier Research Corporation.

Load Eccentricity from Improper Modeling
Improper modeling of stiffened panel concpets may cause load eccentricity Load eccentricity causes rolling and high moments near end frames High moments cause localized buckling Lip Curls in due to Mx Caused by Eccentric Load Path Improper modeling Causes Rolling Static Deformation Rolling moment causes localized buckling before ultimate load © 2010 Collier Research Corporation.

Current End Frame Modeling Technique
A. Correct Sandwich B. Incorrect Stiffened Panel Nodes at 198” radius Weight Penalty: approximately 1000 lbs Sandwich Panel Neutral Axis Stiffened Panel Neutral Axis © 2009 Collier Research Corporation.

Let’s flip flop the joint modeling
A. Sandwich B. Stiffened Panel In summary. The sandwich is getting the benefit of a joint design that best suits it, while the hat is getting penalized with the worse possible joint design for it. In retrospect this is just flip flop of what joints are seen commonly for sandwich that have ramp downs to one solid laminate. Nodes at 198” radius Weight Penalty approximately 500 lbs Additional Weight Savings approximately 1000 lbs Sandwich Panel Neutral Axis Stiffened Panel Neutral Axis © 2009 Collier Research Corporation.

Bonded Skin to Flange Joints
Both the peel and interlaminar stresses in the laminates increase dramatically near the flange end HyperSizer computes these stress variations © 2010 Collier Research Corporation.

Bolted Joint Analysis - BJSFM
BJSFM is used to predict the stress field surrounding a loaded hole. Stress Field Prediction Hole Loaded in Bearing and Bypass Peak Tension Stress Load direction Hole Elongation This analysis method requires material-dependent “characteristic dimension” values defined. BJSFM measures out a Characteristic Distance from the loaded hole and determines the stress in 4 directions 0, +45, 90, -45 at 5 degree increments, 360 degrees around the loaded hole. Then performs a lamina material strength analysis at every location. Peak Compressive Stress Performs Lamina analysis in 4 directions Characteristic Distance D0 © 2009 Collier Research Corporation.

Define Characteristic Distances for Orthotropic Material, Calibrate based on laminate geometry
The characteristic dimension is a fundamental data entry passed to the BJSFM analysis routine. For composites, the characteristic dimension is the proper distance away from the hole edge to apply stress/strain failure criteria and to calibrate to test data. The bottom figure is from MIL-HDBK-17-3E, fig (b)). Strain distributions near an open hole for different hole diameters and tape laminates. Applied far field load is equivalent to the expected failure load. Laminates are given as percentages of 0°/±45/90°. Crossing point of curves defines characteristic distance (d, or as noted in above figure, Rc) .016” © 2009 Collier Research Corporation.

Hat Lighter than Honeycomb?

71” RF Spacing Honeycomb Sandwich
Honeycomb Sandwich Panel – Optimum Dimensions © 2010 Collier Research Corporation.

57” RF Spacing Hat Stiffened Panel
Hat Stiffened Panel – Optimum Dimensions Hat is 21% lighter than Honeycomb Sandwich © 2010 Collier Research Corporation.

Axial Load Distributed in Panel Objects
1 1 2 3 4 2 3 5 1 2 1 3 4 3 5 Composite Weight = 1.03 lb/ft2 Core + Adhesive = lb/ft2 Composite Weight = 1.41 lb/ft2 All of the stiffened panel does carry load. The honeycomb core does not carry any load. Therefore 0.58psf (40%) of material is not supporting any load. © 2009 Collier Research Corporation.

Comparison of Load Carrying Panel Weight
Minimum Laminate Weight for Strength Hat 43.86“ Spacing Hat 57" Spacing Sandwich ” Spacing (36% Core Wt) Sandwich 57“ Spacing (40% Core Wt) Sandwich 142” Spacing (45% Core Wt) © 2009 Collier Research Corporation.

How Each Concept Prevents Panel Buckling
Buckling Lengths Unit Weight More difficult to size with many sizing variables More engineering More weight savings opportunities Easy to size with fewer sizing variables Less engineering Fewer weight savings opportunities Grow the Stiffener Increase Core Depth 71” 57” To Prevent Panel Buckling 43” Ny 0.93 psf Nx High Biaxial Compression © 2009 Collier Research Corporation.

Bending Stiffness Comparison
43.86 Interstage Honeycomb Sandwich 43.86 Interstage Hat 2.43" 1.78" > , , EI1 EI1 , , << , EI2 EI2 , < , GJ GJ , Note: Hat Stiffeners are ‘Closed’ sections So D33,hat >> D33,Tee, D33,I, D33,Z etc. Therefore hat panels are better for panel buckling than open cross-section panels © 2009 Collier Research Corporation.

Panel Buckling Approximation
43.86 Interstage Honeycomb Sandwich 43.86 Interstage Hat Interstage Panel bays have long width to height ratio a << b Shortest path is in the D11 direction, Therefore D11 is more effective than D22 in preventing panel buckling Stiffeners Stiffeners acting like columns or posts under axial compression load © 2009 Collier Research Corporation.

Panel Buckling Approximation
43.86 Interstage Honeycomb Sandwich 43.86 Interstage Hat For talking purposes, consider equation for flat, biaxial symmetric panel buckling AA Craig, mention that for stiffened panels, flat Eigv = cylindrical eigenvalue i.e. Raleigh Ritz gives same panel buckling result as flat panel buckling solution STIFFENED PANELS DO NOT BENEFIT FROM PANEL CURVATURE AS DO SANDWICH PANELS For hat panels, where D11 >> D22, this term is negligible and falls out of the equation, therefore panel buckling becomes a function of D11 only When “a” is much smaller than “b” (i.e. short ringframe spacing), the term dependent on D22 drops form the equation, and the hat panel becomes stable with only D11 and D22 dependence is not that important © 2009 Collier Research Corporation.

Two Hat Panel Designs: Local Buckling at Limit and Local Buckling at Ultimate Loads
RF spacing = 43.86” Local buckling at limit load hat panel = 1.38 psf (blue curve) Local buckling at ultimate load hat panel = 1.41 psf (green curve) Both designs have post buckling strengths > 8000 lb/in © 2009 Collier Research Corporation.

Interstage Hat: 1p474 UW, 57in RFs, +45/90/-45 OML (wide crown)

Post Buckling Model Summary
4 Bays X Symmetry on axial edges (new approach) Y Symmetry on transverse edges Tz = 0 on Bay Edge Facesheets (traditional approach) 2 stiffeners, flat panel Imperfection 0.01” Approximately 10% FS thickness Mesh Refinement: MS = 2, EAR=1 © 2010 Collier Research Corporation.

Non-Linear Buckling Results
Buckling Event Unit Load (% Ult Line Load) Load Factor Local (Crown/Web) 0.72 – 0.76 1.01 – 1.06 Panel 0.83 – 0.84 1.17 – 1.18 © 2010 Collier Research Corporation.

Panel Buckling – λ = 1.14 SE1 = -3100 μ

Panel Buckling – λ = 1.22 Panel Buckling – The local buckling of the web and crown led to a significant reduction in bending stiffness which may have lead to HyperSizer being over conservative. © 2010 Collier Research Corporation.

Interstage Hat: 1p41 UW, 43.86in RFs, +45/-45 OML

Verification: Panel Buckling

FEA#1: Full Barrel FEA with Ringframes
UW = 1.41 psf. RF spacing = 43.86”… LC103 Nx=-4561, Ny=-495 Full Barrel FEA Eigv = 2.19 Half wavelength ~ 31” FEA#1 HyperSizer Generated Equivalent Properties used in the FEM Half Wavelength ~ 31 inches FEA#1 HyperSizer Buckling Eigv = 2.20 Half wavelength ~ 31” Many full barrel FEMs were constructed. Shown here is one that has been meshed so that it can capture the mode shape with at least five elements per half wave. The FEA solution uses the HyperSizer generated PSHELL and MAT2 NASTRAN data types for representing a stiffened panel with an equivalent 2D FEM mesh. For this verification, both the eigenvalue and half wave length match between FEA and HyperSizer. *** Five different FEA solutions were performed with four different unique FEM constructions to verify the panel buckling analysis. This slide and the next two show those FEA results ***

Equivalent Mesh Panel Buckling Modes
UW = 1.41 psf. RF spacing = 43.86”… LC103 Nx=-4561, Ny=-495 HyperSizer gets five half wave lengths and has Eigv = 2.20 HyperSizer Generated Equivalent Properties used in the FEM FEA#2 FEA#3 This FEM is one panel bay long. The width used in HyperSizer was 1/8 circumfernce (155.5”). The same width was used in the FEM. The left is a NASTRAN eigenvalue buckling solution, and on the right is Abaqus. Both FEA solutions show the same mode shape as HyperSizer and nearly the same buckling load. In this FEM, the stiffened panel is represented with HyperSizer generated [A,B,D] matrix. The stiffeners are not discretely meshed. Abaqus (Eigv 1) gets five wave lengths and has Eigv = 2.163 Nx (Eigv 2) gets five wave lengths and has Eigv = 2.160 Nx (Eigv 1) gets six wave lengths and has Eigv = 2.158

Discrete Mesh Panel Buckling Modes
UW = 1.41 psf. RF spacing = 43.86”… LC103 Nx=-4561, Ny=-495 Discrete 4 Bay Eigv = 2.15 FEA#5: HyperSizer Independent Properties PCOMP and MAT8 used do discretely mesh the stiffener shape. FEA#4: Two additional FEMs were made that discretely meshed the hat shaped stiffeners. Because of the high element count, the width was made smaller than for the FEA#2 AND 3. For these discretely meshed models, buckling is dominated by local modes – to get global modes, mesh was coarsened substantially, and the range of eigenvalues was restricted to 2.0 – 2.5 FEA#4 did not have proper ringframe boundary conditions, so a larger FEM was made (FEA#5) to include three ringframes and panel bays. Though not visible, the stiffeners were also discretely meshed. The sum total of these FEA solutions with different modeling approaches prove out that the HyperSizer hat stiffened panel design can support 2.15 limit load without cylindrical panel buckling. Discrete 1 Bay

Verification: Local Buckling

Hat local buckling designed for Ultimate Load
UW = 1.41 psf. RF spacing = 43.86” The controlling load set for local buckling is 101 = (-4,561 lb/in Axial, 5.0 PSI internal pressure causing Ny = 990 lb/in). The first span object to local buckle is the web. For the skin load set 103 that has Ny = -495 controls. Using the MS from the failure tab, the backed out local buckling load is -6413, which is limit load. The is the eigenvalue when limit loads are applied and is used as the comparison to the FEA eigenvalues on the following slides. HyperSizer Buckling Eigenvalue Nx = = (- 4561*1.4056)

Hat local buckling designed for Ultimate Load: FEA Verification
UW = 1.41 psf. RF spacing = 43.86” The first attempt to predict local buckling with FEA was with this FEM. Because the half wave length is short, there were not enough elements (< 5 elements) to capture this mode. The next slide shows another more refined FEM to capture these modes. Local Buckling of Panel Laminates HyperSizer Buckling Eigenvalue = 1.41 NEi/Nastran FEA Buckling Eigenvalue = 1.73 (mesh is too coarse to capture lowest mode) Abaqus FEA Buckling Eigenvalue = 1.60 (mesh is too coarse to capture lowest mode)

Hat local buckling designed for Ultimate Load: FEA Verification
UW = 1.41 psf. RF spacing = 43.86” HyperSizer Eigenvalue = 1.41 NEi/Nastran Eigenvalue = 1.53 This slide shows a shorter span width model that is more refined to capture the local modes. Note that different FEA solvers (Nei Nastran and Abaqus) produce different eigenvalues. More importantly, for this problem the non-linear solution that has been perturbed shows a lower bifurication than does the linear FEA results. The geometric non-linear solution better matches the HyperSizer prediction. More models and load conditions need to be studied to quantify more cases. Abaqus NL bifurcation = 1.43 Abaqus Geometric non-linear deformation Abaqus Buckling Eigenvalue = 1.48

Verification: Crippling

Hat Crippling at Ultimate
UW = 1.41 psf. RF spacing = 43.86” The controlling load set for crippling is 101 = (-4,561 lb/in Axial, 5.0 PSI internal pressure causing Ny = 990 lb/in). Using the MS from the failure tab, the backed out crippling load is and the resulting strain is micro in/in. Note that to get the strain value, all other load sets had to be toggled off. These values are plotted on the next graph. HyperSizer crippling Nx = = (- 4561*1.4* ) HyperSizer crippling Strain x = = (- 4087* ) only turn on LS 101, the controlling load

Crippling Strength is above Ultimate Load
Abaqus geometric non-linear post buckling collapse load >> -4561*1.4 (-) Crippling lb/in lb/in Crippling 6600 lb/in Comparison of FEA load-strain response vs. HyperSizer crippling predictions. The blue curve is the load-strain response predicted by non-linear Abaqus FEA for a hat design that is not allowed to local buckle until after limit load is reached, but is allowed to local buckle before ultimate loads. The green curve is for a hat design that is not allowed to local buckle until ultimate load. Note that both designs are predicted with the non-linear FEA to have nearly the same collapse load which is well after ultimate loads. Also note that HyperSizer is predicting nearly the same empirical based crippling strength for both. The green markers represent HyperSizer local buckling prediction of the different panel spans. The blue markers are the HyperSizer prediction of crippling strength. Typically, crippling strength is predicted to happen well after the first local buckling event. In this case, because damage tolerance allowables are used where the strength of the laminate (ply strain allowable) is reduced, but the material’s modulus (stiffness) is not, presents this scenario where the local buckling load is nearly the same as the crippling load. By definition, crippling is a post local buckling event that becomes a material strength issue. For metallics the customary strain allowable, or in this case stress allowable, is compressive yield (Fcy). For composites, the appropriate value to use for strain crippling cutoff is not as well understood nor defined by the aerospace community. Currently a compressive strain allowable approximately = to (120F) micro in/in is being used. If a higher compressive strain allowable were used, then the HyperSizer crippling failure load prediction would be higher as well, while the local buckling load prediction would remain the same. HyperSizer like all other aerospace companies uses a crippling analysis based on empirical log-log curves. Specifically HyperSizer implements the log-log curves from Mil Hndbk 17. For this design, these curves inherently predict local buckling to occur before reaching ultimate loads, approximately around 1.2 limit load. Though not as important as the damage tolerance strain cutoff, this is another reason why the crippling analytical method indicates collapse failure before the FEA results. On the next contract effort, HyperSizer’s local post buckling method will be used as a higher fidelity method to predict this failure.

Verification: Scissor Buckling

Hat Stiffener Buckling at Ultimate
UW = 1.41 psf. RF spacing = 43.86” The controlling load set for hat stiffener ‘scissor’ buckling is 103 = (-4,561 lb/in Axial, PSI external crush pressure causing Ny = -495 lb/in). Using the MS from the failure tab, the backed out crippling load is which is limit load. The is the eigenvalue when limit loads are applied and is used as the comparison to the FEA eigenvalues on the following slides. HyperSizer Stiffener ‘scissor’ buckling Nx = = (- 4561*1.406)

Hat at Ultimate: Nastran FEA Scissor Verification
UW = 1.41 psf. RF spacing = 43.86” Hat_RF43.86_UW1p41_LB_at_Ult_Scissor and LTB_Eigv1p41.nas Flat Panel, Multi bay FEA Eigenvalue = 1.42 (mode 1) HyperSizer Eigenvalue = 1.41 Curved Panel (r=198”), Multi bay FEA Eigenvalue = 1.47 (mode 1) [3.5% increase] Two FEMs were constructed and executed to analyze this specific failure mode. The first was flat. The HyperSizer result and FEA match quite well. The second FEM had curvature. The curvature increased the eigenvalue and this FEM is used in all following slides.

Hat at Ultimate: Nastran FEA Scissor Verification
UW = 1.41 psf. RF spacing = 43.86” Hat_RF43.86_UW1p41_LB_at_Ult_Scissor and LTB_Eigv1p41.nas The Scissor failure mode caused by biaxial compression (-Ny due to -2.5 psi crush pressure) These images show different views of this buckling mode as predicted with NASTRAN FEA. Panel bottom Panel top

Hat at Ultimate: Abaqus FEA LTB Verification
Lateral Torsional Eigenvalue mode 1 = 1.486 These images show different views of this buckling mode as predicted with Abaqus linear FEA.

Hat at Ultimate: Abaqus FEA ‘Scissor’ Verification
Scissor Eigenvalue mode 2 = 1.495 These images show different views of this buckling mode as predicted with Abaqus linear FEA.

Hat at Ultimate: Abaqus FEA Verification
Eigenvalue mode 3 = 1.56 These images show different views of what appears to be a lateral torsion stiffener buckling mode as predicted with Abaqus linear FEA. For this mode, HyperSizer predicts an eigenvalue = 1.55, which is close to the FEA value of 1.56.

Flexural Torsional Buckling
Verification: Flexural Torsional Buckling

Torsional Buckling: I- Stiffened Panel
This slide is an AVI movie showing the progression of the post buckling as predicted with Abaqus non-linear FEA. The line plot shows the load-strain response to 8000 lb/in when the solution was stopped. Torsional stability and is applicable to general open cross-sections including closed sections and stiffened rods.

Flexural Torsional Buckling: L- Stiffened Panel
This slide is an AVI movie showing the progression of the post buckling as predicted with Abaqus non-linear FEA. The line plot shows the load-strain response to 8000 lb/in when the solution was stopped. Flexural-torsional buckling refers to a coupled buckling mode observed for columns where bending of the section is accompanied by a twist Unsymmetic Stiffeners PRSEUS

Flexural Torsional Buckling: I- Stiffened Panel
This slide is an AVI movie showing the progression of the post buckling as predicted with Abaqus non-linear FEA. The line plot shows the load-strain response to 8000 lb/in when the solution was stopped.

Flexural Buckling: I- Stiffened Panel
This slide is an AVI movie showing the progression of the post buckling as predicted with Abaqus non-linear FEA. The line plot shows the load-strain response to 8000 lb/in when the solution was stopped.

Summary of FEA Verifications

Summary of Verificaitons
Full Interstage Cylindrical Buckling Panel Buckling Local Buckling Crippling Stiffener Buckling Material Strength Bonded Joint Description Buckling and Ringframe sizing of full interstage Buckling of panels between ringframes Local Buckling of web/skin/crown, etc. Entire cross-section collapse in post-local buckling Scissor Mode & Lateral/Torsional buckling Stress/Strain Comparison Interlaminar stresses between skin and stiffener Knockdown Factor 0.65 1.0 Load Type Ultimate Required Eigenvalue 2.15 1.4 Verification FEM(s) Solvers: NX/Nastran ABAQUS Full Barrel FEMs with ringframes modeled as either CBAR 1-D elements or discretely using shell elemts Single and multiple panel bay FEMs with the assumption of perfectly stiff ringframes. Both discretely meshed and smeared stiffener FEMs were used NEiNastran Local discrete FEMs with a single stiffener or two stiffeners were made to capture local failure modes in the skin and/or the stiffener web ABAQUS (Geometric Non-Linear) Post-buckling analysis of a single stiffener FEM from the onset of local facesheet or web buckling until total panel collapse Local discrete FEMs with up to four stiffeners for single or multiple panel bays to capture stiffener buckling including scissor mode and/or torsional buckling Local discrete FEMs with a single or multiple stiffeners match HyperSizer’s stress and strain distribution throughout the cross-section N/A No independent FEA performed as part of this effort, however many FEA verifications and validations to test are available in the HyperSizer documentation Margin-of-Safety achieved? YES HyperSizer Match Excellent © 2010 Collier Research Corporation.

Future Weight Saving Techniques

Component Breakup to Properly Capture Flight Loads
Since the flight loads decrease up the span of the barrel, the panels near the top of the Interstage may be lighter than the panels at the bottom © 2010 Collier Research Corporation.

Summary Hats are lighter
HyperSizer is needed to achieve lightest weight stiffened panel designs HyperSizer can be used in next phase of design © 2010 Collier Research Corporation.

Craig Collier James Ainsworth Phil Yarrington Ryan Lucking

Similar presentations

Presentation on theme: "Craig Collier James Ainsworth Phil Yarrington Ryan Lucking"— Presentation transcript:

Similar presentations

About project

Feedback

Log in

Auth with social network:

Craig Collier James Ainsworth Phil Yarrington Ryan Lucking

Similar presentations

Presentation on theme: "Craig Collier James Ainsworth Phil Yarrington Ryan Lucking"— Presentation transcript:

Similar presentations

About project

Feedback