You are on page 1of 12

THE USE OF HIGH-FIDELITY ANALYSIS FOR RELIABLE BUCKLING LOAD CALCULATIONS

THE USE OF HIGH-FIDELITY ANALYSIS FOR RELIABLE BUCKLING LOAD CALCULATIONS
Jan Hol1 and Johann Arbocz2 Delft University of Technology, The Netherlands SUMMARY It is shown that using the measured midsurface and boundary imperfections, the so-called manufacturing signature of a fabrication process, it is possible to derive a safe, verified knockdown factor by applying a hierarchical high fidelity analysis approach (Ref [22]).Thus the analysis and design phase will be completed faster and only the reliability of the final configuration needs to be verified by structural testing. 1: Introduction With the arrival of the era of supercomputing there is a tendency to replace the relatively expensive experimental investigations by numerical simulation. The use of large general purpose computer codes for the analysis of different types of aerospace, marine, and civil engineering structures is by now well accepted. These programs have been used successfully to calculate the stress and deformation patterns of very complicated structural configurations with the accuracy demanded in engineering analysis. However, there exist numerous complex physical phenomena where only a combined experimental, analytical, and numerical procedure can lead to an acceptable solution. One such problem is the prediction of the behaviour of buckling sensitive structures under the different loading conditions that can occur in everyday usage. The axially compressed cylindrical shell represents one of the best known examples of the very complicated stability behaviour that can occur with thin-walled structures. For thin shells that buckle elastically initial geometric imperfections [1,2] and the effect of the different boundary conditions [3-5] have been identified as the main cause for the wide scatter of the experimental results. However, this knowledge had not been, as yet, incorporated into the current shell design manuals.

1

Assistant Professor, Aerospace Structures, Faculty of Aerospace Engineering

2

Professor Emeritus, Aerospace Structures, Faculty of Aerospace Engineering, Fellow AIAA

ν = Poisson’s ratio.S. = factor of safety. The empirical knockdown factor γ is so chosen that when it is multiplied with Pcrit . a lower bound to all available experimental data is obtained. γ = knock-down factor. Using this approach for isotropic shells under axial compression. and F. h = shell wall-thickness. The improvements with respect to the presently recommended shell design procedures are primarily sought in a more selective approach by the definition of the knockdown factor γ . c= 3(1 − ν 2 ) and R = shell radius. E = Young’s modulus. in 1965 Weingarten et al [9] have published the following expression for a lower bound curve 1 R − P γ= = 1 − 0.THE USE OF HIGH-FIDELITY ANALYSIS FOR RELIABLE BUCKLING LOAD CALCULATIONS These design recommendations [6-8] all adhere to the so-called lower bound design philosophy and as such recommend the use of the following buckling formula: γ Pcrit F.901 (1 − e 16 h ) Pcl (2) where Pcl = classical critical load = 2πR Ncl (3) Eh cR 2 Nc l = . the lowest buckling load of the perfect structure.S. Pa ≤ (1) where Pa = allowable applied load. The proposed new improved shell design procedure can be represented by the following formula λa Pc F. Pa ≤ (4) . Pcrit = lowest buckling load of the perfect structure.S. The central goal of the research reported in this paper is the development of improved shell design criteria.

calculate its critical buckling load quite accurately. which involve an extension of the solution into the postbuckling response region. and F. if one carries out imperfection sensitivity investigations.18] formulation. In both cases. is provided by DISDECO. as has been pointed out in the past by Byskov [10]. With this open ended. and make a reliable prediction of the expected degree of imperfection sensitivity of the critical buckling load. Unfortunately. Pc = lowest buckling load of the perfect structure computed via one of the shell codes. used for an accurate prediction of the critical buckling loads and a reliable estimation of their imperfection sensitivity. This approach allows different aspects of the nonlinear response to be represented properly as these aspects begin to affect the response. the Delft Interactive Shell DEsign COde [12]. Furthermore. The Level-3 solutions used in this paper are based either on a two-dimensional finite difference [16] or finite element [17. The steps involved in the derivation of such a verified high-fidelity (higher) knockdown factor λa are the subject of this paper. The proposed procedure consists of a hierarchical approach. the accuracy of the discrete model used should be checked against available analytical or semi-analytical results or finite element models with appropriate mesh convergence studies to assure that the analysis using the discrete model converges to a meaningful result. further mesh refinement may be needed since the wavelength of the dominant large deformation pattern may often decrease significantly. A test series of seven isotropic shells carried out by Arbocz and Babcock [11] at Caltech is used to illustrate how a hierarchical high fidelity analysis can be carried out. Each likely response mode must be properly represented in the discrete model. if one uses the appropriate . hierarchical interactive computer code the user can access from his workstation a succession of programs of increasing complexity that are suitable for various nonlinear aspects of the response.THE USE OF HIGH-FIDELITY ANALYSIS FOR RELIABLE BUCKLING LOAD CALCULATIONS where Pa = allowable applied load. In an earlier paper [13] the buckling predictions obtained using Level-1 methods with membrane prebuckling analysis [3-5] and Level-2 methods with rigorous nonlinear prebuckling analysis [14. where the analyst proceeds step-by-step from the simpler (Level-1) methods used by the early investigators to the more sophisticated analytical and numerical (Level-2 and Level-3) methods used presently. Thus the shell designer can study the buckling behavior of a specified shell.S. = factor of safety. λa = verified high fidelity (higher) knockdown factor. This step is part of a mandatory study needed in order to establish the dependence of the buckling load predictions on the mesh distribution used and to assure that the analytical solution is consistent with the physics of the problem being solved. 2: Mesh Convergence Study At the beginning of any stability investigation. The platform for the multi-level computations.15] have been reported. some of the critical shorter wave-length modes associated with the nonlinear response of the structure may not be activated until well into the load-response history for the structure.

as can be seen from Figure 1. In the convergence study.THE USE OF HIGH-FIDELITY ANALYSIS FOR RELIABLE BUCKLING LOAD CALCULATIONS meshes. whereby the earlier results obtained with the Level-2 DISDECO module ANILISA [19] listed in Reference 13 serve as a reference. As can be seen from Fig. Thus. the results converge to a limiting value from below at about NC = 261. for a fixed number of mesh points in the axial direction (NR = 161 for this example) the number of mesh points in the circumferential direction (NC) was increased until the bifurcation load approached a horizontal tangent. for a fixed number of mesh points in the circumferential direction (NC = 161). 1. This time the convergence is from above. For this purpose. the number of rows (NR) was varied. at first. and the horizontal tangent is reached at about NR = 261. Figure 2: Maximum normal displacement versus axial compression curves using refined meshes . one can obtain rigorous solutions where all nonlinear effects are properly accounted for. initially a convergence study must be carried out in order to establish the mesh size needed to model accurately the response and buckling behavior of the shell in question. Next. the asymmetric bifurcation responses from a nonlinear prebuckling path solution option was used. Figure 1: Figure 1STAGS-A [10] convergence study using Caltech isotropic shell A-8 [11].

for the first time.THE USE OF HIGH-FIDELITY ANALYSIS FOR RELIABLE BUCKLING LOAD CALCULATIONS To illustrate the difference between using coarser meshes to speed up the computations and finer meshes which produce (nearly) converged solution. Figure 3: Measured initial imperfections of Caltech isotropic shell A-8 [11]. the axial compression versus maximum normal displacement curves for the isotropic shell A-8 with a short-wave asymmetric imperfection [13] are displayed in Figure 2. As can be seen from Figure 3. the measured initial midsurface imperfections of shell A-8 [11] show a rather general distribution dominated by an n = 2 mode. Thus to speed up the computations in the following the coarser 161x161 mesh will be used. One can use the following half-wave cosine double Fourier series 7 99 W x 18 y 19 y x y ′ l sin l + ∑ ∑ Wkl cos k π cos l = ∑ Wio cos i π + ∑ Wol cos l + ∑ Wo h L R R L R i =1 l =1 l =1 k.l x y ′ l cos k π sin l + ∑ ∑ Wk L R k. v = w = Mx = 0) the results do not deviate significantly from the results of a mesh with 261 rows and 261 columns (a model with 207508 degrees of freedom and a maximum semi-bandwidth of 1037) and the same SS-3 boundary conditions. l 116 (5) . In 1969 Arbocz and Babcock [11] published the results of buckling experiments where. Using a mesh of 161 rows and 161 columns (a model with 79708 degrees of freedom and a maximum semi-bandwidth of 635) and SS-3 boundary conditions (NX = -N0. the actual initial imperfections and the prebuckling growth of the midsurface of electroplated isotropic shells were carefully measured and recorded by means of an automated scanning mechanism. one must know the type of imperfections that occur in practice. 3: Midsurface Initial Imperfections In order to apply the theory of imperfection sensitivity with confidence.

Figure 4: Measured flatness of the Caltech load-cell end-ring [11]. the 161x161 STAGS model with SS-3 boundary condition (NX = -N0. It is shown in References [20.1 lbs. The flatness of the end ring attached to the load-cell of the Caltech test set-up used to test shell A-8 was measured. with maximum deviations of about 3 wall thicknesses. whereas the same model with C-4 boundary condition (u = u0. The measured boundary imperfections are decomposed into a one-dimensional Fourier series 1 uo = ub (y) = ξbh{ a o + 2 n= 1 ∑ (a n cos n R + b n sinn R )} y y (6) .001 are neglected.21] that boundary imperfections can have a significant degrading effect on the buckling loads of axially compressed cylindrical shells. one must use the measured initial imperfections in codes like STAGS [16.x = 0) yielded a collapse load of Ps = −976.17] to carry out two-dimensional nonlinear collapse analysis. v = w = w. Employing a user-written subroutine WIMP to input the double Fourier series of equation (5). For a more accurate estimate of the buckling load.9 lbs.5 lbs. v = w = Mx = 0) yielded a collapse load of Ps = −900. 4 there is a very large amplitude wavy pattern in the plane of the end support. where Fourier coefficients with absolute values less than 0.THE USE OF HIGH-FIDELITY ANALYSIS FOR RELIABLE BUCKLING LOAD CALCULATIONS to represent the measured initial imperfections accurately. As can be seen in Fig. Both of these values are significantly higher than the experimental buckling load of Pexp = −825.

one obtains the simulated buckling loads Ps ( ξB = 0.x = 0 and by taking advantage of the dual loading systems provided by the STAGS program. the calculated collapse load of Ps = −748. 5. 4: Numerical Results It is shown in Ref. (7) Figure 5: End rings used in the Caltech test set-up for buckling tests [11]. v = w = w. it is to be expected that after cooling the hardened liquid metal has filled-in all of the gaps.04 . The effect of this unknown end support was modeled by varying the amplitude of the boundary imperfection ξb between 0 and 0.04) tabulated in Table 1. 6. The experimental buckling loads Pexp exceed the simulated buckling loads Ps in all cases. see Fig.1 as can be seen in Fig.THE USE OF HIGH-FIDELITY ANALYSIS FOR RELIABLE BUCKLING LOAD CALCULATIONS This series is then used in STAGS [16] to model the effect of boundary initial imperfections by using a modified C-4 boundary condition u = ub(y). looking at the results presented in Figure 6. one sees that if . Because the grooves of the end rings were filled with liquid Cerrolow at the time shell A-8 was installed in the test apparatus. [11] using the appropriate measured initial midsurface imperfections and the same measured boundary imperfections of Equation (6) with an amplitude of ξB = 0. However.6 lbs agrees closely with the experimental buckling load of shell A-8 of Pexp = −825.9 lbs.04 . Repeating the buckling load calculations for all 7 A-shells of Ref. except for the 2 shells that buckle initially in a stable local buckle. [22] that including both the measured initial midsurface imperfections of Equation (5) and the measured boundary imperfections of Equation (6) with an amplitude of ξb = 0.

6 -823.0 -876.6 -694.9 -1037.9 -825.5 -667.4 -685.7 -1014.2 -1009.) Figure 6: Combined effect of measured midsurface and boundary imperfections.1 -673. buckling loads in lbs.9 -718.5 -682. .2 -698.04) Plocal Pexp A-7 A-8 A-9 A-10 A-12 A-13 A-14 -880.1 -911.6 -905.2 -892.4 -850.5 -689.7 -866.4 -910.9 -805.3 -633. PSS −3 PC − 4 Ps ( ξB = 0.5 -926.0 -900.0 Table 1 Summary of Level-3 Buckling Load Calculations (161x161 mesh.058 .0 -1044.2 -976. then the simulated buckling loads yield a lower bound to all seven shells tested.9 -837.9 -774.9 -748.THE USE OF HIGH-FIDELITY ANALYSIS FOR RELIABLE BUCKLING LOAD CALCULATIONS one uses a boundary imperfection amplitude of ξB = 0.

since the analysis and design phase will be completed faster and only the reliability of the final configuration needs to be verified by structural testing.7 -866.0 Table 2 Comparison of simulated high-fidelity versus experimental buckling loads (buckling loads in lbs.058) P0.599 0. the so-called manufacturing signature of a fabrication process. .570 -546.8 -553. 5: Conclusions It is shown that using the measured midsurface and boundary imperfections.1 -585. It is believed that in the end the use of high fidelity numerical simulation will also lead to overall cost reduction.3 -216.3 -222.0 -543.4 -516.648 0.5 -255.608 0.058 are normalzed by the lowest buckling load of the corresponding perfect shell with C-4 (u = u0. it is possible to derive a safe.2 -626.9 -556. the simulated buckling loads using both midsurface and boundary imperfections with ξB = 0.1 -622. verified high fidelity knockdown factor λa by the numerical simulation approach described in Ref. the verified high fidelity knockdown factor of an imperfect shell.597 0.4 -526.8 -212. From the results listed in Table 2 it is clear that using λa = 0.2 -595.7 -677.0 -659.8 -207.2 -698.9 -837.4 -718.636 0.1 -256.x = 0) boundary conditions. [22] and in this paper.THE USE OF HIGH-FIDELITY ANALYSIS FOR RELIABLE BUCKLING LOAD CALCULATIONS To obtain λa .6 -825.6 one obtains indeed a safe allowable axial load for all seven shells tested. v = w = w.8 -6821.3 0.9 -774.058 PC − 4 λaPC − 4 γPC − 4 Pexp A-7 A-8 A-9 A-10 A-12 A-13 A-14 -545.0 -533. Ps ( ξB = 0.) It is interesting that the proposed high fidelity simulated allowable buckling loads are about twice as big as the allowable buckling loads of the traditional lower bound design criteria of NASA SP-8007.610 0.6 -237.

. REFERENCES 1. J. pp. Oslo. W. 1968. 1-86.. ANONYMOUS.” Proceedings of the 11th IUTAM Congress.” Proceedings of the IUTAM Symposium on Buckling of Structures. Budiansky. 227-250. This support is gratefully acknowledgement. pp. “Dynamic Buckling of Imperfection Sensitive Structures. 2. OHIRA. edited by B. 37-40. “Buckling of Thin Shells. V. Springer-Verlag. N. P. KOITER. “Beulsicherheitsnachweise für Schalen. 5. of aeronautical Sciences.J. SINGER J.. 1962. 1961.” DnV (Det norske Veritas). Amsterdam. Thesis (in Dutch). HOFF. 1980. H. Springer-Verlag. 8. Theodore von Kármán on his 80th Anniversary. “Buckling of Thin-Walled Circular Cylinders.. Science Council of Japan. 7. pp. 1977. 1965. DASt Richtlinie 013. ANONYMOUS. Tokyo.” NASA SP-8007.J. “Elastic Stability of ThinWalled Cylindrical and Conical Shells Under Axial Compression. “Rules for the Design. pp.. 1-833. 500-505.T. and ROSEN. 4. “Local Buckling Theory of Axially Compressed Cylinders. edited by H. English translation NASA TTF-10. Norway.” Deutscher Ausschuss für Stahlbau. and SEIDE. 1961. New York. Görtler. 1945.I.. H. pp. THDelft. Paris.W. . 3. The Netherlands. Japan National Committee for Theoretical and Applied Mechanics. and HUTCHINSON. Vol. Inst. 9. 6. “The Influence of Boundary Conditions on the Buckling of Stiffened Cylindrical Shells.” Ph. ANONYMOUS.D. 1967. May 11. MORGAN. A. “On the Stability of Elastic Equilibrium.THE USE OF HIGH-FIDELITY ANALYSIS FOR RELIABLE BUCKLING LOAD CALCULATIONS 6: Acknowledgement Part of the research reported in this paper has been carried out during the second author’s tenure as an NRC Research Associate at the NASA Langley Research Center. WEINGARTEN.J. 636-651.” Proceedings of the 11th Japan National Congress for Applied Mechanics.” Proceedings of an Aerospace Scientific Symposium of Distinguished Lecturers in Honor of Dr..” AIAA Journal. pp. Berlin. E. B. BUDIANSKY. Construction and Inspection of Offshore Structures. 1964. 3. Berlin. 1976.

111-119. Solids Structures. J. “The Effect of Imperfect Boundary Conditions on the Collapse Behavior of Anisotropic Shells. Air Force Flight Dynamics Laboratory. 14.O.M. 1969. Structural Dynamics and Materials Conference. Flugwissenschaften.. 1963. Vol. Norfolk.” Proceedings 44th AIAA/ASME/ASCE/AHS/ASC Structures. 19.. “Smooth Postbuckling Stresses by a Modified Finite Element Method. Solids and Structures. Vol. J.A. Karlson & Sorensen. pp. Lyngby.. “A Hierarchical High-Fidelity Analysis Procedure for Buckling Critical Structures. F. “The Effect of General Imperfections on the Buckling of Cylindrical Shells. J.f.” J.A.” DCAMM Report No. 12. 16. Vol. 11.T.” Int... G. pp. California. WrightPatterson Air Force Base. F. M. 26.” Z. JR.M. pp. ANONYMOUS. California... “Über den Einfluss der gelenkingen Lagerung auf die Stabilität dünnwandiger Kreiszylinderschalen unter Axiallast und Innendruck. J. 18. March 1973.. Structural Dynamics and Materials Conference. H. STAGS User Manual. February 1964. II. 1993. . Lockheed Palo Alto Research Laboratory..THE USE OF HIGH-FIDELITY ANALYSIS FOR RELIABLE BUCKLING LOAD CALCULATIONS 10. “Shell Stability Analysis in a Computer-Aided Engineering (CAE) Environment. June 1994.M. Palo Alto. pp. 7-10 April.0. J. (Paper AIAA2003-1844) . ARBOCZ.C. User's Manual for the STAGS-A Computer Code. Version 2. MILLER. J. Appl. J. BROGAN.D. 9/10. BROGAN.. and CABINESS. 6891-6915. 20.H. 300-31. "Koiter's Stability Theory in a Computer Aided Engineering (CAE) Environment. 13.” NASA TR-190. Rhode Island. 28-38. LMSC P032594. ZELE.. 11. and STARNES. "Collapse Analysis for Shells of General Shape... La Jolla. Pawtucket.. F. B. and BABCOCK. ARBOCZ.D. FISCHER. ARBOCZ. ABAQUS. Virginia. J. Ohio. RANKIN. 1988. Numbers 46-47. J. 36. 17." Int. 945-973. pp. BYSKOV.M. and HOL. April 19-22. Technical University of Denmark..." Technical Report AFFDL-TR-71-8. “The Influence of Prebuckling Deformations and Stresses on the Buckling of Perfect Cylinders. 27. H.A. and PETERSON.. E. Hibbitt. Vol. ARBOCZ. ARBOCZ. No. 1990. 380. Mech.” Proceedings 34th AIAA/ASME/ASCE/ AHS/ASC Structures.A. 2003. 15. C. ALMROTH. Inc. J. C. and HOL. 2000. E. STEIN.

J. and STARNES. (Paper AIAA-2002-1513) .THE USE OF HIGH-FIDELITY ANALYSIS FOR RELIABLE BUCKLING LOAD CALCULATIONS 21. Georgia. “Buckling Load Calculations of the Isotropic Shell A-8 Using a High-Fidelity Hierarchical Approach..H. Atlanta. ARBOCZ. 22-24 April 2002. “Effects of Imperfections on the Buckling Response of Compression-Loaded Composite Shells. HILBURGER. .” Proceedings 41st AIAA/ASME/ASCE/AHS/ASC Structures. Colorado. and STARNES. J. Structural Dynamics and Materials Conference. M.” Proceedings 43rd Structures. Denver. J. Structural Dynamics and Materials Conference.. 3-6 April 2000. 22.H.W. (Paper AIAA-2000-1382).