Available online at www.sciencedirect.

com

ScienceDirect
Procedia Engineering 172 (2017) 823 – 830

Modern Building Materials, Structures and Techniques, MBMST 2016

An experimental buckling study of column-supported cylinder

Olgerts Ozolins*a, Kaspars Kalninsa
a
Riga Technical University, Institute of Materials and Structures, Riga, LV-1048, Latvia

Abstract

An accuracy of numerical prediction of buckling load of axially or locally loaded thin-walled circular shells was always been a
challenge for engineers. With the introduction of new materials and manufacturing methods, well known design rules can be too
conservative in the case of buckling performance. New advanced or more sophisticated analyses/design methods should be
considered to fully utilize buckling load capacity of modern structural design. The major factors which can affect numerical
prediction of buckling load are combination of initial geometrical imperfections and shell imperfection sensitivity. Shells having
high imperfection to wall thickness ratio a/t are most affected, in contrast thicker shells are less sensitive. The current study deals
with less sensitive shell, subjected to combined axial/bending loading introduced by three support posts which acts locally
introducing local skin buckling. Prediction accuracy obtained by employing detailed finite element modeling of real case boundary
conditions and interaction of involved components. Combination of shell elements with 3D solid elements and utilization of contact
algorithm for sliding spherical ball supports produce highly reliable buckling load experimental and numerical prediction.
© 2017 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
© 2016 The Authors. Published by Elsevier Ltd.
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-reviewunder
Peer-review under responsibility
responsibility oforganizing
of the the organizing committee
committee of MBMST
of MBMST 2016 2016.

Keywords: Buckling, composite cylindrical shells, numerical analyses.

Introduction

A typical local buckling of column supports one may observe in frame supported silos design or in launcher
structures, where booster loads form a specific compression load path to primary structure. These structures by the
scale and size is rarely tested and reported in literature with sufficient level of details thus designers are exposed to
over conservative designs as nor does do aerospace nor building design codes give a detailed estimate of such a load

*Corresponding author: Olgerts Ozolins

E-mail address: [email protected]

1877-7058 © 2017 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the organizing committee of MBMST 2016
doi:10.1016/j.proeng.2017.02.130
824 Olgerts Ozolins and Kaspars Kalnins / Procedia Engineering 172 (2017) 823 – 830

cases. An extensive review on shell buckling and particularly on the case covered by this paper is given by Teng [1].
There were conducted number of studies in the 90`s covering mainly steel silos structures. More recent study of column
supported steel cylindrical structure was analyzed numerically by Hotała et al [2]. Extensive studies on silos buckling
problems was covered by Rotter [3] and integrated in Eurocodes. Design of silos under various load conditions
considering shell buckling was assessed by Zhao [4]. As well as studies on eccentric discharge related buckling [5].
In general observations from referred studies are summarized in Eurocode 3 [6], but in relation to composite design
there are the only option of Eurocode 4 [7] on concrete structures. Composite structures, in this case composite silos,
should be designed according to specified Eurocode 3 [8] section on silos. Nevertheless Eurocode design practice
explicitly allows designer to address material, loading imperfections and non-linearity as well as manufacturing
process related issues as design criteria which properly treated may reduce the over conservative designs. Therefore
a study which address all those issues has been conducted and obtained numerical experimental validation show
capacity potential for higher load carrying capacity or lighter/less expensive designs for the future.

Nomenclature

ܲ Axial compressive load

ܲ௖௥ Critical buckling load for a perfect shell
LVDT Linear variable differential transformer
ܴ Radius
‫ݐ‬ Thickness
‫ܧ‬௝௧ Elastic modulus along the fiber / transverse direction, tension
‫ܧ‬௝஼ Elastic modulus along the fiber / transverse direction, compression
‫ܩ‬ଵଶ Shear modulus
‫ݒ‬ଵଶ Poisson`s ratio
S1T/C Strength along the fiber / transverse direction, tension and compression
SG Strain gauge

1. General description and manufacturing

A tested specimen representing scaled down model of the free standing silos/reservoir type structure, considering
three stand support, was manufactured of IM7/8552 pre-preg system. General lay-up considered for shell consisted of
18 plies [0°/45°/-45°/90°/0°/90°/-45°/45°/0°]sym for shell structure with additional 12 plies [60°/0°/-60°/-
60°/0°/60°]sym for reinforcement area under the load support structure. General geometry of shell shown on Fig.1.

Fig. 1. Geometry of the shell.

 Olgerts Ozolins and Kaspars Kalnins / Procedia Engineering 172 (2017) 823 – 830 825

Instead of standard curing practice for IM7/8552 system, out-of-autoclave curing employed, skipping autoclave
phase. Different stages of manufacturing shown on Fig. 2. Shell made of IM7/8552 was manufactured employing 9
stage intermediate debulking, with the purpose to eliminate slack in consecutive layers generated by entrapped air.
Three equally spaced reinforcement patches consisting of 12 additional layers were prepared and located on the shell
before final vacuum bagging and curing. Curing cycle for IM7/8552 system was carried out as specified by
manufacturer, except autoclave and longer hold time at 180°C, Fig.3. Cure cycle temperature readings are carried out
with two thermocouples, located on inside surface of the mandrel close to the top and in the middle, manufacturer cure
curve shown on graph as green curve.

Fig. 2. Shell manufacturing on different stages.

Fig. 3. Curing cycle for shell made of IM7/8552.

Reinforcement placed on the outside only, due to manufacturing aspects, such pre-preg co-curing and use of internal
mandrel. Height of the reinforcement patch was 200 mm. Lower edge of the shell was not reinforced and left free
during testing. Support structures were CNC machined from EN AW 6082 aluminum blocks to desired shape, formed
as concave and convex two piece structure matching shell geometry, attached to the shell structure on both outside
and inside with PU (polyurethane) two component adhesive and bolted by DIN 912 M6 10.8 class bolts, Fig. 4. Quality
control and thickness measurements were performed by employing ultrasound inspection equipment, as well as, initial
geometry scan was performed by employing industrial 3D measurement system EXAscan scanner. More detailed
imperfection scanning results are given by Kalnins et al [9].
826 Olgerts Ozolins and Kaspars Kalnins / Procedia Engineering 172 (2017) 823 – 830

Fig. 4. Shell structure with attached support blocks and casted upper boundary.

2. Experimental set-up

To obtain desirable test height for shell a custom-made support frame for testing on INSTRON 8002 test frame was
designed and produced. Support ball bearing joints consists of 40 mm bearing ball and two conical seats at each side
at each support point, see Fig 5. Figure 5, shows FE analyses of deformations for design load of 125 kN. Support
frame mounted on INSTRON 8002 test frame allowed to obtain extra space for data acquisition by hiding lower
grip/load cell assembly inside the cylindrical shell.

Fig. 5. Support frame for shell specimen, considers two out of three adjustable height seats to compensate non-parallel machine plates.

Shell was tested on servo hydraulic test frame, considering incremental displacement loading with rate of 1
mm/min. Shell was equipped with 9 strain gauges equally spaced along the circumference near the top loading edge
used to check uniform load distribution, Fig. 6. Additionally back to back strain gauges were located at predicted
buckle spots - 45 mm at the center above support reinforcement patch upper edge, Fig. 6, and used to detect buckling
of the shell. Shell were equipped with three LVDT located in-line with the supports, Fig. 7. Two adjustable support
columns allowed to align shell upper edge to the machine plate, to allow simple shimming of the interface.
 Olgerts Ozolins and Kaspars Kalnins / Procedia Engineering 172 (2017) 823 – 830 827

Fig. 6. Strain gauge arrangement for tested specimen shell.

Fig. 7. Test set-up.

3. Experimental results and numerical validation

Experimental and numerical load-shortening curves were compared on Fig. 8. Measured strain gauge readings are
shown on Fig. 9. Due to non-uniform support translation it was considered to involve non-uniform loading in the finite
element analyses. It should be noted that only minor improvement in buckling load prediction was achieved, see Table
1 whereas the complexity of numerical solution raise rapidly.
Numerical models
100

80
EXPERIM
ENT
Load, kN

60
FEM
40 perfect
model
20

0
0,00 0,50 1,00 1,50 2,00 2,50 3,00
Shortening, mm

Fig. 8. Load-shortening curve for experimental test and numerical analysis.

828 Olgerts Ozolins and Kaspars Kalnins / Procedia Engineering 172 (2017) 823 – 830

Table 1. Experimental vs. numerical predictions.

Load (kN) KDF
Experimental test 93.57 0.96
FEM perf 97.45 1
FEM perf + loading imp. 95.68 0.98

Fig.10 lower edge shows photogrammetry based buckling shape captured in comparison to the numerical analyses
obtained for finite element model containing loading asymmetry. It should be noted that photogrammetry approach
can produce some scattered results due to presence of column support structure in point cloud caused by cleaning out
support block and SG wiring. In order to improve visualization of deformed state the reinforcement patch should be
cut out too from the point cloud in order to provide as uniform surface as possible (patch was considered as out-of-
plane deformation by 3-D processing software Autocad Recap 360 [10]). So photogrammetry based buckling shape
captures can be used only for buckling shape observation, not actual out-of-plane deflection measurements.

Fig. 9. Loading edge SG readings and back-to-back SG readings for tested shell.

Numerical simulations were carried out on commercially available ANSYS [11] finite element code. Additionally
numerical model was corrected for non-uniform load distribution by applying actually measured LVDT translations
to the corresponding supports, causing non-linear reaction force to develop associated to individual supports, Fig. 8.
Obtained buckling predictions was presented in Table 1, as well as corresponding knockdown factors. Numerical
analyses was carried out with non-linear (Newton-Raphson large deformation solver) analyses under displacement
control. Numerical model represents FEM composed of SHELL281 shell elements used for shell structure blended
with 3D solid elements SOLID186 employed for support structures made of aluminum/steel, Fig. 11.

Table 2. Material properties used for the shell.

Stiffness Strength
Mean value Std. Deviation [%] Mean value Std. Deviation [%]
T T
E1 171.5 GPa 2.6 S1 2300 MPa 13.8
E1C 150.2 GPa 4.6 S1C 857 MPa 10.1
T T
E2 8.9 GPa 4.2 S2 40 MPa 20.4
E2C 9.4 GPa 10.9 S2C 203 MPa 3.9
G12 5.1 GPa 7.8 S12 51 MPa 8.4
v12 0.32 13 tply 0.1308 mm
 Olgerts Ozolins and Kaspars Kalnins / Procedia Engineering 172 (2017) 823 – 830 829

Nodal plane of shell elements was shifted to cylinder inner surface to represent actual reinforcement patch and to
connect with solid elements, Fig. 11. Connection between shell and solid elements was realized as merging on
coincident nodes. Table 2, summarizes material properties used for finite element model. Compression material
properties were used for finite element analyses which even though produced in out of autoclave method [12] correlate
with material properties provided in product data sheet [13,14].

Fig. 10. Buckling mode shape for shell – numerical versus photogrammetry measurements.

Three supports are realized as 3D solid modelled support block with spherical joint, which incorporate surface to
surface sliding contact between concave conical support surface and support semi-sphere surface, with the reason to
reproduce actual boundary conditions as precisely as possible.

Fig. 11. 3D solid support block and half of the spherical ball.

For model simplification both support block and semi-sphere were considered represented of steel, since this
elements are deformation free. Bottom surface of the spherical support (semi-sphere) is clamped
(UX=UY=ROTX=ROTY=ROTZ=0), except loading direction UZ, via the coupled nodes for UZ. Bottom surface of
830 Olgerts Ozolins and Kaspars Kalnins / Procedia Engineering 172 (2017) 823 – 830

the spherical support (semi-sphere) is clamped (UX=UY=ROTX=ROTY=ROTZ=0), except loading direction UZ, via
the coupled nodes for UZ. Upper edge of the shell considered clamped (UX=UY=UZ=ROTX=ROTY=ROTZ=0).
Coupled set UZ applied for the loading edge nodes for load extraction from single pilot node reaction force. Bottom
edge of the shell was left free. Loading was considered as displacement controlled and applied equally for all three
semi-sphere pilot nodes. This was done with the purpose of capacity to apply non-uniform loading in accordance with
LVDT displacement measurements during experimental test campaign. Buckling load was extracted as a reaction force
(FZ) for the upper edge coupled set master node.

Conclusions

Even though the complexity of column support introduced buckling numerical and experimental tests are high the
current research highlight that very good agreement between experimental and numerical prediction has been achieved.
This involved complex finite element model implementation which took into account the sensitivity to precise load
application off-set eccentricity, which was not achieved during the preliminary studies employing simplified point
support approach. It was shown that structural robustness is relatively high as application of the loading imperfection
didn`t cause structure to change the load/displacement pattern. It was visually shown that photogrammetry may be a
powerful tool for buckled mode shape extraction even though the capture could be partly blocked by adjacent structure.

References

[1] J. Teng, Buckling of Thin Shells: Recent Advances and Trends. ASME. Appl. Mech. Rev. 49(4) (1996) 263-274.
[2] E. Hotała, Ł Skotny, Journal of Constructional Steel Research 96 (2014) 81–94
[3] Bulk Solids Handling: Equipment Selection and Operation Edited by Don McGlinchey.
[4] Y.Zhao et al., Thin-Walled Structures73 (2013) 337–349
[5] Adam J. Sadowski and J. Michael Rotter, A study of buckling in steel silos under eccentric discharge flows of stored solids, ASCE Journal of
Engineering Mechanics 136(6) 769-776.
[6] Eurocode 3: Design of steel structures - Part 1-6: Strength and stability of shell structures (2007). Comite Europeen de Normalisation (CEN).
[7] Eurocode 4: Design of composite steel and concrete structures (2004). Comite Europeen de Normalisation (CEN).
[8] Eurocode 3: Design of steel structures - Part 4-1: Silos (2007). Comite Europeen de Normalisation (CEN).
[9] K. Kalnins, O. Ozoliņš, M.A. Arbelo, R. Degenhardt, S.G.P. Castro, Verification study on buckling behaviour of composite cylinder with
eccentric supports, 54th Israel Annual Conference on Aerospace Sciences 2 (2014) 1216-1221.
[10] Autodesk® ReCap 360™
[11] ANSYS® Academic Research, Release 16.2.
[12] M.A. Arbelo, K. Kalnins, O. Ozolins, E. Skukis, S.G.P. Castro, Degenhardt, R. Experimental and numerical estimation of buckling load on
unstiffened cylindrical shells using a vibration correlation technique, Thin-Walled Structures 94 (2015) 273-279.
[13] E. Clarkson, A comparison of Equivalence Criteria and Basis Values for HEXCEL 8552 IM7 Unidirectional Tape computed from the NCAMP
shared database, NCP-RP-2013-015 N/C.
[14] G. Jacobsen, Mechanical characterization of stretch broken carbon fiber materials – IM7 fiber in 8552 resin. HEXCEL corp.