You are on page 1of 4

Challenges, Opportunities and Solutions in Structural Engineering

and Construction Ghafoori (ed.)


2010 Taylor & Francis Group, London, ISBN 978-0-415-56809-8

Finite element analysis of wind induced buckling of steel tank


S. Borgersen
SEBCorp, Eagan, MN, USA

S. Yazdani
Department of Civil Engineering, North Dakota State University, Fargo, ND, USA

ABSTRACT: During construction, large diameter steel plate storage tanks could experience severe damage
and buckling under environmental loading conditions, such as gusting wind loads, if inadequate bracing is
provided. One such case is presented in this paper involving a 2,760 cubic-meter steel plate tank that experienced
localized buckling under a severe straight line wind pressure with a velocity exceeding 100 km per hour. This
paper discusses the failure mode, simulation of the failure using FEA with wall material non-linearity, and
recommendations for design and construction methods which could be used to prevent similar occurrences on
other tank erection projects.
1

INTRODUCTION

Construction of a 2,760 cubic-meter ethanol storage


tank was nearly complete, requiring only the installation of a conical roof structure. The completed tank
(Figure 1) was to be a welded plate cylindrical shell
structure 15.24 m diameter and 15.24 m wall height,
constructed of rolled stainless steel plates 1.22 m
2.44 m sections with plate thickness varying with wall
elevation, as shown. The tank wall was of welded plate
construction, and supported on a reinforced concrete
ring wall. Stainless steel threaded anchor rods, 32 mm
diameter, were used to attach the tank to the reinforced concrete ring wall foundation by use of plate
gussets. The gussets were welded to the tank wall at
a circumferential spacing of 305 mm around the tank
base perimeter.
The interior tank floor was a 0.61 m inverted conical
surface, constructed with parallel strips of 1.22 m

Figure 1.

Schematic representation of the tank.

2.44 m 8 mm stainless steel plates, overlapping


25 mm and seam welded using a 5 mm fillet weld.
A pipe drain and reinforcing pad was located at the center or apex of the conical floor. The tank floor extended
approximately 38 mm past the base of the tank wall.
The bottom wall course consisting of 38 mm wall plate
was continuously welded to the 8 mm plate floor along
its outer perimeter.
A 2.7 m high conical plate roof, constructed of
5 mm welded stainless steel plate segments had been
assembled at ground level, next to the tank wall and
foundation. The roof structure was reinforced by eighteen, radially oriented, W10 17 beam members, on a
20 degrees circumferential spacing. The outer edge of
the roof assembly included a circumferential 13 mm
152 mm 203 mm compression angle, with the 8 leg
extended outward in the radial direction.
The entire roof assembly, including access hatches,
vents, peripheral equipment, and perimeter safety railing, was assembled on the ground next to the erected
tank wall.
The roof assembly was scheduled to be lifted into
position and attached to the top of the tank wall, using
a 9 kN capacity crane. As the crane operator prepared
to initiate the installation, a weather warning, citing
potentially very strong, straight-line winds, caused the
installation process to be halted. The roof structure and
crane boom were then securely anchored, as a safety
precaution.
Shortly thereafter, a series of severe, straight-line
wind gusts struck the construction site. Unfortunately,
the straight line wind gusts were localized, and peak
gust magnitudes and durations were not recorded by
the local airport weather station. However, the wind
gust direction was directly perpendicular to the tank

157

Figure 2.

Schematic representation of the damage zone.

wall, as evidenced by the location of the damaged


wall section and the presence of nearby surrounding
structures.
Security cameras allowed plant personnel to witness the event. Their statements indicated the wind
gusts were so severe that the tank wall underwent large
amplitude, dynamic deformations during the storm.
The wind loading continued to affect the tank structure, finally causing the cylindrical, open-end tank
wall to buckle along an elliptically shaped failure surface shown in Figure 2. Examination of the damaged
tank wall structure determined that the damage consisted of large, plastic deformation along the elliptical
wall segment covering approximately 30 around the
top circumference and extending approximately 7-1/2
courses down from the top of the tank wall.
The plates were so severely buckled that removal
and replacement of several plates were required. The
remaining plates were corrected in-place, using acceptable construction techniques. However, removal and
replacement of the damaged wall plate segments
created a situation where the new welds had to be
re-inspected using radiographic methods. Also, hydrostatic leak testing of the tank shell had to be reconducted to certify the structure after the repairs
were completed. Once the roof structure had been
installed, a second series of hydrostatic pressure tests
were required to ensure all safety valves, manhole
access hatches and welds were sufficient to meet API
code (American Petroleum Institute) requirements.
Added costs associated with the delay included:
added material costs, radiological inspection costs,
labor, cost of water required for the tests and delays
due to required code verification testing. However the
largest cost was downtime and loss of plant production
capability, due to inability to utilize the storage tank
as scheduled.

THEORY

Shell buckling is a complex, stability based event,


which takes into account geometry, loading and nonlinear material properties. There appears to be no

directly applicable closed form solutions for the large,


plastic deformation of cylindrical thin wall shells open
and unsupported at one end and cantilevered at the
other.
A forensic replication of the wind induced failure of
the tank wall would therefore require the use of a finite
element analysis (FEA) modeling technique using
software that would be able to account for the applicable parameters, including distributions of pressure
loading applied over portions of the structure, incrementally varying loading amplitudes, and non-linear
material characteristics. In order to meet these analysis
requirements, the MARC software provided by MSCSoftware Corporation was used. The MARC software
is one of a family of technical analysis packages
available from MSC Software Corporation, including:
Patran, Nastran, Adams, Dytran, and others.

PRESSURE DISTRIBUTIONS

Evaluation of AISC, API and IBC Codes with their


assumed design wind velocities, shape coefficients
and associated design pressures were found to be
not applicable to this type of construction geometry:
a cylindrical shell structure cantilevered at the base
and open at the top. The codes are design guidelines
intended for fully enclosed structures, or structures
with openings in one or more vertical surfaces.
It was therefore decided that initial wind pressure distributions on the structure type being investigated could best be evaluated by using the computational fluid dynamics (CFD) capabilities in the MARC
software.
Preliminary wind pressure distributions were
derived using 2-dimensional CFD models: the first
using a two-dimensional plan view (Figure 3), representing a rigid circular structure to evaluate initial
pressure distributions around the circumference of the
tank wall perimeter; the second represented a twodimensional, vertical elevation cut through the center
of the tank.
Results obtained from the two-dimensional plan
view wind model provided locations of peak positive,
zero and negative wind pressure distributions around
the tank perimeter. The two dimensional vertical elevation cut provided additional information with respect
to vertical pressure distributions up the tank wall, as
well as the effects of turbulence on pressure at the
open ends of the tank, upwind, downwind and internal
to the open end.
The results obtained from these two CFD studies were then implemented as pressure loadings in
the FEA model, and correlated with ASCE, API
and IBC code design wind pressure recommendations for closed ended cylindrical structures of this
geometric type.

158

Figure 3.

Study Case 4: The stepped wall tank model used in


CASE 3, but modified with strut stiffeners tying the
top of the center column to the top of the tank wall at
22.5 around the circumference.
Study Case 5: The stepped wall tank model used
in Study Case 2, but modified with a continuous
13 mm 152 mm 203 mm stiffener angle added
to the top edge of the tank with the 203 mm leg being
horizontal.
The FEA model used for this investigation incorporated 12,240 nodes and 12,000 elements. The elements
were a six-degree-of-freedom per node, type 75, thin
shells.
Analysis studies incorporating strut stiffeners and
a cantilevered center post used additional nodes and
element types, as needed. The axial only members
used a three-degree-of-freedom per node, type 9 truss
elements, and a three degree-of-freedom per node,
type 7, solid element was used for the center post.

Wind pressure distribution on the tank.

MATERIAL PROPERTIES

The material used in the model was assumed to be


stainless steel, isotropic, elastic-plastic with Youngs
Modulus E = 200 GPa; Poissons ratio = 0.3; yield
stress Fty = 206 MPa; ultimate stress Ftu = 310 MPa.
A non-linear stress strain model was used for the stainless steel material.

MODEL CORRELATIONS

To simulate the over all wind effect, an incrementally increasing static loading condition using the
wind pressure load was applied to the tank. The incremental loading was linearly increased until a stress
level in the tank wall was developed which provided
an assumed equivalent von Mises yield stress of
206 MPa, matching that of the observed damage on
the tank wall.

The incremental application of wind pressure on the


external surfaces of the tank wall correlating with the
CFD analysis resulted in the representative deformation of the wall structure summarized in Table 1. The
analysis shows that the wind pressure forms an elliptical yield surface extending from the top of the cylindrical shell towards the lower, thicker course of the tank
wall. The FEA model predicts that the primary yield
surface extends down to approximately 7-1/2 plate
courses, approximately 8 m to 9 m, from the top of
the tank wall. The plan view of the deformed surface
also indicates that the maximum deformed shape of
the wall structure occurs around the circumference of
the tank wall.

APPLIED LOADS

ANALYSIS MODELS

Five FEA model studies were evaluated to determine


the effects of various stiffening options which might
have been used to prevent this type of failure during
construction:
Study Case 1: A constant thickness tank wall model
using 13 mm plate thickness fixed at the base and cantilevered and open at the top edge was developed as a
baseline for comparison purposes.
Study Case 2: A stepped wall tank model reflecting
actual plate wall thickness at 2.4 m course heights, as
shown in Figure 1.
Study Case 3: The stepped wall tank model used
in CASE 2, but modified with a full height 457 mm
diameter pipe column cantilevered at the base, and
with strut stiffeners tying the top of the center column
to the top of the tank wall at 45 around the circumference.

159

ANALYSIS RESULTS

FEA analysis studies evaluated three potential damage prevention options that may have prevented the
tank wall buckling, had they been incorporated prior
to installing the roof structure. Study Cases 1) & 2)
are baseline studies using the constant 13 mm thickness tank wall and the actual, as constructed, stepped
thickness tank wall which failed due to wind load.
This produced a maximum calculated radial displacement Equivalent Stress of 2.1 meter, 214 MPa and
1.48 meter and 201 MPa respectively. Study Case 3)
provided radial stiffening of the tank top perimeter
wall, with the axial stiffeners located at 45 around
the perimeter. This produced a maximum inward radial
deflection of the wall with magnitude of 1.2 m, approximately 280 mm less than the unsupported tank wall,
but still produced a yield surface in the tank wall with
a predicted stress level of 208 MPa. Study Case 4)

consisting of tension rods or cables at the 1/8


points along the open end of the tank wall
perimeter, or
ii. Installing a circumferential stiffener flange of
appropriate size on the perimeter of the tank wall
open end.

provided radial stiffening of the tank top perimeter


wall, with the axial stiffeners located at 22.5 around
the perimeter. This produced a maximum 0.132 meter
inward radial deflection of the wall and a peak wall
plate stress of 89 MPa, providing a Factor of Safety of
over 2.0 against yielding. Study Case 5) incorporated a
circumferential stiffener angle installed at the top edge
of the tank wall. This produced a maximum radial
deflection of 0.460 meter and a maximum 97 MPa
stress level in the tank wall plates with a Factor of
Safety = 2.14 preventing yield and permanent damage
to the tank wall.
9

6. Based on estimated construction costs for labor and


materials, the installation of a center column support for radial stiffeners, and placement of radial
stiffeners, would be a less cost effective construction procedure than installing a circumferential
stiffener flange along the perimeter of the open top
edge of the tank wall.

CONCLUSIONS

The FEA model representing the stepped thickness


wall results correlated well with the actual damage
resulting in the tank wall structure due to high wind
gusts. Based on these study results, a number of conclusions are presented.
1. Wind pressure distributions specified for cylindrical, large volume tanks based on API, ASCE and
IBC codes are not adequate for field construction
conditions.
2. The codes assumptions are based on closed ended,
rigid structure and small deformations.
3. Additional turbulence and higher wind pressures
generated by the open end cylindrical plate structures subjected to severe wind gust conditions are
not covered by the design codes.
4. Peak gust wind pressures on an open ended cylindrical structure, represented by the tank wall in this
study case, may be 2 to 4 times higher than values
predicted by the codes, due to turbulence effects at
the open end.
5. Failure of the structurei.e. tank wall snap-thru
buckling, could have been prevented by:
i. Installing a circumferential stiffening angle
around the top perimeter of the tank wall consisting of either temporary construction bracing

REFERENCES
ASCE/SEI 7-05, 2007. Minimum Design Loads for Buildings and Other Structures, American Society of Civil Engineers, Structural Engineering Institute.
Argon, Ali S., 1975. Constitutive Equations in Plasticity,
MIT Press.
Baker, E.H., Kovalevsky, L., Rish, F.L. 1972. Structural
Analysis of Shells. McGraw Hill Book Company.
Calladine, C.R.1985. Plasticity for Engineers, John Wiley &
Sons.
Chen, W.F. & Saleeb, A.F., 1982. Constitutive Equations for
Engineering Materials, Vol 1: Elasticity and Modeling,
John Wiley & Sons
Hodge, Jr. P.G., 1959. Plastic Analysis of Structures, McGraw
Hill Book Company.
International Building Code 2008. International Code Council, Inc. (ICC).
Kollar, L. & Dulacska, E. 1984. Buckling of Shells for Engineers, John Wiley & Sons, Inc.
Kraus, H. 1967. Thin Elastic ShellsAn Introduction to the
Theoretical Foundations and the Analysis of Their Static
and Dynamic Behavior. John Wiley and Sons, Inc.
MSC-MARC 2007 r3, Volume A:Theory and User Information, MSC-Software Corporation.
William, P. & Hodge Jr, P.G. 1968. Theory of Perfectly Plastic
Solids, Dover Publications.

160

You might also like