Orthotropic Toroidal Shell
Orthotropic Toroidal Shell
Orthotropic Toroidal Shell
on the Optimization
of an Orthotropic Composite
Toroidal Shell
Matthew J. Vick
e-mail: matthew.j.vick-1@ou.edu In this research, an analysis technique is developed to model orthotropic composite tor-
oids and optimize the fiber layup, accounting for the natural variation in thickness due to
Kurt Gramoll fiber stacking. The behavior of toroids is difficult to model using membrane shell theories
e-mail: gramoll@ou.edu due to a singularity in the strain-displacement relations occurring at the toroid crest that
yields discontinuous displacement results. A technique is developed here where the con-
Department of Aerospace stitutive properties of multilayered toroidal shells are determined using lamination
and Mechanical Engineering, theory, and the toroid strains and line loads are determined using finite element analysis.
University of Oklahoma, The toroid strains are rotated into the fiber directions, allowing the fiber stress and trans-
865 Asp Avenue, Felgar Hall, Room 237, verse stress distributions to be determined for each layer. The fiber layup is modified heu-
Norman, OK 73019 ristically until an optimum is found. An optimum is reached when the maximum fiber and
transverse direction stresses of each shell layer are equal, minimizing wasted fibers and
excess weight. Test cases are analyzed to verify the accuracy of the finite element model
and an example composite toroid with Kevlar/epoxy material properties is optimized. The
analysis technique developed here can decrease the time and cost associated with the de-
velopment of orthotropic toroidal pressure vessels, resulting in lighter, cheaper, and
more optimal structures. The models developed can be expanded to include a steel liner
and a broader range of fiber winding patterns. [DOI: 10.1115/1.4005873]
Introduction These solutions only consider the global behavior of the toroid,
neglecting the distributions of fiber direction and transverse direc-
A toroid is an axisymmetric shell of revolution with an arbitrary
tion stresses within each layer of the fibrous composite. An analy-
cross-section that does not intersect the axis of revolution. Toroi-
sis technique is presented here to evaluate orthotropic toroids and
dal vessels have been used in industry for the storage of pressur-
optimizes their construction. Tailoring winding patterns to the ge-
ized liquids and gases; notable applications include fuel tanks in
ometry of the toroid can better control the stress distribution
liquefied petroleum gas vehicles, and propellant storage in rockets
within the toroid, evenly stressing fibers thus reducing the propor-
and missiles such as the Russian R-36M ICBM [1]. Toroids offer
tion of wasted fibers. Manufacturing processes for filament wound
advantages over other pressure vessel geometries, such as cylin-
toroids pose unique challenges that are outside the scope of this
ders in certain situations, by providing a more optimal use of
research. Layups are assumed to be hand placed, and this study
space while eliminating wasted material required by end caps.
focuses on the design aspects of toroidal layups.
Manufacturing toroidal pressure vessels using composite mate-
Given an initial fiber layup and toroid geometry, the constitu-
rials offers several distinct advantages over more common con-
tive properties of the multilayer toroidal shell are determined
struction materials such as steel, including lighter weight, and the
using basic lamination theory and calculated using MATLAB, taking
ability to tailor the composite material to the design requirements.
into account thickness variations that arise due to fiber buildup.
By varying construction parameters such as fiber orientations for
The deformation of the toroid is determined using ANSYS, and the
fiber-wound composites, the fiber layup can be optimized to
resultant strains and line forces are output for postprocessing. The
account for the stress variation within the toroid and minimize the
strains experienced by the toroid are rotated into the fiber direc-
weight of the pressure vessel.
tion for each layer, allowing the fiber and transverse direction
Obtaining stress-displacement solutions for toroids are more diffi-
stresses to be determined for each layer of the toroid.
cult than other axisymmetric shells of revolution due to a disconti-
The aim of the optimization is to produce equal stress distribu-
nuity occurring in linear membrane solutions at the crest of the
tions in each layer of the toroid, minimizing wasted fibers thus
toroid. Several stress-displacement solutions for internally pressur-
minimizing weight. The fiber layups are modified in an iterative,
ized isotropic toroids are available in the literature: (a) linear mem-
heuristic process until an optimal fiber layup is reached; optimal-
brane stress approximations [2,3]; (b) closed-form linear
ity is determined using two quantitative fitness measures, the max-
displacement solutions [4–8]; (c) numerical stress-displacement sol-
imum stress in each layer and the ratio of maximum to minimum
utions based on nonlinear formulations [9,10]; (d) finite element sol-
stress in each layer. Due to the computational efficiency of the
utions based on modified linear and nonlinear formulations [11–13].
model, an optimum can be reached quickly.
Solutions also exist for orthotropic toroids [14,15], and layup opti-
mization has been considered for fiber-wound toroids [16,17].
Model Formulation
Contributed by the Pressure Vessel and Piping Division of ASME for publication
in the JOURNAL OF PRESSURE VESSEL TECHNOLOGY. Manuscript received February 5,
Geometry. A circular toroid is formed by revolving a circular
2011; final manuscript received December 21, 2011; published online August 27, cross-section with radius a about an axis with major toroid radius
2012. Assoc. Editor: Young W. Kwon. R, shown in Fig. 1. The thickness t of the cross-section is usually
1X n
k
½ B ¼ Q z2k z2k1 (3b)
2 k¼1
1X n
k
Fig. 1 Circular toroid geometry: (a) section view; (b) top view;
½ D ¼ Q z3k z3k1 (3c)
3 k¼1
and (c) global view
k
for a laminate containing n-layers. Q corresponds to the rotated
assumed constant for metallic materials but can vary for ortho- stiffness matrix of the kth layer of the laminate and zk corresponds
tropic materials. The meridional direction / for a circular toroid is to the distance from the midplane of the laminate to the edge of
tangent to the cross-section and the circumferential direction h layer k. These matrices differ from other constitutive constants
follows the circle of revolution for the toroid. The upper half of such as Young’s modulus owing to the fact they contain geometric
the toroid is defined from 90 < / < 90, with the crest occurring information about the laminate carried by the zk terms.
at / ¼ 0. The bulk engineering constants for a laminate can be deter-
Due to its geometry, a fiber overwound toroid will have a natu- mined from the in-plane stiffness matrix A. The laminate compli-
ral thickness variation due to fiber buildup. The fiber volume is ance, a*, is defined by [18]
constant across the toroid; thus, the thickness of a toroidal shell
2 3
will decrease radially from the axis of revolution due to the a11 a12 a16
increase in the surface area of the toroid. The shell thickness can
0 6 7
be defined as a function of the meridional coordinate, /, and a ref- e ¼ ½a frg where ½a ¼ 4 a12 a22 a26 5 ¼ t½ A1
erence thickness taken to be the thickness at the crest of the toroid, a16 a26 a66
tc [17] (4)
Rtc where r is the vector of the average planar stresses. From a*, the
tð/Þ ¼ (1)
R a sin / bulk engineering constants for the laminate can be defined [18]
1
The larger the ratio of toroid cross-section radius to toroid major Ex ¼ (5a)
radius, the more substantial this thickness variation becomes, a11
As all fibers are oriented within the plane of the laminate, the con-
stants defining behavior through the thickness of the laminate, Ez,
xz, and Gxz, are defined using the fiber and matrix properties of Fig. 3 Laminate fiber orientations
the laminate and are independent of fiber orientation.
(a) Two-dimensional axisymmetric finite element formula-
Finite Element Models. ANSYS offers a variety of tools that aid tion with material properties defined by engineering con-
in the analysis of orthotropic shells, most importantly shell ele- stants and individually defined element thicknesses
ments that allow axisymmetric modeling and material properties (b) Two-dimensional axisymmetric finite element formula-
to be input as preintegrated shell sections using laminate con- tion with material properties defined by engineering con-
stants. ANSYS provides several built-in shell elements for structural stants and element thicknesses defined by a global
analysis; two elements used in this analysis are SHELL209 and function using the SECFUNCTION command in ANSYS
SHELL281. SHELL209 is a one-dimensional, three-node finite (c) Three-dimensional axisymmetric finite element formula-
strain axisymmetric shell element that requires the cross-section tion with material properties defined by preintegrated
of the toroid to be modeled in the x-y plane using lines; the y-axis shell sections.
serves as the axis of revolution. SHELL281 is a two-dimensional,
eight-node finite strain shell element that requires a three- Test cases include:
dimensional model of the toroid; the toroid can be modeled by the
cross-section swept through a small arc with axisymmetric bound- (a) isotropic toroid with constant thickness
ary conditions applied to the free edges of the sweep. SHELL281 (b) isotropic toroid with variable thickness
is one of two ANSYS shell elements that allow material and thick- (c) orthotropic toroid with variable thickness.
ness data to be applied as preintegrated shell sections; for this rea-
son, it is the main element used in this analysis. Each finite element model is based upon a circular cross-section
When utilizing preintegrated shell sections for composite lay- of the toroid. The cross-section is constructed using 360 line arcs
ups, ANSYS requires the values of the A, B, and D matrices of the each spanning 1 deg, thus forming a complete circle. For two-
laminate to be input along with values of an E matrix that defines dimensional models, the lines are directly meshed into elements;
shear behavior through the thickness of the laminate. The E ma- for three-dimensional models, the lines are revolved 1 deg to form
trix is specified by ANSYS as [20] areas which are meshed into elements. Each model contains 360
elements. Pressure loads are applied directly to element faces using
5 Gxz t 0 ten substeps to account for possible nonlinear deformation. The
½ E ¼ (6)
6 0 Gxz t y-displacement of nodes at / ¼ 90 deg is constrained to be zero,
and the results are determined using the default nonlinear solver
Layer Stresses. When considering composite shells, it is con- within ANSYS.
venient to output the element line forces and element planar
strains for further postprocessing. Planar strains allow simple rota-
tions to determine fiber direction and transverse direction stresses Isotropic, Constant Thickness Toroid. An isotropic toroid
in individual layers (transverse stress through the thickness of the with parameters R ¼ 15 in. (381 mm), a ¼ 10 in. (254 mm),
shell is negligible for shell formulations). Utilizing the planar t ¼ 0.5 in. (12.7 mm), E ¼ 10 Msi (68.9 GPa), ¼ 0.3, p ¼ 1000 psi
strains, e1 (meridional strain), e2 (circumferential strain), and e12 (6.89 MPa) is considered for the first test case. Two finite element
(planar shear strain), output by ANSYS, the stress state of each layer models are considered: a two-dimensional model with constitutive
of the shell is determined by [18] properties defined as engineering constants referred to as EC (for
2 38 9 engineering constants) with thickness defined for each element
m2 n2 mn < e1 = individually, and a three-dimensional model with constitutive prop-
frg ¼ ½Qfeg where feg ¼ 4 n2 m2 mn 5 e2 erties defined as laminate constants referred to as PI (for preinte-
: ;
2mn 2mn m2 n2 e12 grated shell sections).
(7) This toroid configuration has been analyzed by Tielking et al.
[11] using modified linear theory (MLT) with published stress and
Q is the layer stiffness matrix, e is the layer strain rotated by ak displacement results. Figure 4 presents displacement and normal-
from the global coordinate system (Fig. 3) into the direction of the ized line force results for the toroid (displacement is shown at
fiber, m ¼ cos(ak) and n ¼ sin(ak). Equation (7) directly gives the 100 times actual magnitude for clarity). The center of the toroid
fiber direction, transverse direction and planar shear stresses cross-section is given by x ¼ 0 and the line force resultants are
within each layer, allowing a comparison of stress states within normalized by p a and presented as a function of meridional posi-
each layer through the cross-section of the toroid and providing tion, /. There is a good agreement between the displacements
the basis of the layup optimization process. predicted by MLT and EC, and the in-plane strain and line force
results obtained from EC and PI are identical. The oscillations
seen in the circumferential line forces differ from those predicted
Model Verification by linear, equilibrium based models. This is due to the fact that
Three models and three test cases are considered to verify the the deformation and thus stresses experienced by toroids are
accuracy and consistency of the finite element formulations for inherently nonlinear, especially approaching the crest of the
internally pressurized toroids. Models include: toroid. This gives rise to oscillation in circumferential stress that
Parameter Value
individual lamina (tl = 0.005 in.) with orientations other than 90,
allowing the overall shell thickness and proper preintegrated shell
constants to be calculated. The total toroid crest thickness, tc, is
defined to be 0.75 in.
For the given toroid geometry, the maximum fiber and trans-
verse stresses within each layer appear to converge to a single
value for an angle-ply near 9 deg off the meridian, as shown in
Fig. 9. Planar shear stresses are calculated as well but not consid-
ered in the optimization; shear stresses are around 2 orders of
magnitude smaller than circumferential stresses and 3 orders of
magnitude smaller than meridional stresses in all cases analyzed.
This convergence is interesting for two reasons: (a) the large
number of unalterable variables (thickness variation, variation in
meridional, and circumferential loads) affecting the stress within
the toroid; (b) it corresponds to the overall minimum of the maxi-
mum stresses within the toroids. In other words, every other fiber
orientation results in a higher maximum stress in some layer
within the toroid. It is also interesting that the same layup is opti-
mum for both the fiber direction and transverse direction stresses.
The relationship between the fiber stress and the angle-ply is
inverted for the transverse stress, which is logical; a fiber rotation
that causes an increase in stress within that fiber will cause a cor-
responding decrease in transverse stress given a constant loading.
To examine the evenness of the stress distributions through the
toroid, the ratios of the maximum to the minimum stress within
each layer for each layup are plotted in Fig. 10. Each stress state
oriented along the meridional direction (fiber directions near
0 and transverse directions near 90) experiences a relatively con-
stant stress ratio between 1.8 and 1.9 owing to the relatively invar-
Fig. 9 Maximum layer stresses with respect to angle ply orien- iant nature of the meridional line load (refer to Figs. 4–6). There
tation: (a) fiber direction; (b) transverse direction is a notable variation in the stress ratio for stress states oriented