You are on page 1of 10

Optimal Design of Anisotropic

(Fiber-Reinforced) Flywheels*
RICHARD M. CHRISTENSEN AND EDWARD M. WU

Lawrence Livermore Laboratory


University of California
P.O. Box 808
Livermore, CA 94550

(Received March 11, 1977)

ABSTRACT

An analysis is given of the kinetic energy storage capacity of anisotropic


flywheels. Using a uniform strain failure criteria, the optimal shapes of
flywheels are determined as a function of the degree of anisotropy. Within
this spectrum of shapes, practical design considerations are shown to favor
the case where there is equal reinforcement in the radial and circumferen-
tial directions. Comparisons are made between the present solid-wheel-type
design and the ring design.

INTRODUCTION
DESIGN OF flywheels has longstanding subject of useful, practi-
been a

HE
cal application. Interest in thesubject now becoming quite active with the
is
prospective use of flywheels part integrated energy conservation systems, both
as of
in stationary powerplants and transportation vehicles. Certainly optimal energy
conservation systems require optimal design of the individual components, and this
work is concerned with the optimal design of flywheels.
In the case of isotropic material design, the optimal configuration is well known.
The optimal shape is that of a radially tapered section, the thickness of which is
governed by the exponential function. The criterion for optimization is, relative to
a prescribed weight or volume, that the flywheels should store the maximum

amount of kinetic energy. Of necessity, the failure criterion for the material enters
into the consideration because maximum energy could be stored only at the thresh-
old of material failure. Furthermore, the maximum energy storage would occur
*This work was partially supported by a contract from the U-S- Army Research Office and
the U.S. Energy Research & Development Administration (Contract W-7405-Eng048).

395
Downloaded from jcm.sagepub.com at UNIV OF PENNSYLVANIA on May 14, 2015
only when all points of the medium are at the same stage of incipient failure, not
just one point or a set of points. Thus the design criterion for isotropic materials is
taken to be the maintenance of a uniform state of stress throughout the entire disk.
The closed-form analyses are restricted to states of plane stress. In this context the
two principal stresses are required to be equal and constant throughout the disk;
therefore the design is often referred to as the constant stress flywheel. On the
other hand, the corresponding problem of the optimal design of anisotropic fly-
wheels has not been successfully treated.
Interest in anisotropic flywheels relates
to modern developments in fiber tech-
nology. However, theoretical
some work on the subject predates the modern fiber
technology era. Specifically, Carrier [1] in 1943 provided a stress analysis for the
case of anisotropic rotating disks of uniform thickness, while Sen Gupta [2] pro-
vided the stress analysis for certain types of anisotropic rotating disks of nonuni-
form thickness. Neither of these works, however, considered the subject of opti-
mum design with the attendant complications of incorporating realistic failure cri-
teria.
The objective here is to provide an optimal design strategy for the flywheel. The
quantity to be optimized is the kinetic energy stored per unit mass. The variables to
be adjusted to achieve the optimal configuration are the thickness of the disk as a
function of radius and the degree and type of anisotropy. The analysis will be
restricted to plane stress conditions. After having obtained exact, closed-form solu-
tions,,.for this problem, the practical questions of designing the wheel with presently
available materials will be considered.

ANALYSIS

The material is taken to be cylindrically orthotropic and in a state of plane


appropriate linear elasticity form of the stress (a)/strain (e) relations is
stress. The
given by

where

r and 0 are the polar coordinates, and the constitutive coefficients are taken to be

independent of position. The moduli Qr’ Q(}, Qr(}’ and QBwill be related to the
common engineering moduli measures later.
The appropriate Poisson ratios are given by

396 Downloaded from jcm.sagepub.com at UNIV OF PENNSYLVANIA on May 14, 2015


where lJr(} is the measure of contraction in the e direction resulting from an applied
load in ther direction.
There arises the necessity of specifying the failure criterion. There are many
different criteria in use and in development for fiber-reinforced composites. Many
of them are quite complicated and would preclude the type of analytical treatment
sought here. A reasonably simple criterion is needed, and yet it is even more
important that it be a realistic specification. To this end, the constant strain criter-
ion is used. At this point it is observed that the strain at failure may be taken to be
that of the fiber phase at failure, if the matrix phase is sufficiently compliant Thus
the fiber phase characteristics will be the governing factor for the composite
Maximum energy storage will be achieved when the material is used in the most
efficient manner. Clearly this corresponds to the condition of simultaneous material
failure at all points of the medium (at the critical speed). A condition consistent
with the use of the strain failure criterion is given such that the strain state is
independent of radius r:

Note that Eq. (4) requires more than the usual statement of a maximum strain
criterion because both components of strain are specified to have the failure value,
rather than just one component.
Under plane stress conditions, the compatibility equation has the form

Obviously this governing condition is satisfied by Equation (4). It remains to satisfy


the equilibrium equation, under steady-state conditions, which has the form

where the thickness of the disk is an unknown function of radius

p is the density, and oj the angular speed.

Downloaded from jcm.sagepub.com at UNIV OF PENNSYLVANIA on May 14, 2015


397
Combining Equations (1) to (4) and (6) gives

This equation can be put into the form

where

Equation (8) may be integrated to obtain

where

and k is the constant of integration.


The objective here is to maximize the kinetic energy stored per unit mass of the
flywheel. The kinetic energy is given by

where the integral is over the plane of the disk. Similarly, the total mass is given by

Forming the ratio of Equations (12) and (13) gives

398 Downloaded from jcm.sagepub.com at UNIV OF PENNSYLVANIA on May 14, 2015


where the polar integration has been performed. Substituting the thickness func-
tion, Equation (10), into Equation (14) gives

The problem now is to evaluate these integrals.


A change of variable facilitates integral evaluation. Let

Inserting Equation (16) into Equation (15) yields

These integrals are the standard form for gamma functions, and Equation (17) can
be written as

Now the following identity (see Reference 3) for gamma functions will be useful:

and Equation (18) can be put into the form

Next inserting the definitions of a from Equation (9) and X from Equation (11)
into Equation (20) gives

Finally, Equations (3) are inserted into Equation (21) to produce

Downloaded from jcm.sagepub.com at UNIV OF PENNSYLVANIA on May 14, 2015


399
It is the right-hand side of Equation (22) that is to be maximized by varying the
degree of the anisotropy. The thickness variation has already been obtained through
the restriction that the material fail simultaneously throughout the entire disk. An
examination of the terms in Equation (22) reveals that there is, in fact, no further
optimization to be performed. The combination of terms QY + QB + 2Qre is an
invariant with respect to coordinate rotation even though the individual terms vary
depending upon the orientation of an individual lamina relative to the given coor-
dinate system. This property of invariance has been derived by Tsai and Pagano [4]
using the tensor transformation laws for anisotropic media properties; thus, whe-
ther the entire reinforcement is radial, or circumferential, or any combination, it is
still the same maximum energy density that is attainable. Of course, to achieve this
maximum energy density, the disk must have the thickness variation dictated by
Equation (10), and very different shapes are involved depending upon the type of
anisotropy involved.
Equation (22) can be put into a form using more familiar properties:

with Er and Eo representing the corresponding modulus measures. Inserting Equa-


tions (23) into Equation (22) yields

The corresponding isotropic material results can be extracted from Equation (24)
by letting

with E and v being the corresponding isotropic properties. Thus for the isotropic (I)
case, the result is

Downloaded from jcm.sagepub.com at UNIV OF PENNSYLVANIA on May 14, 2015


400
A LIMITING CASE RESULT

A first-order evaluation of the results just obtained is given by considering limit-


ing cases. The most obvious limiting case would be that of a fiber phase material in
which the fibers are so stiff in comparison with the matrix phase that the latter
contribution can be neglected.
With a matrix phase of negligible shear modulus, 11m 0, and also using v,, vm,
= =

there results, for the case of radial reinforcement,

where c is the volume fraction of the fiber phase and the subscripts f and m refer to
fiber and matrix phases, respectively. Combining Equations (26) with Equation
(24) produces

where the inequality results from the fact that any matrix contribution can only
increase the prediction of Equation (27).
Christensen and Waals [5] showed that, with vanishing matrix phase modulus,
the two-dimensional random-fiber orientation scheme gives

Inserting Equations (28) into Equation (25) yields for the isotropic results the same
form, Equation (27), which, of course, results from the invariant property of the
right-hand side of Equation (24).
It is important to note that Equation (27) is exactly the same as that obtained

401
Downloaded from jcm.sagepub.com at UNIV OF PENNSYLVANIA on May 14, 2015
from the analysis of a rotating ring. The implications of this result will be discussed
in the next section.

DESIGN CONSIDERATIONS
The final results of the optimization analysis are both Equation (24), which gives
the kinetic energy stored per unit mass as a function of the material properties, and
the corresponding shape solution, Equation (10), which also depends upon the
material properties. A remarkable result emerges in the derivation. As long as the
appropriate optimum shape is used for a given ratio of radial to circumferential
reinforcement, the corresponding energy storage capacity is the same for the entire
spectrum of optimum shapes. The next problem, then, is how to select the &dquo;best&dquo;
shape with the corresponding type of anisotropy. The answer to this question lies in
the realm of practical design. Consider the shape solution of Equation (10). For a <
1, radial reinforcement dominates, and it is seen that the thickness solution is
singular at r 0. On the other hand, for a > 1, circumferential reinforcement
=

dominates and from Equation (10) it is seen that at r 0, h 0. It is only in the


= =

case of a =
1, which corresponds to equal radial and circumferential stiffness, that
there is not an unappealing and, in fact, an impossible design condition at r 0. In =

the case of predominant circumferential reinforcement, it is interesting to note


from Equation (10) that the optimal shape involves a section that thickens with
increasing radius near the origin but has a maximum thickness at some determinable
radius and thereafter tapers to vanishing thickness. Obviously the case of equal
radial and circumferential reinforcement is the most practical configuration.
A flywheel could be constructed with equal radial and circumferential reinforce-
ment. This certainly is more general than the isotropic case. However, in the limit-
ing case results of the preceding section, it was shown that identical results emerge
for the energy storage whether the system is isotropic or just of equal stiffness in
the radial and circumferential directions. (The latter case is orthotropic but not in
general isotropic.) Accordingly, the isotropic result of Equation (25) will be used
for further examination.
One question to be raised at this point is whether the fiber-reinforced flywheel,
with fibers arranged to give an effectively isotropic property characterization, has
any advantage over a ring-type fiber-reinforced flywheel. The limiting case for fiber
effects, shown in the preceding section, revealed these two configurations to give
the same energy storage. First, the solid wheel has a design advantage over the ring:
it is more efficient in space use, having less swept volume. Second, it is likely to be
more efficient and stable under conditions of acceleration and deceleration because
the attachment of the ring-type wheel to spokes and hubs causes problems.
The most basic question to be answered here is which configuration uses matrix
properties in the most efficient manner. (The comparisons of the preceding section
neglected matrix properties.) To answer this question we next examine a flywheel
made by laminating individual uniaxial laminas at various angles to obtain effective

402
Downloaded from jcm.sagepub.com at UNIV OF PENNSYLVANIA on May 14, 2015
isotropic properties. An asymptotic estimate of the resulting isotropic modulus has
been derived by Christensen [6] :

This formula is based upon isotropic property characterizations for both phases.
Deviations from individual phase isotropy, as with Kevlar, would have an effect that
is difficult to assess at this point.* Insertion of Equation (29) into Equation (25)
gives a prediction of the energy storage capacity in terms of both fiber and matrix
phase properties. Equation (29) has been evaluated for Kevlar fiber, epoxy matrix
systems in two separate cases. For a fiber volume fraction of c = 2/3, the results are
as follows:

The former result corresponds to a very flexible epoxy resin while the latter result
is that for
a stiff epoxy resin. For the fiber case alone,

The matrix phase is seen to provide a considerable stiffening effect in the case of
the stiff epoxy. These results should be compared with ring behavior, which is
governed by the rule of mixtures. According to the rule of mixtures, the matrix
phase increases the stiffness by 1.3% over that of fibers alone in the case ofEm/Ef
=
5/190. Compare this result with the 38% improvement shown by the data for the
case of the isotropic solid disk design. A caveat must be inserted at this point. For a
laminated system producing isotropic properties, and using a stiff matrix phase, the
matrix phase would fail before the fiber phase, thus the fiber phase could not be
used to full capacity. However, this situation can be reversed by going to a more
flexible matrix phase, such as the one in the first example in which 1 /190. Em/Ef =

*
Reference to a company or product name does not imply approval or recommendation of
the product by the University of California or the U.S. Energy Research and Development Ad-
ministration to the exclusion of others that may be suitable.

403
Downloaded from jcm.sagepub.com at UNIV OF PENNSYLVANIA on May 14, 2015
In this case the matrix phase still provides far more reinforcement than does the
matrix phase in the ring configuration.
Finally, some mention should be made of other pertinent design approaches.
Although no attempt has been made to survey the many different design ap-
proaches, it is interesting to note that the numerical optimization scheme of Gerstle
and Biggs [7] for circumferentially reinforced flywheels produces shapes in accord-
ance with expression (10) for a > 1.

REFERENCES

1. G. F. Carrier, Trans. ASME Vol. 65, (1943) A-117.


2. A. M. Sen Gupta, Bull. Calcutta Math. Soc. Vol. 41, (1949), p. 129.
3. M. Abramowitz and L. A. Stegun (Eds.), Handbook of Mathematical Functions (National
Bureau of Standards, Applied Mathematics Series 55, Washington, D.C., 1964).
4. S. W. Tsai and N. J. Pagano, "Invariant Properties of Composite Materials" inComposite
Materials Workshop, S. W. Tsai, J. C. Halpin, and N. J. Pagano, Eds. (Technomic Publishing
Co., Inc., Westport, Conn., 1968).
J. Compos. Mater. Vol. 6, (1972), p. 518.
5. R. M. Christensen and F. M. Waals,
Int. J. Solids Struct. Vol. 12, (1976), p. 537
6. R. M. Christensen,
7. F. P. Gerstle, Jr. and F. Biggs, Proc. 12th Annual Meeting Society Engineering Science,
(1975), p. 143.

404
Downloaded from jcm.sagepub.com at UNIV OF PENNSYLVANIA on May 14, 2015

You might also like