Вы находитесь на странице: 1из 23

Journal of the Mechanics and Physics of Solids

49 (2001) 445 467


www.elsevier.com/locate/jmps
Analytical solutions and numerical procedures
for minimum-weight Michell structures
Peter Dewhurst

Department of Industrial and Manufacturing Engineering, University of Rhode Island, Kingston,


RI 02881, USA
Received 4 February 2000; received in revised form 11 August 2000
Abstract
A power-series method developed for plane-strain slip-line eld theory is applied to the con-
struction of minimum-weight Michell frameworks. The relationship between the space and force
diagrams is dened as a basis for weight calculations. Analytical solutions obtained by the
method are shown to agree with known solutions that were obtained through virtual displace-
ment calculations. Framework boundary conditions are investigated, and matrix operators used
in slip-line eld theory are shown to apply to the force-free straight framework boundary-value
problem. The matrix operator method is used to illustrate the transition from circular arc-based
to cycloid-based Michell solutions. Finally, an example is given in the use of the method for
evaluation of support boundary conditions. c 2001 Elsevier Science Ltd. All rights reserved.
Keywords: Michell; Minimum-weight; Framework; Slip-line eld
1. Introduction
It has long been recognized (Chan, 1960; Johnson, 1961; Chan, 1963) that a
remarkable coincidence exists between plane-strain slip-line-eld theory for metal
deformation and two-dimensional minimum-weight structures. The partial dierential
equations, which dene the shapes of the lines of maximum shear stress (the slip
lines) in the former, are precisely the same as the equations dening the shape of
the tension and compression members in the latter. However, while large numbers
of solutions have been determined for two-dimensional metal deformation problems,
the number of non-trivial minimum-weight structural design solutions is small. A

Fax: +1-401-874-5540.
E-mail address: dewhurst@egr.uri.edu (P. Dewhurst).
0022-5096/01/$ - see front matter ? 2001 Elsevier Science Ltd. All rights reserved.
PII: S0022- 5096( 00) 00053- 3
446
P. Dewhurst / J. Mech. Phys. Solids 49 (2001) 445 467
comprehensive account of the properties of Michell structures, and descriptions of
known solutions can be found in Hemp (1973).
One reason for the dierence, in the numbers of solutions, may be that in plane-strain
metal deformation theory, the challenge has been to establish solutions for specic
boundary-value problems dened by commonly used tool geometries. In contrast, for
structural design the search has been for spatially unrestricted optimum structures of
absolute minimum weight. The analogy to this in metal deformation theory would
be the search for a network of maximum shear-stress lines that produce a required
deformation with least work. The tool boundary then becomes part of the solution,
and for this type of problem only a very few slip-line-eld theory solutions are known
(Richmond and Devenpeck, 1962; Richmond, 1968; Alexandrov, 1999).
The aim of this paper is to suggest that a more fruitful approach to optimum struc-
ture design may be to dene the structure boundary, or its general form, and then to
use a boundary-value approach to generate possible solutions. It will be shown how
power-series solution and matrix operator methods, developed for the generation of
slip-line elds, may be used in such a boundary value approach. Also, a new method
for calculating structure weight, based on the relationship between the structure space
diagram and its associated force diagram, is established in the paper.
2. Optimum two-dimensional structures
In a remarkable paper, Michell (1904) dened the requirements for least-weight
structures to support a given set of loads. He showed that for minimum weight, all
of the structural elements must be strained, under the action of the external forces,
by exactly the same strain magnitude, c, in either simple tension or pure compression.
Michells paper, and the analyses presented below, are based on the concept of network
continua. These contain an innitely dense arrangement of innitesimally thin tension
and compression members. The real-world expression of such frameworks must contain a
nite number of discrete members, each with cross-sectional area proportional to the
force it must carry. A solid model of a discrete approximation to a Michell structure
is illustrated in Section 3. The material used for such a structure should have the same
stressstrain relationship in tension and compression, so that equal strain magnitudes
are maintained under increasing load. This gives rise to the additional property that
a Michell structure cannot be less sti than any other structure of the same weight
(Hegemier and Prager, 1969).
Two dierent structural types are possible for the separate regions of a full Michell
structure. First, all of the members in a region may be in tension, or all in compression.
For these regions the tie bars, or struts, can readily be laid out in ways suitable for
supporting the external forces. The cross-sectional areas of the members are made
proportional to the forces they carry so that the requirement for equal strain is achieved.
If a region of the structure must carry both tensile and compressive loads then,
from elementary two-dimensional strain distribution theory, tension and compression
members must everywhere intersect orthogonally. The most general case of such a
structural region is shown in Fig. 1. In this gure, the family of lines to which oa
P. Dewhurst / J. Mech. Phys. Solids 49 (2001) 445 467
447
Fig. 1. General region of Michell framework.
and bp belong, will be referred to as :-lines. The orthogonal family of lines, which
include ob and ap, will be called [-lines. Parameters : and [ also represent the angles
turned through by the corresponding lines, from the base point o, as represented by
the arrows in Fig. 1.
The fact, that the right angles at o, a, b and p must be unchanged during equal
extensions and contractions of the adjacent sides, requires that the angles turned through
by members of the same family (e.g. ob and ap in Fig. 1) must be equal (Michell,
1904). This is the property satised by slip-line elds, which leads to the relationship
dening the radii of curvature of the lines as (Hill, 1950)
cR
c[
=S,
cS
c:
=R, (1)
where R and S denote the radii of curvature of the : and [-lines, respectively, and
radius of curvature is positive for a counter-clockwise turning line. Eqs. (1) only apply
if both families of lines are curved in the region under consideration. However, one or
both families of lines may be straight in a particular region. In these cases the lengths
of straight-line segments, cut o by lines of the other family, are equal (Hill, 1950). A
particular case occurs when the straight lines converge to a single point and members
of the other family are concentric circular arcs. These, so-called centered fans, occur
frequently in slip-line eld theory, and have formed the basis for a number of simple
Michell structures (Hemp, 1973). They also form the basis for a Michell cantilever
solution discussed in the next section.
If the radii of curvature of the base lines oa and ob are represented by
R
0
(:) =

n=0
a
n
:
(n)
, S
0
([) =

n=0
b
n
[
(n)
, (2)
448
P. Dewhurst / J. Mech. Phys. Solids 49 (2001) 445 467
where [
(m)
= [
m
}m!, then Ewing (1967) has shown that the radii of curvature of the
lines bp and ap passing through the general point p are given by
R(:, ) =

m, n=0
(a
n
:
(m+n)

(m)
b
n
:
(m)

(m+n+1)
), (3)
S(0, [) =

m, n=0
(a
n
0
(m+n+1)
[
(m)
+b
n
0
(m)
[
(m+n)
). (4)
Based on Ewings series solutions and a matrix algebraic representation of slip-line
geometries by Collins (1968), Dewhurst and Collins (1973) established a formal nu-
merical procedure using a number of fundamental matrix operators. This provides an
ecient numerical method, since Eqs. (3) and (4) are highly convergent. Matrix opera-
tors truncated to only six rows and columns provide radius of curvature values accurate
to eight signicant gures (Dewhurst and Collins, 1973). Since one of the main pur-
poses of this paper is to encourage the use of the matrix numerical method for Michell
framework analysis, a brief review of the method is given below. Represent R
0
(:) and
S
0
([) by the column vectors of the coecients in Eq. (2):
R
0
=

a
0
a
1
a
2

, S
0
=

b
0
b
1
b
2

(5)
then R(:, ) and S(0, [) can be represented by
R
1
=

r
0
( )
r
1
( )
r
2
( )

, S
1
=

s
0
(0)
s
1
(0)
s
2
(0)

, (6)
where
R
1
=P

R
0
+Q

S
0
, (7)
S
1
=P
0
S
0
+Q
0
R
0
(8)
and
P
[
=

[
(0)
0 0
[
(1)
[
(0)
0
[
(2)
[
(1)
[
(0)


, (9)
P. Dewhurst / J. Mech. Phys. Solids 49 (2001) 445 467
449
Fig. 2. :-bundle in neighborhood of general point :, [.
Q
[
=

[
(1)
[
(2)
[
(3)

[
(2)
[
(3)
[
(4)

[
(3)
[
(4)
[
(5)


. (10)
The weight of a network of tension and compression members can be determined
from the radius of curvature relationships provided the associated force distribution
is known. Fig. 2 shows a short span of :-members in the neighborhood of general
coordinate point (:, [). Let 1(:, [) dene the tensile force radius associated with
the :-members such that the increment of force acting on the bundle of :-members,
lling width So[ is equal to 1o[. Let oA be the cross-sectional area of the bundle.
The weight, of the short span of :-members in Fig. 2, is then
oW =jR(:, [):oA,
=jR(:, [):1(:, [)o[}o, (11)
where o is the constant stress magnitude in the :-members, and j is the material
density. Thus, the weight of the :-members in the network of Fig. 1 is
W
:
=
j
o

0
0
R(:, [)1(:, [) d: d[. (12)
Similarly, the weight of the [-members is
W
[
=
j
o

0
0
S(:, [)C(:, [) d: d[, (13)
where C(:, [) denes the compressive force radius acting on the [-members.
In the following sections the weight of Michell frameworks will be established by
applying these equations to the separate frame regions. The force distributions 1(:, [)
and C(:, [) will be obtained from associated force diagrams. For simplicity of pre-
sentation the ratio j}o will be omitted so that the results presented will have units of
work. French (1999) has previously proposed using the product of force and length, a
quantity that he dened as pertinacity, for analyzing minimum-weight frameworks.
The weight of a Michell structure can also be obtained indirectly from the product
of applied forces and the displacements produced by them (Chan, 1960). Using this
450
P. Dewhurst / J. Mech. Phys. Solids 49 (2001) 445 467
Fig. 3. Basic network representing punch indentation or tip-loaded cantilever.
approach the weight is given by
W =
j
oc
n

i=1
F
i
u
i
, (14)
where F
i
is the external forces acting on the structure, i = 1, 2, . . . , n, u
i
the virtual
displacements at the points of force application, and c the constant strain magnitude
acting throughout the structure.
In the following sections, the results of applying Eqs. (12) and (13) will be compared
with known solutions based on Eq. (14).
3. Basic cantilever solution
The most basic network, dened by two orthogonal circular arcs, has been used to
develop a wide range of solutions in plane-strain metal deformation (Hill, 1950). In
particular, the network shown in Fig. 3, based on equal-span circular arcs, represents
the solution for punch indentation where at punch face cd moves towards point p. It
appears that Chan (1960) was the rst to recognize that this network also represents
the Michell framework for a minimum-weight cantilever; referred to here as the basic
cantilever. In this case the cantilever is subjected to a point load at p, parallel to dc.
Cantilever support forces exist at c and d. Region dob contains only tension members
radiating outwards from point d. Likewise, cao comprises only radial compression
members. There is a smooth distribution of forces throughout the structure, except
along the boundaries cap and dbp where a force discontinuity occurs. This produces
concentrated forces that require a compression ange along cap and a tensile ange
along dbp. This force discontinuity corresponds to a velocity discontinuity around the
deformation zone in punch indentation. Points c and d where tension and compression
members converge, correspond to stress singularities at the corners of die contact in
punch indentation. Region cod represents a dead-metal zone in punch indentation that
P. Dewhurst / J. Mech. Phys. Solids 49 (2001) 445 467
451
moves with the punch. This corresponds to a region of uniform strain, in the Michell
structure, that requires no internal support and is therefore stress free. In addition
to the dead-metal zone, the regions outside of dbp and cap are also assumed to be
non-deforming during indentation. Assumed non-deforming regions in slip-line-eld
theory often impose limits on the range of valid solutions. For example, for [ greater
than approximately 77

, the material beyond db and ca is actually deforming and the


process changes to surface indentation (Hill, 1950). Analogous limitations may exist
on the permissible range of Michell framework solutions, caused by the incompatibility
of assumed stress-free regions with the forces applied to the structure. It seems that
the only possible violation of the basic cantilever solution would be if the line of
action of the support forces at c and d did not pass within the sweep of angles [.
The corresponding force diagram, described later in this section, shows that this cannot
occur.
Chan (1960) noted that since the elements of the basic cantilever are all subjected to
strain magnitude, c, then radial elements db to do will all increase by rc, and elements
co to ca will all decrease by rc. The changed circular arcs oa and ob thus form
the new boundary for calculating displacements for any point within oapb. With this
approach Chan showed that the displacement of the force at point p is given by
u() =

2cr[(1 + 2[)I
0
(2[) + 2[I
1
(2[)]. (15)
Combined with Eq. (14), this provides a truly elegant result, for the cantilever weight,
from the knowledge of just a single force displacement. Chan (1960) was also able
to determine the thickness distribution throughout the basic cantilever. However, it is
unlikely that the analytical solutions obtained by Chan, through the use of Riemanns
method, can be extended to more complex structures with more dicult to handle
boundary conditions.
An alternative approach is to establish the relationship between the framework and
the distribution of the forces in its members in order to utilize weight expressions of
the type given by Eqs. (12) and (13). A key to this relationship, is the insight by
Johnson (1961) that a more powerful similarity exists, between Michell structures and
analogous slip-line elds, than just the sharing of Eqs. (1). Johnson recognized that the
velocity (or hodograph) diagram in slip-line eld theory (Green, 1954) corresponds to
the force diagram of a Michell structure. It is clear that since all of the members of a
Michell structure are in pure tension or compression, then the forces in the members
must trace out a network of lines which remain orthogonal during deformation, thus
satisfying Eqs. (1). This result can be established rigorously from the force equilibrium
equations; see Chan (1960).
Fig. 4 illustrates the relationship between the space diagram and the force diagram for
the basic cantilever. It shows the remarkable result that the force diagram is identical
in form to the cantilever itself. The straight radial lines, in the force diagram, represent
the magnitude of the force discontinuity to be carried by the boundary anges of the
cantilever. The curved network of the force diagram, within b

, represents the
force radii as dened in Fig. 2 and incorporated into Eqs. (12) and (13). The positions
within the two diagrams in Fig. 4 are essentially reversed. The triangle of forces
at point are represented by c

, and in particular d

represents the applied


452
P. Dewhurst / J. Mech. Phys. Solids 49 (2001) 445 467
Fig. 4. Relationships between (a) space diagram and (b) force diagram.
vertical force F

at . In general, co-ordinates (0, ) in the space diagram correspond


to ([ , [ 0) in the force diagram. Thus, if R(0, ) is the radius of curvature of
the tension member passing through general point q in Fig. 4(a), then R

([ , [0)
is the radius of the corresponding force line passing through corresponding point q

in Fig. 4(b). Also, in Fig. 4, length r

, equal to R

o , is the magnitude of force oF


acting on tension members passing through qr of length So .
Thus, applying Eq. (12), and substituting R

([ , [ 0) for 1(0, ), the weight


of the tensile elements within oapb is given by
W
:1
=

[
0

[
0
R(0, )R

([ , [ 0) d0 d. (16)
P. Dewhurst / J. Mech. Phys. Solids 49 (2001) 445 467
453
Since oa and ob in Fig. 4(a) are represented in Eq. (2) by
a
0
=r, a
1
= 0, a
2
= 0, . . . ,
b
0
=r, b
1
= 0, b
2
= 0, . . .
then Eq. (3) becomes
R(0, ) =r

m=0
(
(m)
+
(m+1)
)0
(m)
. (17)
Similarly,
R

([ , [ 0) =s

m=0
(([ 0)
(m)
+ ([ 0)
(m+1)
)([ )
(m)
. (18)
Substituting Eqs. (17) and (18) into Eq. (16) gives
W
:1
=rs

m, n=0

[
0
(
(m)
+
(m+1)
)([ )
(n)
d

[
0
(([ 0)
(n)
+([ 0)
(n+1)
)0
(m)
d0 (19)
which can be integrated by parts to give
W
:1
=rs

m, n=0
([
(m+n+1)
+[
(m+n+2)
)
2
. (20)
Turning now to the radial tension members db to do in Fig. 4(a), the corresponding
forces exist normal to the boundary bo dened by 0 = 0 and = 0 to [. From
the co-ordinate transformation in Fig. 4(b) the corresponding force line is b

. The
incremental weight for region dob is thus
oW
:2
=rR

([, [ )o (21)
and so
W
:2
=rs

[
0

m=0
([
(m+1)
+[
(m)
)([ )
(m)
d (22)
=rs

m=0
([
(m+1)
+[
(m)
)[
(m+1)
. (23)
The outer tension boundary dbp is dened by =[, 0 =0 to [, which corresponds
to radial lines d

to d

in the force diagram. Thus the tension ange, dbp, of the


cantilever carries force s =F

2, as does the compression ange, cap. The weight of


db is the product of lengths db and d

which has value rs. The increment of weight


for bp is
oW
:3
=rR

(0, [)o0 (24)


454
P. Dewhurst / J. Mech. Phys. Solids 49 (2001) 445 467
Table 1
Volume and proportions of the basic cantilever
Angle Volume Radial Orthogonal Outer
[(

) (r = 1, F

= 1) members bracing anges


10 2.0539 0.1431 0.0252 0.8317
20 3.0692 0.2287 0.0818 0.6895
30 4.6317 0.2724 0.1498 0.5778
40 6.9959 0.2896 0.2186 0.4918
45 8.5899 0.2918 0.2517 0.4565
50 10.5371 0.2911 0.2835 0.4254
60 15.8059 0.2839 0.3427 0.3734
70 23.6069 0.2721 0.3959 0.332
80 35.1146 0.2582 0.4433 0.2985
90 52.0393 0.2437 0.4853 0.2709
and thus
W
:3
=rs

1 +

[
0

m=0
([
(m+1)
+[
(m)
)0
(m)
d0

(25)
or
W
:3
=rs +W
:2
. (26)
By symmetry, the compression elements contribute the same weight as the tensile
ones. The total weight of the cantilever is thus twice the sum of W
:1
, W
:2
and W
:3
,
corresponding to the radial members, the internal orthogonal bracing, and the outer
anges, respectively. These can be combined to give
W =rs

2 + 4

m=0
([
(m)
+[
(m+1)
)[
(m+1)
+ 2

m=0

n=0
([
(m+n+1)
+[
(m+n+2)
)
2

.
(27)
Eq. (27) can be veried by inspection of the coecients to be equivalent to Chans
weight solution; Eqs. (14) and (15). Table 1 presents the results of Eq. (27) for [
values between 10 and 90

. The second column gives total volume values that agree


with those calculated by Chan (1960).
As noted earlier, Eq. (27) represents the weight of framework continua, that must be
approximated by a nite number of discrete structural members. Fig. 5 illustrates such
a discrete structure. It is a computer printout from a solid model of a basic cantilever
for which [ = 60

, and discrete members have been generated every 12.5

.
4. Framework boundaries
In this section we will dene the conditions for an external force-free boundary
alongside an intersecting network of both tension and compression members. For the
boundary to be free of external shear tractions, it is clear that either the tension or
P. Dewhurst / J. Mech. Phys. Solids 49 (2001) 445 467
455
Fig. 5. CAD solid model of a basic cantilever.
compression members must meet the straight boundary tangentially. In this way the
boundary becomes a consistent part of the framework, subjected to either tension or
compression. At any intersection angles other than 0 and 90

, an external shear force


is required to support the combined tangential components of the tension and compres-
sion members. An external shear force-free boundary of a Michell framework, with
intersecting tension and compression members, must therefore be an envelope of either
the tension or the compression members.
Fig. 6 illustrates a small region adjacent to a straight boundary, and the corresponding
portion of the force diagram, for the condition of zero normal and shear force on the
boundary. The condition of tangential intersection at all points along the boundary,
implies that both bc and b

must intersect the boundary at right angles. Hence, the


small angular spans of all four curved line segments are the same and will be denoted
:. In terms of the co-ordinate system in Fig. 1, the boundary is thus dened by
0 = . We will assume that the angle between the chord ab and the boundary in
Fig. 6(a) is equal to m: for some multiplier value m1. The condition for zero
normal force increment on ac, using the small-angle approximation is
tan(m:) m: =R

}S

(28)
that in the limit as : tends to zero implies that R

must be zero on the boundary. It


has been shown, in slip-line eld theory, that the envelope of one family of slip-lines is
always the locus of cusps of members of the other family, where the radii of curvature
pass through zero; see for example Kachanov (1973). Thus the condition, R

equal
to zero, is a consequence of the boundary being an envelope, and so in Fig. 6(a) the
radius of tension members is also zero along the boundary.
The boundary condition can thus be written, with reference to Fig. 7,
R(0, ) = 0 when =0. (29)
456
P. Dewhurst / J. Mech. Phys. Solids 49 (2001) 445 467
Fig. 6. (a) small region of network adjacent to straight boundary, (b) corresponding region of force diagram.
Fig. 7. Michell framework adjacent to straight shear force-free boundary.
Substituting this condition into Eq. (3) gives
a
n
=b
n1
for n = 1, 2, 3, . . . ; a
0
= 0. (30)
The boundary value problem dened by Eq. (29), and for which Eq. (30) provides a so-
lution, has been investigated by Dewhurst and Collins (1973) for deformation against a
perfectly rough tool surface.
The simplest approach to determining R
1
from S
0
in Fig. 7 is to use the matrix
operators in Eq. (7). If the tangential member cb in Fig. 7 is dened by column vector
S
0
={s
0
, s
1
, s
2
, . . .} and the opposite boundary member ba is R
0
={r
0
, r
1
, r
2
, . . .} then
R
1
=P
[
R
0
+Q
[
S
0
, (31)
where from Eq. (30), the coecients of R
0
are {0, s
0
, s
1,
s
2
, . . .}.
P. Dewhurst / J. Mech. Phys. Solids 49 (2001) 445 467
457
Alternatively, Eq. (31) can be written as
R
1
= (P
[
J +Q
[
)S
0
, (32)
where
J =

0 0 0 . .
1 0 0 . .
0 1 0 . .
. . . . .
. . . . .

. (33)
Referring back to Fig. 6, it can be seen that a horizontal force increment oF
f
, equal
to S

:, is added to the boundary over length o!, equal to S:. Hence the force, F
f
,
to be carried along the boundary ange, changes along the ange according to
dF
f
d!
=
S

S
. (34)
More general forms of Eq. (32) have been used by Dewhurst (1978, 1985) to
investigate the deformation characteristics of metal machining. The condition that one
family of framework members has zero curvature, along an envelope of members of
the opposite family, remains true if the envelope is curved. A more general linear
representation of the envelope boundary condition can be represented given by
R(0, ) = 0 when =n0, (35)
where n is a constant multiplier. For n not equal to unity, this will generate a curved
boundary that emerges as part of the solution. Dewhurst (1985) has established a
general matrix operator that can be applied to this curved boundary problem.
For now we will restrict attention to the problem of straight framework boundaries.
5. Extensions of the basic Michell cantilever
The basic cantilever, in Fig. 4(a), can be extended to provide a wider range of
least-weight framework solutions for cantilevers supported by two forces at the root
and subjected to a single external load at the tip. For a valid solution both a space
and corresponding force diagram must satisfy Eqs. (1), where the force diagram is
everywhere aligned with the corresponding elements in the space diagram and satises
the equilibrium requirements of the external forces.
The rst such extension of the solution range, proposed by Chan (1960), is obtained
by selecting dierent values for the radii co and do, in Fig. 4(a), and for the angles
aco and odb. An example is shown in Fig. 8(a). The angle of inclination, z, of the tip
force to be at point e can take any value between 0 and }2, and a force diagram can
be constructed as shown in Fig. 8(b). Dierent values of z change the ratio of s
1
and
s
2
in Fig. 8(b), and hence the relative magnitudes of tensile and compressive force in
the framework. When [
1
is not equal to [
2
, no value of z exists for which the force
diagram is identical to the space diagram. However, when and z are equal, the two
diagrams become mirror images of each other, although no special signicance appears
458
P. Dewhurst / J. Mech. Phys. Solids 49 (2001) 445 467
Fig. 8. (a) space diagram and (b) force diagram for the generalized basic cantilever.
to be attached to this symmetry. When [
1
and [
2
are the same, setting z equal to
}2 gives a force diagram identical to the space diagram and for which the tip load
is vertical (parallel to dc).
A second extension of the solution range of the basic cantilever can be obtained by
continuing the straight boundary lines db and ca in Fig. 8(a). In Fig. 9, the straight
boundaries have been extended to points g and f. Regions beg and aef are established
using the boundary value solution from the previous section. Then region efgh is
constructed using Eqs. (7) and (8) but with the vector solutions for ef and eg substituted
for R
0
and S
0
, respectively. The angle z of the tip load at point h can again take any
angle between 0 and }2. However, when z and are equal, the force diagram is
identical to the space diagram and the tip load is vertical (parallel to dc). It can be
noted that if the tip force in Fig. 9 was applied at point [, then regions bghfeb would
be unstressed. Regions dbefac then become the simplest one-sided extension of a basic
cantilever. This structural form was proposed by Kozlowski and Mroz (1970), and
deections of this structure with [
1
=[
2
=}4 were obtained graphically by Johnson
et al. (1971).
The cantilever in Fig. 9 will be referred to as having one boundary extension. Clearly,
straight lines dbg and caf can be extended further and the process repeated through
P. Dewhurst / J. Mech. Phys. Solids 49 (2001) 445 467
459
Fig. 9. Extended basic cantilever.
Fig. 10. Symmetrical constant-thickness extended cantilever.
any number of boundary extensions. It can be shown that for an odd number of such
extensions, the space diagram and the force diagram are identical if the tip load is
vertical. For an even number of extensions this condition can only be achieved for [
1
equal to [
2
.
The general range of extended basic cantilevers described above are only amenable
to numerical evaluation, and the position of the cantilever tip with respect to the sup-
port points emerges as part of the numerical solution. A particular solution for given
co-ordinate positions of the support points and the cantilever tip, would require the ap-
plication of a numerical search procedure. As an example of the numerical procedure,
the special case of [
1
= [
2
= }4 and r
1
= r
2
will be presented. This is an interest-
ing example in that it illustrates a link between this special solution and a Michell
framework composed of only cycloid members (Hemp, 1973).
Fig. 10 shows the rst two boundary extensions of this special case, which is equiv-
alent to the slip-line eld for compression between perfectly-rough dies (Hill, 1950).
The constant height of the cantilever is H and so column vector R
0
is given by
R
0
={H}

2, 0, 0, 0, . . .} (36)
460
P. Dewhurst / J. Mech. Phys. Solids 49 (2001) 445 467
Table 2
Successive column vectors of the extended cantilever
R
0
S
1
R
1
S
2
R
2
S
3
R
3
S
4
R
9
S
10
1.000 1.785 1.773 2.594 1.968 2.796 1.997 2.826 2.000 2.828
0.000 1.094 1.140 0.665 1.792 0.182 1.969 0.029 2.000 0.000
0.000 0.389 2.334 2.564 2.491 3.166 2.157 2.962 2.000 2.828
0.000 0.097 1.768 3.785 1.283 1.041 2.205 0.091 2.000 0.000
0.000 0.018 0.879 3.047 5.390 5.520 3.253 4.196 2.000 2.828
0.000 0.003 0.330 1.767 6.716 11.121 3.784 0.749 2.000 0.000
0.000 0.000 0.100 0.813 5.518 12.515 14.889 16.718 2.002 2.830
0.000 0.000 0.026 0.311 3.484 10.338 22.819 35.389 2.008 0.006
and Eqs. (8) and (32) can be used successively to determine boundary elements,
S
1
, R
1
, S
2
, R
2
, and so on. Table 2 gives the results of such calculations for H =

2,
using 8 8 matrices, up to the compression member dened by S
10
. It can readily be
seen that the elements of the column vector representations are converging to
R =

2
2
2
2
.
.

, S =

2
0
2

2
0
.
.

(37)
and substituted into the original power series these give, with respect to the co-ordinates
dened on region ghj in Fig. 10,
R(0) = 2(sin 0 cos 0), (38)
S( ) = 2

2 cos . (39)
These are the equations for the radii of curvature of cycloids. R(0) is generated by a
circle of diameter

2 rolling clockwise along the upper boundary and starting at the
45

position. S( ) is generated by the circle rolling counter-clockwise along the upper


boundary. We should note that if the direction of R(0) is reversed, starting for example
from point ) in Fig. 10 then its equation becomes
R(0) =2

2 sin 0. (40)
It is easily recognized that opposing families of cycloids dened by Eqs. (39) and
(40) satisfy Eqs. (1). Prandtl (1921) established that cycloid slip-lines could describe
the deformation between perfectly rough compression dies, and Hill et al. (1951)
showed that the solution is only valid for dies of innite width. The numerical re-
sults above simply illustrate the link between the Prandtl and Hill solutions.
Hemp (1973) showed that a cantilever beam with cycloid inner structure could sat-
isfy Michells conditions for minimum weight. Hemps solution, however, requires a
combination of both a vertical force and a couple applied at the tip for equilibrium
with two-point support at the root. He obtained an analytical solution for the virtual
P. Dewhurst / J. Mech. Phys. Solids 49 (2001) 445 467
461
Fig. 11. Proposed support system for a cycloidal cantilever.
displacements of the structure, and following the approach used by Chan (1960) was
able to obtain an estimate for the total structure weight. He also estimated the error
in the cantilever weight resulting from the undesired couple at the tip. The nature of
the support boundary for a cycloid cantilever subjected to only vertical tip loading will
be investigated in the next section.
Returning to Fig. 10, the radii of curvature at any given point in the boundary
regions aeg, ghj, etc., can be shown to be
R(0, ) =

m, n=0
s
n
(0
(m+n+1)

(m)
0
(m)

(m+n+1)
), (41)
S(0, ) =

m, n=0
s
n
(0
(m)

(m+n)
0
(m+n+2)

(m)
). (42)
The weight of the extended cantilever can thus be obtained following the same
region by region integration as presented in Section 3. However, in this case analytical
solutions are improbable and resort must be made to numerical methods. For the present
we will restrict attention to the support boundary conditions associated with an entirely
cycloidal cantilever, and then proceed to calculate its weight.
6. The cycloid cantilever
In order to eliminate the undesired tip moment in a cycloid-frame cantilever, it is
necessary to apply an additional moment of the same magnitude into the root support
system. Such a support system is illustrated in Fig. 11, where spline-type connections
around a curved boundary surface supply the necessary shear tractions. The shape of
the boundary curve, shown in Fig. 11, assumes the results that will be established
below.
462
P. Dewhurst / J. Mech. Phys. Solids 49 (2001) 445 467
Fig. 12. (a) Support boundary conditions for a cycloidal cantilever, and (b) corresponding portion of the
force diagram.
Fig. 12(a) shows two cycloid members starting from the support points at b and
a, and two more continuing from points q and [. For ease of interpretation, only
two intermediate members are shown, starting from point on the support boundary.
Consistent with the desired support system in Fig. 11, it is assumed that a curved
boundary can be drawn between support points a and b, and normal to which there
are no external forces. At an arbitrary point on this boundary, the structure members
intersect the curve at angles [ and }2 [, respectively, and the angular spans of
the members are dened by z and }2 z. In Fig. 12(b), F

represents the vertical


load carried at the tip of the cantilever and F
h
represents the force carried in the upper
tension boundary passing through bf, and in the lower compression boundary extending
through aq. Also c

and d

represent the force distributions carried by members


passing through cp and dp, respectively. The radii of curvature of cp and dp at point
are given by
R = 2H sin z, (43)
S = 2H cos z. (44)
P. Dewhurst / J. Mech. Phys. Solids 49 (2001) 445 467
463
Fig. 13. (a) Incremental lengths adjacent to boundary at point p, and (b) incremental forces.
Similarly, the radii of curvature of d

and c

at point

are
R

= 2F

sin z, (45)
S

= 2F

cos z. (46)
Using these relationships, the boundary conditions adjacent to point are illustrated
in Fig. 13, where R
b
is the radius of the boundary at and F
s
is the incremental
shear force along the boundary.
From Fig. 13(a) the relationship between adjacent side lengths gives
sin([)cos(z)0 = cos([)sin(z) (47)
and from Fig. 13(b) the condition for zero force component normal to the boundary
gives
cos([)sin(z)0 = sin([)cos(z). (48)
It can be seen that Eqs. (47) and (48) are both satised only if [ =z and 0 = .
Thus, following path pqn in Fig. 13(a), it can be seen that the angular span from the
lower boundary aq to point n is z+2 . The angle of intersection of the compression
member at point n is thus
[ =z + 2. (49)
The angular span of n along the boundary is thus
= 4. (50)
Hence from triangle pnq we obtain
(4R
b
)
2
= 4H
2
(sin
2
+ cos
2
)
2
(51)
or
R
b
=H}2. (52)
The boundary is thus the semi-circle between a and b in Fig. 12(a).
Solving for the incremental shear force in Fig. 13(b) gives
F
s
=R

b
= 2F

(cos
2
+ sin
2
) (53)
and substituting for from Eq. (50) gives
R

b
=F

}2 (54)
464
P. Dewhurst / J. Mech. Phys. Solids 49 (2001) 445 467
Fig. 14. (a) Section of Warren truss cantilever, (b) section of cycloidal cantilever.
consistent with the force diagram in Fig. 12(b). The vertical component of shear forces
around the semi-circular boundary thus equals F

fully supporting the cantilever tip


load, and the support forces F
h
in Fig. 12(a) are therefore horizontal as shown. The
moment produced by the shear force around surface ab is F

H}4 equal to the un-


wanted bending moment identied by Hemp.
The weights of the separate regions of the cycloidal cantilever can readily be deter-
mined. For example, the weight of aeb bounded by the circular boundary is established
as follows. Using co-ordinates 0 measured from a along ae, and from ae to the
boundary on which 0 = , the radii of compression members and corresponding force
curves are
R(0, ) = 2H sin(0 + ), (55)
R

(}4 , }4 0) = 2F

cos(}4 + }4 0) = 2F

sin(0 + ). (56)
The weight of region aeb is thus
W
aeb
= 2(4HF

}4
0

0
sin
2
(0 + ) d0 d =HF

2
}8, (57)
where the multiplier 2 before the integral accounts also for the tension members. In a
similar manner it can be shown that the volumes of aef and efhg are given by
W
ae[
=HF

2
}8 and W
efgh
=HF

2
}4. (58)
The distance between points e and h in Fig. 12(a), equal to H}2, contains weight
of material equivalent to the sum of W
aeq
, W
bef
, and W
efgh
. Thus the weight per unit
length of the cycloid framework is F

. This does not agree with the value F

}2
given by Hemp (1973). The dierence results from the manner in which Hemp chose
to present the results for comparison with other truss designs. Fig. 14 shows one section
of a Warren truss cantilever and one section of a cycloid cantilever, both with depth
H. Assume that both cantilevers have length nH, and that cross sections ab in Fig. 14
are distance mH from the root of the cantilevers. The force carried in member ab of
the Warren truss is thus (n m)F

, and bracing members ac and bd both carry force


magnitude F

2. Thus, multiplying these forces by the lengths of the members, the


volume of section abcd is
W
warren
= 2(n m)F

! + 2F

!, (59)
where ! is the length of the section equal to H.
P. Dewhurst / J. Mech. Phys. Solids 49 (2001) 445 467
465
From Eqs. (34), (44) and (46), the force in the anges of a cycloid cantilever vary
linearly with multiplying coecient F

}H. The force in member ab of Fig. 14(b) is


(n m)F

at a, and decreases linearly by amount (}2)F

over the length !, equal to


(}2)H, from a to b. The weight of section abcd of the cycloid cantilever is thus
W
cycloid
= 2(n m }4)F

! + F

!, (60)
where as in Eq. (59), the rst term represents the weight of the anges and the second
term the weight of the internal bracing. Eq. (60) can be rearranged as
W
cycloid
= 2(n m)F

! + (}2)F

!. (61)
This is the weight representation used by Hemp. The problem arises when the weight
reduction, caused by the tip couple, is applied to the bracing rather than to the anges;
giving the erroneous impression that the last term in Eq. (61) represents the weight of
the bracing.
It should be noted that Eq. (58) is only valid at a sucient distance from the tip,
where the force diagram is also a cycloid structure. Commencing from the tip, the force
diagram starts with two centered fans as in Fig. 10, representing the force discontinuity
at the anges. Thus, numerical methods are required for a full solution of a cycloid
cantilever.
7. Discussion
Power-series solutions and matrix operator methods, developed for plane-strain slip-
line eld theory, have been applied to the generation of minimum-weight structures. The
procedures are intended to form the basis for a numerical analysis method to be used
to investigate minimum-weight solutions for a wider range of boundary conditions. The
intention of the work is to show how frames can be generated region by region, and
how the combination of space and force diagrams can be used for weight calculations.
In the present work attention was restricted to known solution types, for which ana-
lytical solutions for the structure weight could be obtained from the series method. The
intention was to validate the procedures against known solutions. As part of this work
the formal conditions for a force-free boundary have been established, and the solution
for a straight boundary shown to be exactly the same as for a perfectly rough tool
surface in slip-line eld theory. A future extension of this work will be to investigate
the possible form of curved force-free boundaries, using the general boundary operator
developed by Dewhurst (1985).
With the restriction to analytical solutions, no attention was given in the paper to
the determination of the co-ordinate geometry of the frameworks. This can readily be
accomplished numerically using the series or matrix methods; see Ewing (1967) and
Dewhurst and Collins (1973). Also, except in reference to the work by Chan (1960),
consideration has not been given to the calculation of displacements of Michell struc-
tures under applied loads. Johnson et al. (1971) showed, in the appendix to their paper,
that the calculation of displacements along a Michell structural member is identical to
the calculation of force transmitted across a slip-line in plane-strain metal deformation.
466
P. Dewhurst / J. Mech. Phys. Solids 49 (2001) 445 467
Numerical evaluations of displacement can thus be made, following the procedure for
calculating slip-line forces, given by Dewhurst and Collins (1973).
In practice, the continuum framework solutions must be approximated by networks
of closely arranged discrete elements. Discrete network arrangements, of this type,
can readily be evaluated by the series or matrix methods, to obtain geometry, force,
and corresponding weight solutions. It appears that nature may already have perfected
such discrete structures. In a fascinating review of the study of families of curved
load-bearing dense plates in human bone, Newell (1997) presents gures of the human
femur that illustrate a powerful analogy with Michell frameworks. He quotes Bell
(1834, p. 31) as stating that a human bone may be taken to prove that in Natures
work strength is given with the least possible expense of material.
Acknowledgements
The author would like to thank Sriruk Srithongchai, Ph.D. candidate at the University
of Rhode Island (URI), for the development of the solid model illustrated in Fig. 5.
The model has been used to make test samples in the Rapid Manufacturing Center at
URI.
References
Alexandrov, S., 1999. Steady planar streamlined plastic ows: general and special analytical solutions.
In: Rimrott, F.P.J., Schwaighofer, J. (Eds.), The Integration of Material, Process and Product Design,
Balkema, Rotterdam, pp. 103110.
Bell, C., 1834. Cited by Newell, 1997, p. 31.
Chan, H.S.L., 1960. The design of Michell optimum structures. Report No. 142, Craneld College of
Aeronautics, UK.
Chan, H.S.Y., 1963. Optimum Michell frameworks for three parallel forces. Report No. 167, Craneld
College of Aeronautics, UK.
Collins, I.F., 1968. The algebraic geometry of slip-line elds with applications to boundary value problems.
Proc. Roy. Soc. Ser. A 303, 317338.
Dewhurst, P., 1978. On the non-uniqueness of the machining process. Proc. Roy. Soc. Ser. A 360, 587610.
Dewhurst, P., 1985. A general matrix operator for linear boundary value problems in slip-line eld theory.
Int. J. Numer. Methods Engng. 21, 169182.
Dewhurst, P., Collins, I.F., 1973. A matrix technique for constructing slip-line eld solutions to a class of
plane strain plasticity problems. Int. J. Numer. Methods Engng. 7, 357378.
Ewing, D.J.F., 1967. A series-method for constructing plastic slipline elds. J. Mech. Phys. Solids 15,
105114.
French, M.J., 1999. A Second Law for structures: an insightful approach to the design of plane frames.
ASME J. Appl. Mech. 66, 738741.
Green, A.P., 1954. On the use of hodographs in problems of plane plastic strain. J. Mech. Phys. Solids 16,
267276.
Hegemier, G.A., Prager, W., 1969. On Michell trusses. Int. J. Mech. Sci. 11, 209215.
Hemp, W.S., 1973. Optimum Structures. Clarendon Press, Oxford.
Hill, R., 1950. The Mathematical Theory of Plasticity. Clarendon Press, Oxford.
Hill, R., Lee, E.H., Tupper, C.L., 1951. A method of numerical analysis of plastic ow in plane strain and
its application to the compression of a ductile material between rough plates. Trans. ASME, J. Appl.
Mech. 18, 4652.
P. Dewhurst / J. Mech. Phys. Solids 49 (2001) 445 467
467
Johnson, W., 1961. An analogy between upper-bound solutions for plane-strain metal working and
minimum-weight two-dimensional frames. Int. J. Mech. Sci. 3, 239246.
Johnson, W., Chitkara, N.R., Reid, S.R., Collins, I.F., 1971. The displacement eld and its signicance for
certain minimum-weight two-dimensional frames using the analogy with perfectly plastic ow in metal
working. Int. J. Mech. Sci. 13, 547561.
Kachanov, L.M., 1973. Fundamentals of the Theory of Plasticity. Mir Publishers, Moscow.
Kozlowski, W., Mroz, Z., 1970. Optimal design of disks subject to geometric constraints. Int. J. Mech. Sci.
12, 10071021.
Michell, A.G.M., 1904. Limits of economy of material in frame-structures. Philos. Mag. 6 (847), 589597.
Newell, R.L.M., 1997. The Calcar femorale: a tale of historical neglect. Clin. Anat. 10, 2733.
Prandtl, L., 1921. Practical application of the Hencky equation to plastic equilibrium. Z. Angew. Math. Mech.
3, 401413.
Richmond, O., 1968. Theory of Streamlined Dies for Drawing and Extrusion. Mechanics of the Solid State,
University of Toronto Press, 154167.
Richmond, O., Devenpeck, M.L., 1962. A die prole for maximum eciency in strip drawing. Proceedings
of the fourth ASME US Congress on Applied Mechanics, University of California, Berkeley.

Вам также может понравиться