6

Constructing
MoM Members

61

62

Chapter 6: CONSTRUCTING MOM MEMBERS

TABLE OF CONTENTS
Page

6.1.

FORMULATION OF MOM MEMBERS

6.1.1. What They Look Like
. . . . . . . .
6.1.2. End Quantities, Degrees of Freedom, Joint Forces
6.1.3. Internal Quantities . . . . . . . . . .
6.1.4. Discrete Field Equations, Tonti Diagram . . .
6.2. SIMPLEX MOM MEMBERS
6.2.1. The Truss Element Revisited
. . . . . .
6.2.2. The Spar Element
. . . . . . . . . .
6.3. NON-SIMPLEX MOM MEMBERS
6.
Exercises . . . . . . . . . . . . . . . .

62

. .
.
. .
.

.
. .
.
. .

. .
.
. .
.

.
.
.
.

. . . . . .
. . . . .
. . . . . .

63
63
64
64
65
66
67
68
69
611

63

6.1

FORMULATION OF MOM MEMBERS

The truss member used as example in Chapters 24 is an instance of a structural element. Such elements may be formulated directly using concepts and modeling techniques developed in Mechanics
of Materials (MoM).1 The construction does not involve the more advanced tools that are required
for the continuum nite elements that appear in Part II.
This Chapter presents an overview of the technique to construct the element stiffness equations of
MoM members using simple matrix operations. These simplied equations come in handy for
a surprisingly large number of applications, particularly in skeletal structures. Focus is on simplex
elements, which may be formed directly as a sequence of matrix operations. Non-simplex elements
are presented as a recipe, since their proper formulation requires work theorems not yet studied.
The physical interpretation of the FEM is still emphasized. Consequently we continue to speak of
structures built up of members connected at joints.
6.1. FORMULATION OF MOM MEMBERS
6.1.1. What They Look Like
MoM-based formulations are largely restricted to intrinsically one-dimensional members. These are
structural components one of whose dimensions, called the longitudinal dimension, is signicantly
larger than the other two, which are called the tranverse dimensions. Such members are amenable
to the simplied structural theories developed in MoM textbooks. We shall study only straight
members with geometry dened by the two end joints. The member cross sections are dened by
the intersection of planes normal to the longitudinal dimension with the member. See Figure 6.1.
Note that although the individual member will be idealized as being one-dimensional in its intrinsic
or local coordinate system, it is generally part of a two- or three-dimensional structure.
This class of structural components embodies bars, beams, beam-columns, shafts and spars. Although geometrically similar, the names distinguish the main kind of internal forces the member
resists and transmits: axial forces for bars, bending and shear forces for beams, axial compression
and bending for beam-columns, torsion forces for shafts, and shear forces for spars.
The members are connected at their end joints by displacement degrees of freedom. For truss (bar)
members those freedoms are the translational components of the joint displacements. For other
types, notably beams and shafts, nodal rotations are chosen as additional degrees of freedom.
The structures fabricated with these kinds of members are generally three-dimensional. Their
geometry is dened with respect to a global Cartesian coordinate system {x, y, z}. Two-dimensional
idealizations are useful simplications in cases where the nature of the geometry and loading allows
the reduction of the structural model to one plane of symmetry, which is chosen to be the {x, y}
plane. Plane trusses and plane frameworks are examples of such simplications.
In this Chapter we study generic structural members that t the preceding class. An individual
member is identied by (e) but this superscript will be usually suppressed in the equations below
to reduce clutter. The local axes are denoted by {x,
y , z }, with x along the longitudinal direction.
See Figure 6.1.
1

Mechanics of Materials was called Strength of Materials in older texts. The scope of this subject includes bars, beams,
shafts, arches, thin plates and shells, but only one-dimensional models are covered in basic undergraduate courses. MoM
involves ab initio kinematic assumptions such as plane sections remain plane. An identical acronym is used in Electrical
Engineering for something completely different: the Method of Moments.

63

64

Chapter 6: CONSTRUCTING MOM MEMBERS

Figure 6.1. A Mechanics of Materials (MoM) member is a structural element one of

whose dimensions (the longitudinal dimension) is signicantly larger than

the other two. Local axes {x,
y , z } are chosen as indicated. Although the
depicted member is prismatic, some applications utilize tapered or stepped
members, the cross section of which varies as a function of x.

The mathematical model of a MoM member is obtained by an idealization process. The model
represents the member as a line segment that connects the two end joints, as depicted in Figure 6.2.
6.1.2. End Quantities, Degrees of Freedom, Joint Forces
The set of mathematical variables used to interconnect members are called end quantities or connectors. In the Direct Stiffness Method (DSM) these are joint displacements (the degrees of freedom)
and the joint forces. These quantities are linked by the member stiffness equations.
The degrees of freedoms selected at the end joints i and j are collected in the joint displacement
This may include translations only, or a combination of translations and rotations.
vector u.
Component
The vector of joint forces f collects components in one to one correspondence with u.
pairs must be conjugate in the sense of the Principle of Virtual Work. For example if the x-translation

the corresponding entry in f is the x-force

at joint i: u xi appears in u,

f xi at i. If the rotation about

the corresponding entry in f is the z-moment m z j .
z at joint j: z j appears in u,
6.1.3. Internal Quantities
Internal quantities are mechanical actions that take place inside the member. Those actions involve
stresses and deformations. Accordingly two types of internal quantities appear:
Internal member forces form a nite set of stress-resultant quantities collected in an array p. This set
globally characterizes the forces resisted by the material. They are also called generalized stresses
in structural theory. Stresses at any point in the member may be recovered if p is known.
Member deformations form a nite set of quantities, chosen in one-to one correspondence with
internal member forces, and collected in an array v. This set characterizes the deformations experienced by the material. They are also called generalized strains in structural theory. Strains at any
point in the member can be recovered if v is known.
As in the case of end quantities, internal forces and deformations are paired in one to one correspondence. For example, the axial force in a bar member must be paired either with an average
axial deformation, or with the total elongation. Pairs that mutually correspond in the sense of the
Principle of Virtual Work are called conjugate. Unlike the case of end quantities, conjugacy is not
64

65

6.1

FORMULATION OF MOM MEMBERS

End quantities (connectors)

are defined at the joints i, j

(e)

i
y

Internal quantities are

defined within the member
and may depend on x.

Figure 6.2. The FE mathematical idealization of a MoM member. The model is

one-dimensional in x.
The two end joints are the site of end quantities:
joint forces and displacements, that interconnect members. The internal
quantities characterize the stresses and deformations in the member.

a mandatory requirement but simplies some derivations.

6.1.4. Discrete Field Equations, Tonti Diagram
v, p and f are called the discrete eld equations. There are
The matrix equations that connect u,
three of them.
The member deformations v are linked to the joint displacements u by the kinematic compability
conditions, also called the deformation-displacement or strain-displacement equations:

v = Bu.

(6.1)

The internal member forces are linked to the member deformations by the constitutive equations.
In the absence of initial strain effects those equations are homogeneous:
p = Sv.

(6.2)

Finally, the internal member forces are linked to the joint forces by the equilibrium equations. If
the internal forces p are constant over the member, the relation is simply
f = AT p.
where the transpose of A is used for convenience.2
2

If p is a function of x the relation is of differential type and is studied in 6.3.

65

(6.3)

66

Chapter 6: CONSTRUCTING MOM MEMBERS

u
f = AT S B u = K

f = AT p

Kinematic v = B u
p=Sv

Equilibrium

Constitutive
Figure 6.3. Tonti diagram of the three discrete eld equations (6.1)(6.3) and the

stiffness equation (6.4) for a simplex MoM element. Internal and end
quantities appear inside the orange and yellow boxes, respectively.

These equations can be presented graphically as shown in Figure 6.3. This is a discrete variant
of the so-called Tonti diagrams, which represent the governing equations as arrows linking boxes
containing kinematic and static quantities.
Matrices B, S and A receive the following names in the literature:
A

Equilibrium

Rigidity, material, constitutive3

Compatibility, deformation-displacement, strain-displacement

If the element is sufciently simple, the determination of these three matrices can be carried out
through MoM techniques. If the construction requires more advanced tools, however, recourse to
the general methodology of nite elements and variational principles is necessary.
6.2. SIMPLEX MOM MEMBERS
In this section we assume that the internal quantities are constant over the member length. Such
members are called simplex elements. If so the matrices A, B and S are independent of member
cross section, and the derivation of the element stiffness equations is particularly simple.
Under the constancy assumption, elimination of the interior quantities p and v from (6.1)-(6.3)
yields the element stiffness relation
u.
f = AT SB u = K

(6.4)

= AT SB.
K

(6.5)

Hence the element stiffness matrix is

The four preceding matrix equations are diagrammed in Figure 6.3.

3

The name rigidity matrix for S is preferable. It is a member integrated version of the cross section constitutive equations.
The latter are usually denoted by symbol R, as in 6.3.

66

67

6.2
(a)

z
y

Axial rigidity E A, length L

y
(b)

SIMPLEX MOM MEMBERS

EA

fxi , u xi i

j fx j , u x j

L
Figure 6.4. The prismatic truss (also called bar) member: (a) individual

member in 3D space, (b) idealization as generic member.

are conjugate in the sense of the Principle of Virtual Work, it can be shown that
If {p, v} and {f, u}
A = B and that S is symmetric. Then
= BT SB.
K

(6.6)

is a symmetric matrix. Symmetry is computationally desirable for reasons outlined in Part III.
REMARK 6.1

must come out to be symmetric even if S is unsymmetric and

If f and u are conjugate but p and v are not, K
A = B. However there are more opportunities to go wrong.

6.2.1. The Truss Element Revisited

The simplest example of a MoM element is the prismatic truss (bar) element already derived in
Chapter 2. See Figure 6.4. This qualies as a simplex MoM element since all internal quantities are
constant. One minor difference in the derivation below is that the joint displacements and forces
in the y direction are omitted in the generic element since they contribute nothing to the stiffness
equations. In the FEM terminology, freedoms associated with zero stiffness are called inactive.
Three choices for internal deformation and force variables are studied below. They illustrate that
the resulting element stiffness equations coalesce, as can be expected, since the external quantities
are the same.
Derivation Using Axial Elongation and Axial Force. The member axial elongation d is taken as
deformation measure, and the member axial force F as internal force measure. Hence v and p
reduce to the scalars v = d and p = F, respectively. The chain of discrete eld equations is easily
67

68

Chapter 6: CONSTRUCTING MOM MEMBERS

(a)

Shear rigidity G As , length L

z
x

x
y
fyi , u yi

(b)

fy j , u y j

G As

L
Figure 6.5. The prismatic spar (also called shear-web) member: (a) individual

member in 3D space, (b) idealization as generic member.

constructed:
d = [ 1

u
1 ] xi
u x j

= Bu,

EA
d = Sd,

L  
xi
f
1
f =
=
F = AT F
1
fx j

F=

Hence

= AT SB = SBT B = E A
K
L

1
1


1
.
1

(6.7)

(6.8)

Note that A = B because F and d are conjugate.

Derivation Using Mean Axial Strain and Axial Force. Instead of d we may use the mean axial strain
e = d/L as deformation measure whereas F is kept as internal force measure. The only change
is that B becomes [ 1 1 ] /L whereas S becomes E A. Matrix A does not change. The product
as in (6.8), as can be expected. Now AT is not equal to B because F and
AT SB gives the same K
e are not conjugate, but they differ only by a factor 1/L.
Derivation Using Mean Axial Strain and Axial Stress. We keep the mean axial strain e = d/L as
deformation measure but the mean axial stress = F/A, (which is not conjugate to e)
is taken as
T
internal force measure. Now B = [ 1 1 ] /L, S = E and A = A [ 1 1 ]. The product AT SB
shown in (6.8).
gives again the same K
6.2.2. The Spar Element
The spar or shear-web member has two joints, i and j. It can only resist and transmit a constant
shear force V in the plane of the web, which is chosen to be the {x,
y } plane. See Figure 6.5. This
element is often used in modeling high-aspect aircraft wing structures, as illustrated in Figure 6.6.
68

69

6.3

NON-SIMPLEX MOM MEMBERS

The active degrees of freedom for the generic element of length L depicted in Figure 6.5(b) are
u yi and u y j . Let G be the shear modulus and As the effective shear area.4 The shear rigidity is
G As . As deformation measure the mean shear strain = V /(G As ) is chosen. The deformationdisplacement, constitutive, and equilibrium equations are


1
u yi

= Bu,
= [ 1 1 ]
u y j
L
V = G As = S ,
 



f = f yi = 1 V = AT V.
1
fy j

(6.9)

Therefore the member stiffness equations are

f =

fyi
fy j

G As
= A SBu =
L
T

1
1

1
1



u yi
u y j

u.

=K

(6.10)

Note that A = B as V and are not conjugate.

However, the difference is easily adjusted for. See
Exercise 6.2.
The equations (6.10) may be augmented with zero
rows and columns to bring in the node displacements in the x and z directions, and then referred
to the global axis {x, y, z} through a congruential
transformation.

COVER
PLATES
SPAR

6.3. NON-SIMPLEX MOM MEMBERS

RIB

In the general case, the internal quantities p and

v may vary over the member; that is, depend on
x.
This dependence may be due to element type,
variable cross section, or both. Consequently one
or more of the matrices A, B and S, depend on x.

Such elements are called non-simplex.

Figure 6.6. Spar members in aircraft wing (Piper

Cherokee). For more impressive

aircraft structures see CATECS slides.

If the element falls under this category, the straightforward matrix multiplication recipe (6.5) cannot
This fact can be grasped by observing that
be used to construct the element stiffness matrix K.
T

x)
would depend on x.
On the other hand, K must be independent of x because it
A(x)
S(x)B(
relates the end quantities u and f.
The derivation of non-simplex MoM elements requires the use of work principles of mechanics,
for example the Principle of Virtual Work or PVW. More care must be exercised in the selection of
conjugate internal quantities. The following rules can be justied through the variational arguments
discussed in Part II. They are stated here only as recipe.
4

A concept developed in Mechanics of Materials. For a narrow rectangular cross section, As = 5A/6.

69

610

Chapter 6: CONSTRUCTING MOM MEMBERS

Kinematic
(at each section)

f =

BT R B dx u

v = B u

d f = BT dp
p=Rv

Equilibrium

Constitutive
(at each section)
Figure 6.7. Diagram of the discrete eld equations for a non-simplex MoM element.

Rule 1. Select internal deformations v(x)

and internal forces p(x)
that are conjugate in the PVW
sense. Link deformations to node displacements by v(x)
= B(x)u.

Rule 2. From the PVW it may be shown5 that the force equilibrium equation exists only in a
6
differential sense: BT dp = d f. Here d denotes differentiation with respect to x.
Rule 3. The constitutive matrix R(x)
that relates p(x)
to v(x)
must be symmetric.7
Relations that emanate from these rules are diagrammed in Figure 6.7. Internal quantities are now
eliminated using the differential equilibrium relation:

d f = BT dp = BT p d x = BT R v d x = BT RBu d x = BT RB d x u.
Integrating both sides over the member length L yields
 L
 L

u,
f =

df =
BT RB d x u = K
0

(6.11)

(6.12)

because u does not depend on x.

Consequently the local element stiffness matrix is
=
K

BT R B d x

(6.13)

The result (6.13) will be justied in Part II of the course through energy methods. It will be seen
there that this formula applies to arbitrary displacement-assumed nite elements in any number of
dimensions. It is applied to the derivation of the stiffness equations of the plane beam element in
Chapter 13.
5

T
= (BT .dp)T . u.

The proof is done by equating the virtual work of a slice of length d x:

d f . u = dpT .v = dpT .(B u)
Since u is arbitrary, BT dp = d f.

The meaning of dp is simply p(x)

d x.
That is, the differential of internal forces as one passes from cross-section x to a
neighboring one x + d x.
The interpretation of d f is less immediate because f is not a function of x.
It actually means
the contribution of that member slice to the building of the node force vector f. See Equation (6.12).

Note that symbol R replaces the S of the previous section. R applies to a specic cross section, S to the entire member.
See Exercise 6.9.

610

611

Exercises

Homework Exercises for Chapter 6

Constructing MoM Members
EXERCISE 6.1

[A:15] Formulate a prismatic shaft member of length L that can only transmit a torque T along the longitudinal
(x)
direction. The only active degrees of freedom of this member are the twist-angle rotations xi and x j of the
end joints about x.
The constitutive equation furnished by Mechanics of Materials is T = (G J/L), where
G J is the torsional rigidity and = x j xi is the relative twist angle. Draw the Tonti diagram to illustrate
the discrete equations.

The result should be

in which m xi = T and m x j

 GJ 
 
xi
m xi
1 1
=
m x j
x j
1
L 1
= T are the nodal moments about x.

(E6.1)

EXERCISE 6.2

[A:10] Explain how to select the deformation variable v (paired to V ) of the spar member formulated in
6.2.2, so that A = B.
EXERCISE 6.3

[A/C:25] Derive the 4 4 global element stiffness matrix of a prismatic spar element in a two dimensional
Cartesian system {x, y}. Start from the local stiffness (6.10) and proceed as in 3.2.1 and 3.2.2 for the bar
in (6.10) is 2 2, T is 2 4. Show that T consists of rows 2 and 4 of the matrix of (3.2).
element. Since K
j

(xj , yj , z j )

(a)

E, A

uyi

u xj

(b)
j

i (x i , yi , z i )

u xi

y
L

Figure E6.1. Bar element in 3D for Exercise 6.4.

EXERCISE 6.4

[A+N:15] A bar element moving in three dimensional space is dened by the global coordinates {xi , yi , z i },
{x j , y j , z j } of its end nodes i and j, as illustrated in Figure E6.1. The 26 displacement transformation matrix
T relates u (e) = Tu(e) . Here u (e) contains the local axial displacements u xi and u x j whereas u(e) contains the
global displacements u xi , u yi , u zi , u x j , u y j , u z j . Show that
T=

1  x ji
L 0

y ji
0

z ji
0

0
x ji

0
y ji

0
z ji

cx ji c y ji cz ji 0 0 0
0 0 0 cx ji c y ji cz ji

(E6.2)

in which L is the element length, x ji = x j xi , etc., and cx ji = x ji /L, etc., are the direction cosines of the
vector going from i to j. Evaluate T for a bar going from node i at {1, 2, 3} to node j at {3, 8, 6}.

611

612

Chapter 6: CONSTRUCTING MOM MEMBERS

k (xk , yk , zk )

Orientation node k defines plane

as passing through {i,j,k}
{x,y}

u yj

(a)

z
i

i (x i , yi , z i )
y

(b)

G, As

uyi

(xj , yj , z j )

Lc

x
Figure E6.2. Spar element in 3D for Exercise 6.5.

EXERCISE 6.5

[A+N:25 (10+15)] A spar element in three dimensional space is only partially dened by the global coordinates
{xi , yi , z i }, {x j , y j , z j } of its end nodes i and j, as illustrated in Figure E6.2. The problem is that axis y , which
denes the direction of shear force transmission, is not uniquely dened by i and j.8 Most FEM programs use
the orientation node method to complete the denition. A third node k, not colinear with i and j, is provided
by the user. Nodes {i, j, k} dene the {x,
y } plane and consequently z . The projection of k on line i j is point m
and the distance h > 0 from m to k is called h as shown in Figure E6.2. The 2 6 displacement transformation
matrix T relates u (e) = Tu(e) . Here u (e) contains the local transverse displacements u yi and u y j whereas u(e)
contains the global displacements u xi , u yi , u zi , u x j , u y j , u z j .
(a)

Show that
T=

1  xkm
h 0

ykm
0

z km
0

0
xkm

0
ykm

z km

cxkm c ykm czkm 0

0
0
0
0
0 cxkm c ykm czkm

(E6.3)

in which xkm = xk xm , etc., and cxkm = xkm / h, etc., are the direction cosines of the vector going from
m to k.
(b)

Work out the formulas to compute the coordinates of point m in terms of the coordinates of {i, j, k}.
Using
the notation of Figure E6.2(b) and elementary trigonometry, show that h = 2A/L, where A =

p( p a)( p b)( p L) with p = 12 (L + a + b) (Herons formula), cos = (L 2 + b2 a 2 )/(2bL),
cos = (L 2 + a 2 b2 )/(2a L), c = b cos , L c = a cos , xm = xi (L c)/L + x j c/L, etc. Evaluate
T for a spar member going from node i at {1, 2, 3} to node j at {3, 8, 6}. with k at {4, 5, 6}.

EXERCISE 6.6

[A:15] If the matrices B and R are constant over the element length L, show that expression (6.13) of the
element stiffness matrix for a variable-section element reduces to (6.6), in which S = LR.
EXERCISE 6.7

[A:25] Explain in detail the quick derivation of footnote 5. (Knowledge of the Principle of Virtual Work is
required to do this exercise.)
8

This full specication of the local system is also required for 3D beam elements.

612

613

Exercises

EXERCISE 6.8

[A:20] Explain how thermal effects can be generally incorporated in the constitutive equation (6.2) to produce
an initial force vector.
EXERCISE 6.9

[A:20] Consider a non-simplex MoM element in which R varies with x but B is constant. From (6.13) show
that
 L
1
T

R(x)
d x
(E6.4)
K = LB R B, with R =
L 0
is an element-averaged constitutive matrix. Apply this result to form K
for a truss member of tapered
Here R

+ A j x/L,

where
cross section area A(x).
Assume that the area is dened by the linear law A = Ai (1 x/L)
Ai and A j are the end areas at joints i and j, respectively. Take B to be the same as in the prismatic case
discussed in 6.2.1.
EXERCISE 6.10

[A/C+N:30] A prismatic bar element in 3D space is referred to a global coordinate system {x, y, z}, as in
Figure E6.1. The end nodes are located at {x1 , y1 , z 1 } and {x2 , y2 , z 2 }.9 The elastic modulus E and the
cross section
 area A are constant along the length. Denote x21 = x2 x1 , y21 = y2 y1 , z 21 = z 2 z 1
2
2
2
and L = x21
+ y21
+ z 21
. Show that the element stiffness matrix in global coordinates can be compactly
10
written
EA
K(e) = 3 BT B
(E6.5)
L
where
(E6.6)
B = [ x21 y21 z 21 x21 y21 z 21 ]
Compute K(e) if the nodes are at {1, 2, 3} and {3, 8, 6}, with elastic modulus E = 343 and cross section area
A = 1. Note: the computation can be either done by hand or with the help of a program such as the following
Mathematica module, which is used in Part III of the course:
Stiffness3DBar[ncoor_,mprop_,fprop_,opt_]:= Module[
{x1,x2,y1,y2,z1,z2,x21,y21,z21,Em,Gm,rho,alpha,A,
num,L,LL,LLL,B,Ke}, {{x1,y1,z1},{x2,y2,z2}}=ncoor;
{x21,y21,z21}={x2-x1,y2-y1,z2-z1};
{Em,Gm,rho,alpha}=mprop; {A}=fprop; {num}=opt;
9
10

End nodes are labeled 1 and 2 instead of i and j to agree with the code listed below.
There are several ways of arriving at this result. Some are faster and more elegant than others. Here is a sketch of one of
the ways. Denote by L 0 and L the lengths of the bar in the undeformed and deformed congurations, respectively. Then
2
1
2 (L

L 20 ) = x21 (u x2 u x1 ) + y21 (u y2 u y1 ) + z 21 (u z2 u z1 ) + R Bu

in which R is a quadratic function of node displacements which is therefore dropped in the small-displacement linear
theory. But
2
2
1
1
2 (L L 0 ) = 2 (L + L 0 )(L L 0 ) L L
also because of small displacements. Hence the small axial strain is e = L/L = (1/L 2 )Bu(e) , which begins the Tonti
diagram. Next is F = E A e, and nally you should show that force equilibrium at nodes requires f(e) = (1/L)BT F.
Multiplying through you get (E6.5). Another way is to start from the local stiffness (6.8) and transform to global using
the transformation matrix (E6.2).

613

Chapter 6: CONSTRUCTING MOM MEMBERS

If [num,{x21,y21,z21,Em,A}=N[{x21,y21,z21,Em,A}]];
LL=x21^2+y21^2+z21^2; L=PowerExpand[Sqrt[LL]];
LLL=Simplify[LL*L]; B={{-x21,-y21,-z21,x21,y21,z21}};
Ke=(Em*A/LLL)*Transpose[B].B;
Return[Ke]];
ClearAll[Em,A]; Em=343; A=1;
ncoor={{0,0,0},{2,6,3}}; mprop={Em,0,0,0}; fprop={A}; opt={False};
Ke=Stiffness3DBar[ncoor,mprop,fprop,opt];
Print["Stiffness of 3D Bar Element:"];
Print[Ke//MatrixForm];
Print["eigs of Ke: ",Eigenvalues[Ke]];
As a check, the six eigenvalues of this particular K(e) should be 98 and ve zeros.

614

614