Upload
buixuyen
View
223
Download
3
Embed Size (px)
Citation preview
CHAPTER 5
Matrix
Displacement Method
5.1 INTRODUCTION
In the last half-century, considerable progress has been made in the matrix analysis of structures. The topic has been generalized to finite elements, and extended to the stability, non-linear and dynamic analysis of structures. This progress is due to the simplicity, modularity and flexibility of matrix methods.
Many textbooks covering these methods have been published including Argyris [4], McGuire and Gallagher [172], Livesley [160], Meek [173], Kardestuncer [88], ad Vanderbilt [242] among many others. In these books the displacement method of structural analysis is thoroughly developed, and therefore only a brief introduction will be presented here.
5.2 FORMULATION
In Chapter 4, the network-topological formulation of the displacement (stiffness) method of structural analysis has already been presented. In this section, a matrix formulation using the basic tools of structural analysis - equilibrium of forces, compatibility of displacements, and force-displacement relationships - is provided. The notations are chosen from the most popularly encountered versions in structural mechanics.
124 Structural Mechanics: Graph and Matrix Methods
124
Consider a structure S with M members and N nodes; each node having one degree of freedom. The kinematical indeterminacy of S may then be determined as,
β−α=η N)S( , (5-1)
where β is the number of constraints due to the support conditions. As an example, )S(η for the planar truss S depicted in Figure 5.1(a) is given by )S(η = 2×5 − 3 =
7, and for the space frame shown in Figure 5.1(b), it is calculated as )S(η = 6×8 −6×4 = 24.
6
6 6
62 2
210
(a) A planar truss. (b) A space frame.
Fig. 5.1 The degrees of freedom of the joints for two structures.
One can also calculate η(S) by simple addition of the degrees of freedom of the joints of the structure, i.e. for the truss S, η(S) = 2 + 2 + 2 + 1= 7, and for the frame η(S) = 4×6 = 24.
Let p and v represent the joint loads and joint displacements of a structure. Then the force-displacement relationship for the structure can be expressed as,
p = Kv, (5-2)
where K is a ηN×ηN symmetric matrix, known as the stiffness matrix of the structure. Expanding the ith equation of the above system, the force pi can be expressed in terms of the displacements },...,,{ N21 αvvv as:
.... NNi22i11ii αα+++= vKvKvKp (5-3)
A typical coefficient Kij is the value of the force pi required to be applied at the ith component of the structure, in order to produce a displacement vj=1 at j and zero displacements at all the other components.
CHAPTER 5 Matrix Displacement Method 125
125
As has been shown in Chapter 4, the member forces r can be related to nodal forces p by:
p = Br. (5-4)
Similarly, the joint displacements v can be related to member distortions u by:
u = Btv. (5-5)
For each individual member of the structure, the member forces can be related to member distortions by an element stiffness matrix km. A block diagonal matrix containing these element stiffness matrices is known as the unassembled stiffness matrix of the structure, denoted by k. Obviously:
r = ku. (5-6)
This equation together with Eqs (5-4) and (5-5) yields:
p = BkBtv. (5-7)
Therefore,
K = BkBt (5-8)
is obtained. The matrix K is singular since the boundary conditions of the structure are not yet applied. For an appropriately supported structure, the deletion of the rows and columns of K corresponding to the support constraints results in a positive definite matrix, known as the reduced stiffness matrix of the structure.
Let us illustrate the method by means of a simple example. Consider a fixed end beam with a load P applied at its mid span. This beam is discretized as two beam elements, as shown in Figure 5.2(a). The components of element forces and element distortions are depicted in Figure 5.2(b) and those of the entire structure are illustrated in Figure 5.2(c).
L L
P
21 3
1 2
(a) A fixed ended beam S.
126 Structural Mechanics: Graph and Matrix Methods
126
2 2
5 5
r ,u r ,u r ,u r ,u
r ,u r ,u r ,u r ,u1 1 3 3
4 4 6 61 2
7 7
8 8
(b) Member forces and member distortions.
3 3
4 4 6 6
p ,v1 1
p ,v p ,v p ,v
p ,v p ,v
2 2
5 5
1 2
(c) Nodal forces and nodal displacements of the entire structure.
Fig. 5.2 Illustration of the analysis of simple structure.
For each element such as element 1, the element stiffness matrix can be written as,
⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢
⎣
⎡
⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢
⎣
⎡
=
⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢
⎣
⎡
4
3
2
1
44434241
34333231
24232221
14131211
4
3
2
1
uuuu
kkkkkkkkkkkkkkkk
rrrr
(5-9)
and for the entire structure we have:
⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
=
⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
6
5
4
3
2
1
666564636261
565554535251
464544434241
363534333231
262524232221
161514131211
6
5
4
3
2
1
vvvvvv
KKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKKK
pppppp
(5-10)
Element stiffness matrices k1 and k2 can be easily constructed using the definition of kij. For a beam element, ignoring its axial deformation, these terms are shown in Figure 5.3. The structure has a uniform cross section and since both elements have the same length:
CHAPTER 5 Matrix Displacement Method 127
127
u =1
u =1
u =1
u =1
k
k k k
kk
k
k
k k k k
kkkk11 12
23
33
22
44
34
31
41
21
42
32
14
24
43
13
1
2
34
Fig. 5.3 The stiffness coefficients of a beam element ignoring its axial deformation.
⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢
⎣
⎡
−−−
−−−
==
2L/31L/3L/3L/6L/3L/6
1L/32L/3L/3L/6L/3L/6
LEI2
22
22
21 kk . (5-11)
The unassembled stiffness matrix is an 8×8 matrix of the form k:
⎥⎦
⎤⎢⎣
⎡=
2
1k00k
k . (5-12)
Now consider the equilibrium of the joints of the structure, resulting in,
p1 = r1 , p2 = r2 , p3 = r5 + r3,
p4 = r4 + r6 , p5 = r7 , p6 = r8 . (5-13)
or in a matrix form we have,
128 Structural Mechanics: Graph and Matrix Methods
128
⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅
=
⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
8
7
6
5
4
3
2
1
6
5
4
3
2
1
rrrrrrrr
11
1111
11
pppppp
, (5-14)
and more compactly:
p = Br. (5-15)
Consider now the compatibility of displacements as:
u1 = v1 , u2 = v2 , u3 = u5 = v3,
u4 = u6 = v4 , u7 = v5 , u8=v6 , (5-16)
and in a matrix form we have,
,
vvvvvv
11
11
11
11
uuuuuuuu
6
5
4
3
2
1
8
7
6
5
4
3
2
1
⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅
=
⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
(5-17)
and in compact form:
u = Ev = Btv. (5-18)
CHAPTER 5 Matrix Displacement Method 129
129
The reason for matrix E being the transpose of the matrix B, has already been discussed in the previous chapter, however, using the principle of virtual work, a simple proof can be obtained. Consider:
W = work done by external loads = pv t21
U = strain energy = ru t21
Equating W and U leads to E = Bt and completes the proof.
Therefore the stiffness matrix of the entire structure can be obtained as:
.
2L/31L/300L/3L/6L/3L/600
1L/3401L/3L/3L/60L/12L/3L/6
001L/32L/300L/3L/6L/3L/6
LEI2
22
222
22
⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
−−
−−−−
−−−−
=K (5-19)
Applying the boundary conditions,
v1 = v2 = v5 = v6 = 0,
leads to the formation of the following reduced stiffness matrix:
⎥⎦
⎤⎢⎣
⎡
⎥⎥⎦
⎤
⎢⎢⎣
⎡=⎥
⎦
⎤⎢⎣
⎡
4
32
4
3vv
400L/12
LEI2
pp
. (5-20)
Since p4 = 0 and Pp3 −= , therefore
.EI24
PLEI24Lpv
333
3−
==
From this simple example, it can be seen that matrix B is a very sparse boolean matrix and the direct formation of BkBt using matrix multiplication requires a considerable amount of storage. In the following, it is shown that one can form BkBt with an assembling process (known also as planting), as follows:
130 Structural Mechanics: Graph and Matrix Methods
130
Consider an element "a" of a structure, as shown in Figure 5.4, for which the element stiffness matrix can be written as,
,jjji
ijiia ⎥
⎦
⎤⎢⎣
⎡=
kkkk
k (5-21)
i and j are the two end nodes of member a. Multiplication BkBt has the following effect on ka :
⎥⎥⎦
⎤
⎢⎢⎣
⎡
⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
⎥⎦
⎤⎢⎣
⎡
⎥⎥⎦
⎤
⎢⎢⎣
⎡
⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
= 00jj0ji00000ij0ii000
0000I0000I000000
00I000000000I000
jjjiijii
0000I0000I000000
kkkk
kkkk
(5-22)
⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
=
000000000000000000jj0ji0000000000000ij0ii000000000000000000000000000
87654321
kk
kk
8
7
3
6
5
a
4
12
Fig. 5.4 A structural model S.
The adjacency matrix of S is also an 8×8 matrix and the effect of node 4 being adjacent to node 6, is the existence of unit entries in the same locations as the submatrices of the element a. One can build up the adjacency matrix of a graph by the addition of the effect of one member at a time. In the same way, one can also form the overall stiffness matrix of the structure by the addition of the contribution of every member in succession. As an example, for the graph shown in Figure 5.4, the overall stiffness matrix has the following pattern:
CHAPTER 5 Matrix Displacement Method 131
131
.
87654321
87654321
⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
∗∗∗⋅⋅⋅⋅⋅∗∗⋅∗⋅∗⋅⋅∗⋅∗∗∗⋅⋅⋅⋅∗∗∗∗⋅∗⋅⋅⋅∗∗∗⋅⋅∗⋅∗⋅⋅⋅∗⋅⋅⋅⋅⋅∗⋅⋅∗⋅⋅⋅⋅⋅∗⋅⋅∗
(5-23)
Non-zero entries are shown by ∗ . For a stiffness matrix each of these non-zero entries is an α×α submatrix, where a is the degrees of freedom of each node of the structure. As an example, for a planar truss α = 2, and for a space frame α = 6. The formation of the stiffness matrix by the above process is known as assembling or planting of the stiffness matrix of a structure.
In the above example, the stiffness matrices could be assembled because both are constructed with reference to the same coordinate system. However, for a structure in general, the stiffness matrices should be prepared in a single coordinate system. On the other hand, for each element, there exists a coordinate system attached to the element, known as a local coordinate system. In Figure 5.5, local coordinate systems for members 45 and 25, and the global coordinate system for the entire structure are illustrated.
_
_
_x
_y
4 5
1 3
2
y
x
x
y
Fig. 5.5 Local y,x and global coordinate x, y systems.
A global coordinate system can be selected arbitrarily; however, it may be advantageous to select this coordinate system such that the structure falls in the first quadrate of the plane, in order to have positive coordinates for the nodes of the structure. On the other hand, a local coordinate system of a member has one of
132 Structural Mechanics: Graph and Matrix Methods
132
its axes along the member, the second axis lies in its plane of symmetry (if it has one) and the third axis is chosen such that it results in a right handed coordinate system.
The transformation from a local coordinate to a global coordinate system can be performed as illustrated in Figure 5.6, in which xyz is the global system and x2y2z2, often denoted by zyx , is the local system.
The relation between x1y1z1 and xyz can be expressed as:
⎥⎥⎥
⎦
⎤
⎢⎢⎢
⎣
⎡
⎥⎥⎥
⎦
⎤
⎢⎢⎢
⎣
⎡
αα−
αα=
⎥⎥⎥
⎦
⎤
⎢⎢⎢
⎣
⎡
zyx
cos0sin010
sin0cos
zyx
1
1
1 (5-24)
αβ
γ
y y
y
y
zz
z
z
x
x
x
x
1
2
3
2
3
1
2
3
1
*
y
z
x
L
L
x
z
j
i
y
xyz
xyz
i
i
i
j
j
j
ji
ji
ji
(a) (b)
Fig. 5.6 Transformation from local coordinate system to global coordinate system.
Similarly x2y2z2 and x1y1z1 are related by,
⎥⎥⎥
⎦
⎤
⎢⎢⎢
⎣
⎡
⎥⎥⎥
⎦
⎤
⎢⎢⎢
⎣
⎡ββ−ββ
=⎥⎥⎥
⎦
⎤
⎢⎢⎢
⎣
⎡
1
1
1
2
2
2
zyx
1000cossin0sincos
zyx
(5-25)
CHAPTER 5 Matrix Displacement Method 133
133
and ⎥⎥⎥
⎦
⎤
⎢⎢⎢
⎣
⎡
⎥⎥⎥
⎦
⎤
⎢⎢⎢
⎣
⎡
γγ−γγ=
⎥⎥⎥
⎦
⎤
⎢⎢⎢
⎣
⎡
2
2
2
3
3
3
zyx
cossin0sincos0
001
zyx
. (5-26)
Combining the above transformations results in,
.
)sinsinsincos(cos)sincos()sinsincoscos(sin)cossinsincos(sin)cos(cos)cossincossin(sin
)sin(cos)(sin)cos(cos
⎥⎥⎥
⎦
⎤
⎢⎢⎢
⎣
⎡
=
γβα+γαγβ−γβα−γα−γβα−αγγβγβα+γα−
αβββαT
(5-27)
where:
[ ] .zyx
zyx
3
3
3
⎥⎥⎥
⎦
⎤
⎢⎢⎢
⎣
⎡=
⎥⎥⎥
⎦
⎤
⎢⎢⎢
⎣
⎡T (5-28)
A vector in a local coordinate system Γ and in a global coordinate system Γ are related by:
. ΓTΓ = (5-29)
It can easily be proved that T is an orthogonal matrix, i.e.
[T]-1 = [T]t. (5-30)
In the above transformation, γ represents the tilt of the member which is quite often zero. Thus T can be simplified as:
.cos0sin
sinsincossincoscossinsincoscos
⎥⎥⎥
⎦
⎤
⎢⎢⎢
⎣
⎡
αα−βα−ββα−βαββα
=T (5-31)
This matrix can easily be written in terms of the coordinates of the two ends of a vector. Considering Figure 5.6(b), Eq. (5-31) becomes,
,*L/x0*L/z
L*L/zyL/*LL*L/yxL/zL/yL/x
jiji
jijijiji
jijiji
⎥⎥⎥
⎦
⎤
⎢⎢⎢
⎣
⎡
−−=T (5-32)
134 Structural Mechanics: Graph and Matrix Methods
134
where:
xji = xj−xi yji = yj−yi zji = zj−zi
.)xyz(L and )xz(*L 21
2ji
2ji
2ji2
12ji
2ji ++=+= (5-33)
Notice that T transforms a 3-dimensional vector from a global to a local coordinate system and Tt performs the reverse transformation. However, if the element forces or element displacements (distortions) consist of p vectors, the block diagonal matrix with p submatrices should be used. As an example, for a beam element of a space frame, with each node having 6 degrees of freedom, the transformation matrix is a 12×12 matrix of the form:
⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢
⎣
⎡
=
TT
TT
T (5-34)
5.3 ELEMENT STIFFNESS MATRICES
Element stiffness matrices for skeletal structures can be obtained using various methods. For some elements, concepts from mechanics of solids are sufficient for the formation of an element stiffness matrix; for others, energy methods are more suitable. In the following, a general method for the formation of a stiffness matrix is presented and then applied to bar and beam elements. The details of the derivations are omitted for brevity. Such details can be found in any classical book on the matrix analysis of structures.
5.3.1 STIFFNESS MATRIX OF A GENERAL ELEMENT
Consider an elastic body as shown in Figure 5.7. Suppose that some loads are applied at certain points (specified as nodes) 1,2,...,n. Let vit be the displacement of node i along the applied load pit. The loads are applied in a pseudo-static manner, increasing gradually from zero. Assuming a linear behaviour, the work done by an external force p = {p1, p2, ... , pn) through the displacement v = {v1, v2, ... , vn} can be written as:
).vp...vpv(p21W nn2211 +++= (5-35)
CHAPTER 5 Matrix Displacement Method 135
135
According to the principle of the conservation of energy,
W = U, (5-36)
and therefore: )vp...vpvp(21U nn2211 +++= . (5-37)
Now if a small variation is given to vi while keeping the other displacement components constant, then the variation of v with respect to vi can be written as:
].vvp...v
vpv
vpp[
vU
ni
n2
i
21
i
1i2
1
i ∂∂
++∂∂
+∂∂
+=∂∂ (5-38)
According to Castigliano´s theorem:
.pvU
ii=
∂∂ (5-39)
Thus,
],vvp...v
vpv
vp[p n
i
n2
i
21
i
1i ∂
∂++
∂∂
+∂∂
= (5-40)
or in a matrix form for all i=1,...,n we have:
⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅
⋅⋅
⋅⋅
=
⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
⋅⋅
⋅⋅
∂∂
∂∂
∂∂
∂∂
∂∂
∂∂
∂∂
∂∂
∂∂
n
2
1
n
2
1
v
vv
p
pp
nvnp
nv2p
nv1p
2vnp
2v2p
2v1p
1vnp
1v2p
1v1p
. (5-41)
According to the definition, the above coefficient matrix forms the stiffness matrix of the elastic body defined by its n nodes as illustrated in Figure 5.7.
A typical element of the stiffness matrix kij is given by:
.vp
ki
jij ∂
∂= (5-42)
Using Castigliano´s first theorem:
.vv
U)vU(
vk
ij
2
ijij ∂∂
∂=
∂∂
∂∂
= (5-43)
136 Structural Mechanics: Graph and Matrix Methods
136
p
pp
pp
p
,v,v
,v
,v,v
,v
1
2
3
4i
n
i
n
1
2
3
4i
n1
2
3
4
Fig. 5.7 An elastic body, its nodal forces and nodal displacements.
Similarly:
.vv
Uvpk
ij
2
j
iij ∂∂
∂=
∂∂
= (5-44)
Since the order of differentiation should not affect the result, we have,
kij = kji , (5-45)
which is proof of the symmetry of the stiffness matrices, both for a structure and for an element.
A symmetric matrix S is called positive definite, if xtSx > 0 for every non-zero vector x. The stiffness matrix K of a structure is positive definite since,
ptv = (Kv)tv = vtKtv = vtKv = 2W,
and W is always positive.
5.3.2 STIFFNESS MATRIX OF A BAR ELEMENT
Consider a prismatic bar element as shown in its local coordinate system, Figure 5.8. According to the definition of such an element, only axial forces are present.
The strain energy of this bar can be calculated as:
CHAPTER 5 Matrix Displacement Method 137
137
dx 2
EAdxdydz 2Edxdydz
21U 2
xx2xxxxxx ∫∫∫∫∫∫∫ ε=ε=εσ= (5-46)
On the other hand:
.x
ustrain xxx ∂
∂==ε (5-47)
x
y
z
r ,u _
_
r ,u1 14 4
ji
_ __ _
Fig. 5.8 A bar element in its local coordinate system.
Since the strain is constant along the bar, ux can be expressed as:
.AxAu 21x += (5-48)
From the boundary conditions:
0x at uu 1x ==
.Lx at uu 4x == (5-49)
Hence: .uA and L
uuA 1214
1 =−
= (5-50)
By substitution in Eq. (5-48), we have,
,uxL
uuu 114
x +−
= (5-51)
138 Structural Mechanics: Graph and Matrix Methods
138
and from Eq. (5-46) the strain energy of the bar can be calculated as:
].uuu2u[L2
EAU 2114
24 +−= (5-52)
Hence:
,L
EAuUk 21
2
11 =∂
∂= ,
LEA
uuUkk
41
2
4114 −=∂∂
∂== (5-53)
,L
EAuUk 24
2
44 =∂
∂= and 0kij = for all other components.
Therefore, the stiffness matrix of a bar element in the selected local coordinate system is obtained, and
⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
−
−
=
⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
6
5
4
3
2
1
6
5
4
3
2
1
uuuuuu
000000000000001001000000000000001001
LEA
rrrrrr
(5-54)
From Eq. (5-29), we have:
Trr = (5-55) and
Tuu = (5-56)
From the definition of an element stiffness matrix in a local coordinate system:
ukr = . (5-57)
By substitution of Eqs (5-55) and (5-56) in the above equation:
.ut1 TkTTukTr == − (5-58)
CHAPTER 5 Matrix Displacement Method 139
139
By definition of a stiffness matrix in a global coordinate system:
.kur = (5-59)
Comparison of Eq. (5-58) and Eq. (5-59) results in:
.t TkTk = (5-60)
Hence the stiffness matrix of a bar element in a global system, as shown in Figure 5.9, can be written as:
[ ] . t
⎥⎦
⎤⎢⎣
⎡⎥⎦
⎤⎢⎣
⎡=
TT
kT
Tk
x
y
z
1 1
4 4j
i r ,u
r ,ur ,u
r ,u
r ,u
r ,u
2 2
3 3
5 5
6 6
O
Fig. 5.9 A bar element of a space truss.
Denoting T in Eq. (5-32) by,
⎥⎥⎥
⎦
⎤
⎢⎢⎢
⎣
⎡=
333231
232221
131211
TTTTTTTTT
T . (5-61)
k can be written as:
140 Structural Mechanics: Graph and Matrix Methods
140
⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
−−−−−−−−−
=
21313121311
21313121311
21212111312
2121211
21113111211
211
21313121311
2121211
211
TTTTTTTTTTTTTTTTTT
TTTTTTTTTTT
.symTTTT
LEAk . (5-62)
The entries of the above matrix can be found using Tij from Eq. (5.32). As an example, the stiffness matrix of bar 1 in the planar truss shown in Figure 5.10, can be obtained as:
22
21
)zyx(
xT21
212
212
212
2111 ==
++
= ,
.22
21
)zyx(
yT21
212
212
212
2112 −=−=
++
=
x
y
1
2
3
12 L
L
Fig. 5.10 A planar truss and the selected global coordinate system.
Therefore:
⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢
⎣
⎡
−−−−−−
−−
=
5.05.05.05.05.05.05.05.05.05.05.05.0
5.05.05.05.0
2LEA
1k .
5.3.3 STIFFNESS MATRIX OF A BEAM ELEMENT
Consider a prismatic beam element as shown in Figure 5.11. The element forces and the element distortions, are defined by the following vectors:
CHAPTER 5 Matrix Displacement Method 141
141
,}r,...,r,r,r{ t12321=r
and
,}u,...,u,u,u{ t12321=u
where r1 – r3 are the force components at end i and r4 – r6 are moment components at end i. Also r7 – r9 are the force and r10 – r12 are the moment components, respectively at the end j, and ui (i=1,...,12) are correspondingly the translations and rotations at the ends i and j of the element.
x
y
z
_
_
_
ji
Fig. 5.11 A beam element in the local coordinate.
Using energy methods, the stiffness matrix of the beam element in the local coordinate system defined in Figure 5.11 can be obtained as:
⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
−
−ν+ν+−
−
−−−
−
−
−−ν+−ν+
−−−
−
−
=
zI4000L/zI60zI2000L/zI60
0yI40L/yI6000yI20L/yI60000)1(2/J00000)1(2/J000
0L/yI602L/yI12000L/yI602L/yI1200
L/zI60002L/yI120L/zI60002L/zI120
00000A00000AzI2000L/zI60zI4000L/zI60
0yI20L/yI6000yI40L/yI60000)1(2/J00000)1(2/J000
0L/yI602L/yI12000L/yI602L/yI1200
L/zI6002L/zI1200L/zI60002L/zI120
00000A00000A
L
Ek
(5-63)
142 Structural Mechanics: Graph and Matrix Methods
142
In which Iy, Iz and J are the moments of inertia with respect to the y and z axes and J is the polar moment of inertia of the section. E specifies the elastic modulus and υ is the Poisson ratio. The length of the beam is denoted by L.
For the two-dimensional case the columns and rows corresponding to the third dimension can easily be deleted, to obtain the stiffness matrix of an element of a planar frame.
The stiffness matrix in a global coordinate system can be written as:
[ ] .
t
⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢
⎣
⎡
⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢
⎣
⎡
=
TT
TT
k
TT
TT
k (5-64)
For the two-dimensional case,
[ ] . t
⎥⎦
⎤⎢⎣
⎡⎥⎦
⎤⎢⎣
⎡=
TT
kT
Tk (5-65)
The entries of k are as follows,
z4
2211
21111 TTk α+α=
z422211121121 TTTTk α+α= z
42221
21222 TTk α+α=
z22131 Tk α= , z
22232 Tk α= z333k α=
z4
2211
21141 TTk α+α−= , 11112
z4222142 TTTTk α−α−= , z
22143 Tk α−= ,z22144 Tk α−= (5-66)
11112z4222151 TTTTk α−α−= 1
212
z4
22152 TTk α−α−= z
22253 Tk α−=
11112z4222154 TTTTk α+α= 1
212
z4
22255 TTk α+α=
z22161 Tk α= , z
22262 Tk α= , z663k α= , z
22164 Tk α−= , z22265 Tk α−= , z
366k α= .
CHAPTER 5 Matrix Displacement Method 143
143
in which:
LEA
1 =α , ,LEI6
2zz
2 =α ,LEI4 zz
3 =α ,LEI123
zz4 =α and .
LEI2 zz
6 =α
As an example, consider the planar frame, shown in Figure 5.12, with 2114623 N/m102E and m1030I ,m104A ×=×=×= −− . For element 1 we have,
T11 = 0 T12 = 1 T21 = − 1 T22 = 0,
and the stiffness matrix of the element is obtained as,
⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
−−
−−
=
6075.03075.0200002000
25.175.0025.16075.0
.sym200025.1
1061k ,
where "sym." denotes the symmetry of the matrix.
x
y
1
23
1
2
4m
4m
Fig. 5.12 A planar frame.
5.4 OVERALL STIFFNESS MATRIX OF A STRUCTURE
Once the stiffness matrix of an element is obtained in the selected global coordinate system, it can be planted in the specified and initialised overall stiffness
144 Structural Mechanics: Graph and Matrix Methods
144
matrix of the structure K, using the process described in Section 5.2. This is illustrated by the following simple example:
Let S be a planar truss with an arbitrary nodal and element numbering, as shown in Figure 5.13. The entries of the transformation matrices of the members are calculated using Eq. (5-32) and Eq. (5-33) as follows:
For bar 1: 21
201
2xxT 12
11 =−
=−
= and .23
203
2yyT 12
12 =−
=−
=
Similarly for bar 2: 21T11 = ,
23T12 −= and for bar 3 T11 =1, T12 = 0.
2
3
4
11m 1m 1m
x
y 3m1
3
2
2030kN
kN
Fig. 5.13 A planar truss and the selected global coordinate system. Now the stiffness matrices can be formed using Eq. (5-62) as:
For bar 1: ⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢
⎣
⎡
−−−−
=
75.0433.075.0433.025.0433.025.0
75.0433.0.sym25.0
2EA
1k .
For bar 2: ⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢
⎣
⎡
−−−−
=
75.0433.075.0433.025.0433.025.0
75.0433.0.sym25.0
2EA
2k .
CHAPTER 5 Matrix Displacement Method 145
145
For bar 3: ⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢
⎣
⎡
−=
.0000101
00.sym1
2EA
3k .
The overall stiffness matrix of the structure is an 8×8 matrix, which can easily be formed by planting the three member stiffness matrices as follows:
.
10000000010001000075.0433.075.0433.00000433.025.0433.025.0000075.0433.05.1075.0433.001433.025.005.1433.025.0000075.0433.075.0433.00000433.025.0433.025.0
2EA
⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
−−−−
−−−−−
−−−−−−−−
=K
Partitioning K into 2×2 submatrices, it can easily be seen that it is pattern equivalent to the node adjacency matrix of the graph model of the structure as follows:
⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢
⎣
⎡
∗⋅∗⋅⋅∗∗⋅∗∗∗∗⋅⋅∗∗
=*D .
This pattern equivalence simplifies certain problems in structural mechanics, such as ordering the variables for bandwidth or profile reduction, methods for increasing the sparsity using special cutset bases, and improving the conditioning of structural matrices, which will be discussed in Chapters 7 and 8.
The matrix K is singular, since the boundary conditions have to be applied. Consider,
p = Kv
and partition it for free and constraint degrees of freedom as:
146 Structural Mechanics: Graph and Matrix Methods
146
⎥⎦
⎤⎢⎣
⎡⎥⎦
⎤⎢⎣
⎡=⎥
⎦
⎤⎢⎣
⎡
c
f
cccf
fcff
c
fvv
KKKK
pp
. (5-67)
This equation has a mixed nature; pf and vc have known values and pc and vf are unknowns. Kff is known as the reduced stiffness matrix of the structure, which is non-singular for a rigid structure.
For boundary conditions such as vc = 0, it is easy to delete the corresponding rows and columns to obtain,
,ffff vKp = (5-68)
from which vf can be obtained by solution of the above set of equations. In a computer this can be done by multiplying the diagonal entries of Kcc by a big number such as 1020. An alternative approach is possible by equating the diagonal entries of Kcc to unity and all the other entries of these rows and columns to zero. If vc contains some specified values, pc will have corresponding vc values. A third method, which is useful when a structure has more constraint degrees of freedom (such as many supports), consists of the formation of element stiffness matrices considering the corresponding constraints, i.e. to form the reduced stiffness matrices of the elements in place of their complete matrices. This leads to some reduction in storage, also at the expense of additional computational effort.
As an example, the reduced stiffness matrix of the structure shown in Figure 5.13 can be obtained from K, by deleting the rows and columns corresponding to the three supports 1, 3 and 4.
⎥⎦
⎤⎢⎣
⎡⎥⎦
⎤⎢⎣
⎡=⎥
⎦
⎤⎢⎣
⎡
y2
x2uu
5.1005.1
2EA
3020
.
The solution results in the joint displacements as:
.EA40u and
EA5.140u y2x2 ==
The member distortions can easily be extracted from the displacement vector, and multiplication by the stiffness matrix of each member results in its member forces in the global coordinate system. As an example, for member 3 we have:
CHAPTER 5 Matrix Displacement Method 147
147
⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢
⎣
⎡
−=
⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢
⎣
⎡
⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢
⎣
⎡
−=
⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢
⎣
⎡
033.13
033.13
00EA/40
EA5.1/40
0000101
.sym001
2EA
rrrr
y4
x4
y2
x2
.
A transformation yields the member forces in the local coordinate systems, t
1 23.99} 99.23{−=r , t2 10.65} 659.10{−=r and t
3 13.33} 33.13{ −=r .
Example: The truss shown in Figure 5.14 has members each of the same cross sectional area of 15000mm2, and elastic modulus 210 kN/mm2. Vertical loads of 10kN and 5kN are applied at node 3 and node 5, respectively. Determine the forces in all members:
Fig. 5.14 A planar truss S.
The force-displacement relationship for a planar bar member is obtained from Eq. (5-23) as follows:
⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢
⎣
⎡
δ
δ
δδ
⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢
⎣
⎡
−−−−
−−−−
=
⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢
⎣
⎡
yj
xj
yi
xi
2121211
2121211
12112
1112112
11
2121211
2121211
12112
1112112
11
xj
xj
yi
xi
TTTTTTTTTTTTTTTTTT
TTTTTT
LEA
F
FFF
(5-69)
The stiffness matrices for the members of S are determined as:
For members 1 and 2:
148 Structural Mechanics: Graph and Matrix Methods
148
⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢
⎣
⎡
−
−
000000.157500.1575000000.157500.1575
, ⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢
⎣
⎡
−
−
000000.157500.1575000000.157500.1575
For members 3 and 4:
⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢
⎣
⎡
−−−−
−−−−
03.23202.53003.23202.53002.53071.121002.53071.121003.23202.53003.23202.53002.53071.121002.53071.1210
,
⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢
⎣
⎡
−−−−−−
−−
76.51418.58876.51418.58818.58808.76218.58808.67276.51418.58876.51418.588
18.58808.67218.58808.672
For members 5 and 6:
⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢
⎣
⎡
−
−
0.360000.360000000
0.360000.360000000
, ⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢
⎣
⎡
−−−−−−
−−
03.23202.53003.23202.53002.53071.121002.53071.121003.23202.53003.23202.530
02.53071.121002.53071.1210
For member 7:
⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢
⎣
⎡
−−−−−−
−−
03.23202.53003.23202.53002.53071.121002.53071.121003.23202.53003.23202.530
02.53071.121002.53071.1210
Assembling the stiffness matrix of the entire structure and imposing the boundary conditions 0y
4x4
y1
x1 =δ=δ=δ=δ results in:
CHAPTER 5 Matrix Displacement Method 149
149
⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
δδ
δδ
δδ
⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
−−−−−
−
=
⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
−
−
y5
x5
y3
x3
y2
x2
03.23202.5300003.23202.53071.278500.157502.53071.1210
76.411418.5880.3600008.382200
.sym09.429602.53013.3632
5010000
The solution of the above equations results in the joint displacements:
3x2 10716814.4 −×=δ , 2y
2 1012241.2 −×−=δ , 2x3 1009894.1 −×−=δ ,
2y3 1025665.2 −×−=δ 2x
5 10824108.1 −×−=δ and 2y5 10521007.9 −×−=δ .
Once the displacements are calculated, the member forces can easily be obtained using member stiffness matrices.
5.5 GENERAL LOADING
The joint load vector of a structure can be computed in two parts. The first part comes from the external concentrated loads and/or moments, which are applied at the joints defined as the nodes of S. The components of such loads are most easily specified in a global coordinate system and can be entered to the joint load vector p.
The second part comes from the loads which are applied on members. These loads are usually defined in the local coordinate system of a member. For each member the fixed end actions (FEA) can be calculated using the existing classical formulae or tables. A simple computer program can be prepared for this purpose. The fixed end actions should then be rotated to the global coordinate system using the transformation matrix given by Eq. (5-27). The FEA should then be reversed and applied to the end nodes of the members. These components can be added to p to form the final joint load vector. After p has been prepared and the boundary conditions imposed, the corresponding equations should be solved to obtain the joint displacements of the structure. Member distortions can be extracted for each member in the reverse order to that used in assembling p vector.
Example: A portal frame is considered as shown in Figure 5. 15. The members are all made of sections with area A = 150cm2, moment of inertia Iz = 2×104cm4 and elastic modulus E = 2×104kN/cm2. Calculate the joint rotations and displacements.
150 Structural Mechanics: Graph and Matrix Methods
150
4m
5m
1
2 3
4
1
2
3
50kN
12kN/m
Fig. 5.15 A portal frame and its loading.
The equivalent joint loads are illustrated in Figure 5.16:
x
y
74kN
1600kN.m
Fig. 5.16 Equivalent joint loads.
Employing Eq. (5-66), the stiffness matrices for the members are obtained as follows:
For member 1:
⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
−−
−=
40005.120005.175.00075.00
008.05.10008.040005.1
.sym75.00008.0
1041k ,
and for member 2:
CHAPTER 5 Matrix Displacement Method 151
151
⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
−−−
−=
32096.0016096.00004.0096.0004.00
6.0006.032096.00
.sym004.006.0
1042k .
For member 3:
⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
−−
−−=
40005.120005.175.00075.00
008.05.10008.040005.1
.sym75.00008.0
1043k .
Assembling the stiffness matrices and imposing the boundary conditions results in the following equations:
⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
θ
δδθ
δδ
⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
−−−
−=
⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
z3
y3
x3
z2
y2
x2
4
72096.05.116096.00754.0096.0004.00
608.0006.072096.05.1
.sym754.00608.0
10
000
16004.7
Solving these equations leads to:
x2δ = 0.0659167, y
2δ = 2.617764E−04, z2θ = −8.983453E−05,
x3δ = 0.06533767, −=δy
3 2.617704E−04 and −=θz3 1.16855E−04.
5.6 COMPUTATIONAL ASPECTS OF THE MATRIX
DISPLACEMENT METHOD
The main advantage of the displacement method is its simplicity for computer programming. This is due to the existence of a simple kinematical basis formed on
152 Structural Mechanics: Graph and Matrix Methods
152
a special cutset basis known as cocycle basis of the graph model S of the structure. Such a basis does not correspond to the most sparse stiffness matrix, however, the sparsity is good enough, not to look for a better basis in more usual cases. However, if an optimal cutset basis of S is needed, then the displacement method has all the problems encountered in the force method, described in Chapter 6. The algorithm for the displacement method is summarized in the following. The coding for such an algorithm may be found in textbooks such as those of Vanderbilt [242] and Meek [173].
Algorithm
Step 1: Select a global coordinate system and number the nodes and members of the structure. An appropriate nodal ordering algorithm will be discussed in Chapter 7.
Step 2: After initialisation of all the vectors and matrices, read the data for the structure and its members. For multi-member regular structures, data can be generated using the method of Chapter 10.
Step 3: For each member of the structure:
(a) compute L, L*, sinα, sinβ,sinγ, cosα, cosβ, cosγ;
(b) compute the rotation matrix T;
(c) form the member stiffness matrix k in its local coordinate system;
(d) form the member stiffness matrix k in the selected global coordinate system;
(e) plant k in the overall stiffness matrix K of the structure.
Step 4: For each loaded member:
(a) read the fixed end actions;
(b) transform the fixed end actions to the global coordinate system and reverse it to apply at joints;
(c) store these joint loads in the specified overall joint load vector.
Step 5: For each loaded joint:
(a) read the joint number and the applied joint loads;
CHAPTER 5 Matrix Displacement Method 153
153
(b) store it in the overall joint load vector.
Step 6: Apply boundary conditions to the structural stiffness matrix K, to obtain the reduced stiffness matrix Kff. Repeat the same for the overall joint load vector.
Step 7: Solve the corresponding equations to obtain the joint displacements.
Step 8: For each member:
(a) extract the member distortions from the joint displacements;
(b) rotate the member distortions to the local coordinate system;
(c) compute the member stiffness matrix;
(d) compute the member forces and fixed end actions.
Step 9: Compute the final member forces.
For an efficient displacement analysis of a structure, special considerations must be taken into account, which will be discussed in Chapters 7, 8 and 10 of this book.
x
y
z
4@2m
3@2m
P
P P P
PP