DIRECT STIFFNESS METHOD FOR

ANALYSIS OF SKELETAL STRUCTURES

http://web.iitd.ac.in/~sbhalla/cvl756.html
Password: cvl756
Dr. Suresh Bhalla
Professor
Department of Civil Engineering
13/08/2019 Indian Institute of Technology Delhi 1
 SLOPE DEFLECTION METHOD

Applied moment M θ
Carry over moment
4 EI
θ 2 EIθ
L
6 EIθ 6 EIθ L
L2 L2

12 EIδ
6 EIδ L3
L2 δ

6 EIδ
12 EIδ
13/08/2019
L2 2
L3
 Generalized slope deflection equations

2 EI ⎡ 3(δ 2 − δ 1 ) ⎤
M1 θ2 M1 = 2θ +
⎢⎣ 1 2 θ − ⎥⎦
δ2 L L
θ1
δ1 2 EI ⎡ 3(δ 2 − δ 1 ) ⎤
M2 M2 = θ
⎢⎣ 1 + 2θ 2 − ⎥⎦
L L

F1 F2 2 EI ⎡ 6(δ 2 − δ 1 ) ⎤
F1 = 3θ
⎢⎣ 1 + 3θ 2 − ⎥⎦
L2 L

2 EI ⎡ 6(δ 2 − δ 1 ) ⎤
F2 = − 3θ
⎢⎣ 1 + 3θ 2 − ⎥⎦
L2 L

M1 + M 2
F1 = − F2 =
L

13/08/2019 3
 Put in matrix form

⎡ 12 EI 6 EI 12 EI 6 EI ⎤
⎢ L3 − 3
⎡ F1 ⎤ ⎢ 6 EI L2 L L2 ⎥ ⎡δ ⎤
4 EI 6 EI 2 EI ⎥ 1
⎢M ⎥ ⎢ 2
− 2 ⎥ ⎢θ ⎥
⎢ 1⎥ = ⎢ L L L L ⎥⎢ 1 ⎥
− 2 ⎥ ⎢δ 2 ⎥
⎢ F2 ⎥ ⎢ 12 EI 6 EI 12 EI 6 EI ⎢ ⎥
⎢ ⎥ ⎢− L3 − 2
L L3 L ⎥θ
⎣ M 2 ⎦ ⎢ 6 EI 2 EI 6 EI 4 EI ⎥ ⎣ 2 ⎦
⎢ − 2 ⎥
⎣ L2 L L L ⎦

{F} = [K]{δ}
Force Stiffness Displacement
vector matrix vector
(symmetric)
13/08/2019 4
⎡ F1 ⎤ ⎡ K11 K12 . K1n ⎤ ⎡δ1 ⎤ 1 2 3
⎢F ⎥ ⎢K K 22 . K 2 n ⎥ ⎢δ 2 ⎥
⎢ 2 ⎥ = ⎢ 21 ⎥⎢ ⎥
⎢.⎥ ⎢ . . . . ⎥⎢ . ⎥
⎢ ⎥ ⎢ ⎥⎢ ⎥
⎣ Fn ⎦ ⎣ K n1 K n1 . K nn ⎦ ⎣δ n ⎦
n = Degrees of freedom (DKI)
Fi = Force along ith degree of freedom
δi = Displacement along ith degree of freedom Kij =Force generated along ith
degrees of freedom under a unit
jth col. of [K] : displacement along the jth degree of
Forces generated along the various degrees of freedom with all other degrees of
freedom under a unit displacement along the jth freedom locked.
degree of freedom (δj = 1), with all other
degrees of freedom locked (δx = 0, where x ≠j )

Matrix Stiffness Approach (MSA) : The elements of [K] are obtained

by first principles using the definition of kij from the deformation
pattern of the structure and force-deformation relations of the
members
13/08/2019 5
 Matrix Stiffness Approach (MSA) : Human judgment needed
1 2 3

The direct stiffness approach (DSA),

on the other hand, enables
computation of the overall stiffness
method of any complicated structure
using computer program, based on
finite element formulation, without
13/08/2019 human judgement. 6
 SIGNIFICANCE OF DIRECT
STIFFNESS METHOD
ALL COMMERCIAL STRUCTURAL
ENGINEERING ANALYSIS PACKAGES ARE
BASED ON THE DIRECT STIFFNESS
APPROACH.

UNDERSTANDING AND IMPLEMENTING THE CONCEPTS

WILL HELP YOU IN:
1. MAKING YOUR OWN CUSTOMIZED RESULT ORIENTED
SOFTWARE WITHOUT SPENDING ANY PENNY.
2. USING THE EXISTING SOFTWARE IN ERROR FREE
MANNER, WITH UNDERSTANDING, RATHER THAN AS A
“BLACK BOX” APPROACH.
13/08/2019 7
 DIRECT STIFFNESS METHOD: ASSUMPTIONS

1. Restricted to frame and truss structures (skeletal structures) only. Members

assumed as line elements (passing through neutral axis) with lumped
sectional properties. At first, we restrict analysis to prismatic members only.

2. Hooke’s law of elasticity holds.

3. Small deflections => no change in overall geometry of structure.

13/08/2019 8
 DIRECT STIFFNESS METHOD: ASSUMPTIONS
4. Plane sections remain plane after bending.

5. In bending mode, very small slope

d 2 y / dx 2 d2y
Curvature =
[1 + (dy / dx ) ]
2 3/ 2 ≈
dx 2
6. If displacement takes place normal to member, no change in length of the
member. Change in length of an element due to flexural deformation
(curvature effects) is also negligible.

13/08/2019 9
 DIRECT STIFFNESS METHOD: ASSUMPTIONS

7. Principle of superposition holds good.

• Loads can be superimposed
• Boundary conditions can be superimposed
• Displacements can be superimposed
• BMD, SFD can be superimposed.

IMPORTANT:
All assumptions of slope deflection method are repeated
except one…..

We have discarded the assumption regarding

inextensibility of the members……
Unlike manual approach, digital computers will not
have no problem in tackling additional degrees of
freedom.
13/08/2019 10
 DIRECT STIFFNESS METHOD FOR
COMPUTER APPLICATIONS

• Each individual member is treated as structure (called

element).

• Stiffness matrix of each individual element is obtained.

• Total stiffness matrix of the entire structure is then

computationally obtained by superimposing the matrices of
elements, without human intervention.

• Hence, analysis can be broken down into small steps and

programmed, in a finite element procedure.

13/08/2019 11
 GENERATION OF ELEMENT STIFFNESS MATRIX

2D STRUCTURES
x’
5 4
⎡ EA ⎤
⎢ L ⎥
⎢ 12 EI ⎥
Local coordinate ⎡ f1 ⎤ ⎢ 0 ⎥ ⎡ d1 ⎤
system ⎢f ⎥ ⎢ L3 ⎥ ⎢d ⎥
y’ 6 ⎢ 2⎥ ⎢ 6 EI 4 EI ⎥⎢ 2 ⎥
θ ⎢ f3 ⎥ ⎢ 0 L2 L ⎥ ⎢d 3 ⎥
⎢ ⎥ = ⎢ EA EA ⎥ ⎢d ⎥
y ⎢ f 4 ⎥ ⎢− 0 0 ⎥⎢ 4 ⎥
2 ⎢ f5 ⎥ ⎢ L L ⎥ ⎢d 5 ⎥
3 ⎢ ⎥ ⎢ 12 EI 6 EI 12 EI ⎥⎢ ⎥
⎣⎢ f 6 ⎦⎥ ⎢ 0 − − 0
1 Global coordinate L3 L2 L3 ⎥ ⎣⎢ d 6 ⎦⎥
system ⎢ 6 EI 2 EI 6 EI 4 EI ⎥
⎢ 0 0 − 2 ⎥
x ⎣ L2 L L L ⎦
z axis normal to plane of board towards viewer

In short form, {f}= [k] L {d}

[k]L = Element stiffness matrix with respect to

13/08/2019 local coordinate system. 12
 TRANSFORMATION OF
COORDINATE SYSTEM

¾At a joint, members of different orientations may

meet.

¾The forces and displacements at member ends

cannot be easily related.

¾To consider equilibrium of the joint and compatibility

of member displacements, the member end forces
and displacements must be transformed to a
common coordinate system.
13/08/2019 13
 y v
V = A vector (force or displacement)
y’

r ⎡Vx′ ⎤ ⎡ cosθ sin θ ⎤ ⎡Vx ⎤

⎢V ′ ⎥ = ⎢ ⎢ ⎥
Vy
⎥
V y′
V ⎣ y ⎦ ⎣− sin θ cosθ ⎦ ⎢⎣V y ⎥⎦
x’ (Local)
φ
V x′
θ
x (Global)
Vx

v v
V = Vx î + Vy ĵ V = Vx’ î’ + Vy ’ĵ’

= (Vcosφ)î + (Vsinφ)ĵ = Vcos(φ-θ) î’ + Vsin(φ-θ) ĵ’

= [Vcosφcosθ + Vsinφsinθ] î’ + [Vsinφcosθ -Vcosφsinθ] ĵ’

= [Vxcosθ + Vysinθ] î’ + [-Vxsinθ +Vycosθ] ĵ’

Hence Vx’ = Vxcosθ + Vysinθ

13/08/2019 14
Vy’ = -Vxsinθ + Vycosθ
MEMBER FORCES IN GLOBAL & LOCAL COORDINATES

Local coordinate
system
f5 F5
f4
y’ x’ F4

f6 F6

f2 θ y F2

f1 f3 F1 F3
x

Global coordinate
Member end forces system
Member end forces
in local coordinates in global coordinates
⎡ f1 ⎤ ⎡ cos θ sin θ 0⎤ ⎡ F1 ⎤ ⎡ f 4 ⎤ ⎡ cos θ sin θ 0⎤ ⎡ F4 ⎤
⎢ f ⎥ = ⎢− sin θ 0⎥⎥ ⎢⎢ F2 ⎥⎥ ⎢ f ⎥ = ⎢− sin θ
⎢ 2⎥ ⎢ cos θ
⎢ 5⎥ ⎢ cos θ 0⎥⎥ ⎢⎢ F5 ⎥⎥
⎢⎣ f 3 ⎥⎦ ⎢⎣ 0 0 1⎥⎦ ⎢⎣ F3 ⎥⎦ ⎢⎣ f 6 ⎥⎦ ⎢⎣ 0 0 1⎥⎦ ⎢⎣ F6 ⎥⎦
13/08/2019 15
 ⎡ f1 ⎤ ⎡ cosθ sin θ 0⎤ ⎡ F1 ⎤ ⎡ F1 ⎤
⎢ f ⎥ = ⎢− sin θ 0⎥⎥ ⎢⎢ F2 ⎥⎥ =
⎢ 2⎥ ⎢
⎢⎣ f 3 ⎥⎦ ⎢⎣ 0
cosθ
0 1⎥⎦ ⎢⎣ F3 ⎥⎦
R ⎢F ⎥
⎢ 2⎥
⎢⎣ F3 ⎥⎦
1

⎡ F1 ⎤ -1 ⎡ f1 ⎤ T ⎡ f1 ⎤
⎢F ⎥ =
⎢ 2⎥
⎢⎣ F3 ⎥⎦
R ⎢f ⎥
⎢ 2⎥
⎢⎣ f 3 ⎥⎦
= R ⎢f ⎥
⎢ 2⎥
⎢⎣ f 3 ⎥⎦

R-1 = RT R is an orthogonal matrix.

R = JOINT TRANSFORMATION MATRIX

⎡ f4 ⎤ ⎡ F4 ⎤
Similarly ⎢f ⎥=
⎢ 5⎥
⎢⎣ f 6 ⎥⎦
R ⎢F ⎥
⎢ 5⎥
⎢⎣ F6 ⎥⎦
2
 Combining equations 1 and 2

⎡ f1 ⎤ ⎡ F1 ⎤
⎢f ⎥ 0 0 0
⎢F ⎥
⎢ 2⎥
⎢ f3 ⎥ R 0 0
0
0
0
0
0
⎢ 2⎥
⎢ F3 ⎥
⎢ ⎥= 0 0 0 ⎢ ⎥
⎢ f4 ⎥
⎢ f5 ⎥
⎢ ⎥
0
0 0 R
0
0
0
0
⎢ F4 ⎥
⎢ F5 ⎥
⎢ ⎥
⎢⎣ f 6 ⎥⎦ ⎢⎣ F6 ⎥⎦

{ f } = [ T ] {F}
[T] = MEMBER TRANSFORMATION MATRIX

Similarly for displacements { d } = [ T ] {D}

 HOW WILL PROGRAM OBTAIN THE NECCESARY
INFORMATION FOR COORDINATE TRANSFORMATION??
User should provide the coordinates of all joints….
x2 − x1
cos θ =
L
(x2,, y2)
Local coordinate y −y
system sin θ = 2 1
y’ L
θ
y L = ( x2 − x1 ) 2 + ( y2 − y1 ) 2
(x1,, y1) Global coordinate
system

x
⎡ cos θ sin θ 0⎤
R = ⎢⎢− sin θ cos θ 0⎥⎥
⎢⎣ 0 0 1⎥⎦

R 0
[T] = T =
-1
T
T

13/08/2019 0 R 18
 HOW IS TRANSFORMATION UTILIZED??

{f} = [k]L {d}

[T] {F} = [k]L[ T ] {D}

{F} = [ T ]T [k]L [ T ] {D}

[K]G Stiffness matrix of member in global

coordinates

[K]G =[ T ]T [k]L [ T ]
13/08/2019 19
 SPECIAL CASE: TRUSS STRUCTURES

x’ ⎡ EA EA ⎤
⎧ f1 ⎫ ⎢ 0 − 0⎥ ⎧ d 1 ⎫
4 3 ⎪f ⎪ ⎢ L L ⎪ ⎪
⎪ 2⎪ 0 0 0 0 ⎥ ⎪d 2 ⎪
⎨ ⎬ = ⎢ EA EA ⎥⎨ ⎬
Local coordinate system
⎪ f 3 ⎪ ⎢− 0 0⎥ ⎪ d 3 ⎪
⎪⎩ f 4 ⎪⎭ ⎢ L L ⎥ ⎪d ⎪
y’ ⎢⎣ 0 0 0 0⎥⎦ ⎩ 4 ⎭
θ
y
2
1 Global coordinate ⎡ cos θ sin θ ⎤
R=⎢
cos θ ⎥⎦
system
x
⎣− sin θ
No member end moments or
rotations

13/08/2019 20
MEMBER FORCES IN GLOBAL & LOCAL COORDINATES

Local coordinate
system
f5 F5
f4
y’ x’ F4

f6 F6

f2 θ y F2

f1 f3 F1 F3
x

Global coordinate
Member end forces system
Member end forces
in local coordinates in global coordinates
⎡ f1 ⎤ ⎡ cos θ sin θ 0⎤ ⎡ F1 ⎤ ⎡ f 4 ⎤ ⎡ cos θ sin θ 0⎤ ⎡ F4 ⎤
⎢ f ⎥ = ⎢− sin θ 0⎥⎥ ⎢⎢ F2 ⎥⎥ ⎢ f ⎥ = ⎢− sin θ
⎢ 2⎥ ⎢ cos θ
⎢ 5⎥ ⎢ cos θ 0⎥⎥ ⎢⎢ F5 ⎥⎥
⎢⎣ f 3 ⎥⎦ ⎢⎣ 0 0 1⎥⎦ ⎢⎣ F3 ⎥⎦ ⎢⎣ f 6 ⎥⎦ ⎢⎣ 0 0 1⎥⎦ ⎢⎣ F6 ⎥⎦
13/08/2019 21
 GENERATION OF TOTAL STRUCTURAL
STIFFNESS MATRIX
2 3 4
1

6 7
5 8

i+1
i

n
n-1

We shall first derive formulations for simple 2D case:

(1)Supports are fixed (2) All joints are rigid with no internal hinges.
(3) Joints can be sequentially numbered as above

We shall introduce complications into analysis one by one.

13/08/2019 22
 NUMBERING SCHEME
Joints numbered sequentially, restrained joints numbered in end
P2 , u2 Joint loads and joint displacements
P1, u1 ⎡ P1 ⎤ ⎡ k11 k12 . k1( 3n ) ⎤ ⎡ u1 ⎤
2 3 4
⎢P ⎥ ⎢ . . . . ⎥⎥ ⎢u2 ⎥
1 ⎢ 2 ⎥=⎢ ⎢ ⎥
⎢ . ⎥ ⎢ . . . . ⎥⎢ . ⎥
⎢ ⎥ ⎢ ⎥⎢ ⎥
⎣ X 3n ⎦ ⎣⎢k(3n )1 k(3n ) 2 . k( 3n )(3n ) ⎦⎥ ⎣un ⎦
P3, u3 5
6 7 8

Total structural stiffness matrix should

relate global structural loads to global
P3 i - 1 structural displacements
c
P3 i - 2 b i+1
i X3 n - 1
a X3n - 2
n
P3 i n-1

Similar pattern for numbering X3 n

of displacements
13/08/2019 Degrees of freedom shall include those at 23
supports also
 5
6
4

c
1 3
3i-1 2 5
2 3 b
3i-2 Joint i 5 1 4
3j-1
6
4 6
3j-2 Joint
3i a j=i+1
1 3
3j
2

Member degrees of freedom: From element point of view (1..6)

Structural degrees of freedom: From global (overall structures)
point of view (1..3n)
Each member degree of freedom in global coordinates (1,2,…,6)
corresponds to a particular structural degree of freedom (1,2,….,3n).

13/08/2019 COMPATIBILITY u3i −2 = D4a = D1b = D1c 24

 EQUILIBRIUM CONDITIONS
P2 , u2
P1, u1
2 3 4
1

P3, u3 5
6 7 8

P3 i - 1
c
P3 i - 2 b i+1
i

a
n
P3 i n-1

Let a unit displacement be applied along

d.o.f (3i-2) and all other d.o.f. =0

13/08/2019 25
 5
6
4

c
1 3
3i-1 2 5
2 3 b
3i-2 Joint i 5 1 4

6
4 6

Compatibility Joint
3i a j=i+1
1 3 u3i −2 = D4a = D1b = D1c
2

Let a unit displacement be applied along d.o.f (3i-2) and all other d.o.f. =0
By joint equilibrium,

P3i-2 = Sum of member end forces of a, b, c in X-direction

P3i-2 = Fa4 + F1b +Fc1

Recall: kmn = Force induced along d.o.f.’m’ due to unit displacement along d.o.f ’n’, all
other displacements maintained zero.

13/08/2019
K3i-2, 3i-2 = ka44 + kb11 + kc11 26
 K3i-2, 3i-2 = ka44 + kb11 +kc11
Elements of member
Element of total structural stiffness matrices in global
stiffness matrix coordinates

• An element of [K]TS can be obtained by summing the

elements of member stiffness matrices (in global
coordinates) of corresponding d.o.f from members that
frame into that joint.

• In order to carryout smoothly, we follow

Code Number Approach.

13/08/2019 27
Each member degree of freedom in global coordinates (1,2,…,6) corresponds to a particular structural degree
of freedom (1,2,….,3n). This information can be stored in the association matrix of the member.

K34 of [K]L will correspond

to K3i, 3(i+1)-2 of [K]TS
(i.e. will transfer to that
location)

For member ‘b’, the association matrix is

3 4

3i-2 3i-1 3i 3(i+1)-2 3(i+1)-1 3(i+1)

2 5 3i-1 3(i+1)-1
3 b
1 4 3i-2
3i b
3(i+1)-2
6 3(i+1)

13/08/2019 28
 EXAMPLE
1 2 3
⎡ ⎤
4 5 6
⎢ ⎥
⎢ ⎥
7 8 9 ⎢ ⎥
⎢ ⎥
⎣ ⎦
13/08/2019 29
 KTS GENERATION: MATRIX STIFFNESS VS
DIRECT STIFFNESS APPROACH
1 2 3
⎡ ⎤
4 5 6
⎢ ⎥
⎢ ⎥
7 8 9 ⎢ ⎥
⎢ ⎥
⎣ ⎦
KTS (9,9) =??
13/08/2019 30
 KTS GENERATION: MATRIX STIFFNESS VS
DIRECT STIFFNESS APPROACH
1 2 3
⎡ ⎤
4 5 6
⎢ ⎥
⎢ ⎥
7 8 9 ⎢ ⎥
⎢ ⎥
⎣ ⎦
KTS (6,9) =??
13/08/2019 31
 HOW TO GENERATE THE TOTAL STRUCTURAL
STIFFNESS MATRIX

ƒ All joints of the structure should be numbered sequentially,

starting from the unstrained joints.
ƒ Restrained joints should be numbered in the end.
ƒ Initialize the total structural stiffness matrix to ‘0’.
ƒ Consider each member; compute its member stiffness matrix
in global coordinates.
ƒ Then send its elements into appropriate location of the global
stiffness matrix of the entire structure, one at a time.
ƒ Repeat this process for each member; keep adding its
elements to the appropriate elements of the total structural
stiffness matrix.
ƒ Finally, the total stiffness matrix of the structure will result.

13/08/2019 32
 STEPWISE PROCEDURE FOR
PROGRAMMING

1.Label all elements (or members) 1…..m.

2.Label all joints 1….n, first unrestrained, then the

restrained ones. D.O.F associated with ith node: 3i-2,
3i-1, 3i. Hence, all d.o.f are also numbered.

3. Compute the size of structural stiffness matrix &

initialize it to 0.

13/08/2019 33
 STEPWISE PROCEDURE FOR
PROGRAMMING (Contd…)

4. Repeat for each element (from i=1 to m)

• Compute [k]L
• Compute [R] from end coordinates.
• Compute [k]G = [T]T[k]L[T]
• Establish association matrix (from node number of
the two nodes of member.
• Transfer each element of [kG] to appropriate
location of [K]TS
(kTS)ij = Σ(kG)mn
Extends over all members meeting at a joint.
m: Corresponds to ith dof and n to the jth dof.

Need to do this process 36 times for each

13/08/2019
member, no discount from symmetry….Why? 34
 STEPWISE PROCEDURE FOR PROGRAMMING
(Contd…)

5. Obtain nodal loads P

• Direct nodal loads
• Equivalent nodal loads (from member loads such as
distributed loads)

≡ +
Fixed ended
forces (to be Equivalent joint
added in the loads
end)

Equivalent joint load contributed by a member = − [T ]T { f } f

(Global coordinates)
13/08/2019 35
 STEPWISE PROCEDURE FOR PROGRAMMING
(Contd…)
⎡ P ⎤ ⎡ K PP K PX ⎤ ⎡ u P ⎤
6. Set up equations ⎢ X ⎥ = ⎢K K XX ⎥⎦ ⎢⎣u X ⎥⎦
⎣ ⎦ ⎣ XP

ux : prescribed displacements : corresponding to X (unknown reactions)

up : un-prescribed displacements: corresponding to P (known loads)

Let {u x} =0 (no support movement )

{P}= [kpp] {up}

∴ {up}= [kpp]-1{P} and hence {X}= [k xp] {up}

If not zero
Support
{P}= [kpp] {up} + [kpx] {ux}
settlement
{up}= [kpp]-1{P}-[kpx] {ux}

Reactions {X}=[kxp]{up} + [kxx]{ux}

13/08/2019 36
 STEPWISE PROCEDURE FOR PROGRAMMING
(Contd…)
7. Member end forces (go back to member level)
For each member (from 1 to m), do-
a) Using association matrix, get nodal displacements{D}(global
coordinates)
Eg, for this member
D1=u13
2 3
D2=u14 1
D3=u15
D4=u4
D5=u5 5 6
D6=u6 4

(b)[d]= [T][D]
Member end forces (c){f}= [k] L{d}
(d) Correction for fixed ended action
7 8 9

≡ +
{f}=[kL][d]+ [f] f

13/08/2019 37
 EFFICIENT STORAGE SCHEME FOR KPP
[Kpp] is a huge matrix. For example, for n= 100 unrestrained joints, it
will be 300x300 in size. i.e. 9x104 elements.

However, the fact is that it is symmetric and banded (why??)

R
(i,j) i <= j

[K]pp = [B] =

All elements outside the band are zero

Therefore, we only need to generate elements which are within the band. Generally, this
is achieved by storing right half band in a rotated rectangular matrix.

An element (i,j) in the original matrix will go to:

Row = i
Column = c=(j-i+1) in the banded matrix

Diagonal element kii : ki1

13/08/2019 38
Correlation between half banded and full [K]
Element (i, r) (i, c+i-1)
Diagonal (i, 1) (i, i)
HOW TO COMPUTE ‘R’ , THE HALF BAND WIDTH
„ Depends on structure size and also how we do numbering of joints
„ For each member find –
x = (max d.o.f. – min d.o.f) +1
„ The max value of x will be equal to “R”, the half band width. We need to store
[k]pp, [k]px, [k]xx
BASIS OF THIS
FORMULATION ??

Horizontal numbering
[k]pp = 45x45
Half band width = 18

Vertical numbering
[k]pp = 45x45
Half band width = 12

CONCLUSION??
13/08/2019 39
 INVERSE OF [K]PP: CHOLESKY’S ALGORITHM

To solve –
[P] = [kpp] [up] (if [ux]=0
or [P*] = [kpp][up] (if [ux] is not equal to 0)

Ku=P {let us relax notation}

Since K is symmetric
K = VT V
Where V is upper triangular matrix.

⎡ K1 K2 K3 K4 ⎤ ⎡V1 0 0 0 ⎤ ⎡V1 V2 V3 V4 ⎤
⎢K K6 K7 K 8 ⎥⎥ ⎢V V 0 0 ⎥ ⎢0 V V V ⎥
⎢ 5 = ⎢ 2 3 ⎥ ⎢ 5 6 7 ⎥

⎢ K9 K10 K11 K12 ⎥ ⎢V4 V5 V6 0 ⎥ ⎢ 0 0 V8 V9 ⎥

⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎣ K13 K14 K15 K16 ⎦ ⎣V7 V8 V9 V10 ⎦ ⎣ 0 0 0 V10 ⎦

[K] [VT] [V]

13/08/2019 40
 CHOLESKY’S ALGORITHM (contd….)
Ku = P
VT Vu =P
VT w = P where (w = Vu)

w1 = P1/V1
⎡V1 0 0 0 ⎤ ⎡ w1 ⎤ ⎡ P1 ⎤
⎢V V 0 0 ⎥ ⎢w ⎥ ⎢P ⎥
w2 = (P2-V2w1)/V3
⎢ 2 3

⎢V4 V5 0 0 ⎥
⎥ ⎢ 2⎥
⎢ w3 ⎥ = ⎢ 2⎥
⎢ P3 ⎥
⎢ ⎥
⎣V7 V8 V9 V10 ⎦
⎢ ⎥
⎣ w4 ⎦
⎢ ⎥
⎣ P4 ⎦
Similarly we can find w3,
w4 ,……wN

Further, Vu= w

⎡V1 V2 V3 V4 ⎤ ⎡ u1 ⎤ ⎡ w1 ⎤
⎢0 V V V ⎥ ⎢u ⎥ ⎢w ⎥
⎢ 5 6 7 ⎥ ⎢ 2⎥ = ⎢ 2⎥ On similar lines, we can
⎢ 0 0 V8 V9 ⎥ ⎢ u3 ⎥ ⎢ w3 ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ find uN ,uN-1 ,……u1
⎣ 0 0 0 V10 ⎦ ⎣u4 ⎦ ⎣ w4 ⎦

13/08/2019 41
 HOW TO OBTAIN [V] THE UPPER
TRIANGULAR MATRIX

V11 = K11
V1i =k1i/V11
Vii = [(Kii - ∑m =1 V2mi)]
i −1
i>1

∑
i −1
Vij = (kij - m =1
VmiVmj)/Vii j>i
Vij = 0 for i>j

V & VT both will be band matrices, with same bandwidth

as [k]pp. We can over write ‘V’ on [k]pp . Hence, no need to
create a new matrix.

13/08/2019 42
 INTERACTIVE EXERCISE
FORM ALL MATRICES FOR THE STRUCTURE
w = 12 kN/m, L = 6m
W

L L

13/08/2019 43
 HOW TO ANALYSE 3D
STRUCTURES
Option1: A space frame can be broken down into plane frames.

z
y

Member forces to
be superimposed
for columns
13/08/2019 44
 OPTION 2: MATRIX FORMULATIONS
FOR 3D STRUCTURES
Alternatively, the 2D formulations can be extended into 3D

Additional term (dof 4, 10) ⎡ GJ ⎤

T = ⎢ ⎥θ d10-d4
⎣ L ⎦

7 10
Local Rotations
x’ 1
4
Displacements
Displacements
z
Rotations y

Effect of slab in lateral load distribution Global

13/08/2019 ignored….. x 45
 ⎡ EA ⎤
⎢ L ⎥
⎢ 12 EI z ⎥
⎢ 0 ⎥
⎢ L3 ⎥
⎢ 12 EI y ⎥
⎢ 0 0
L3 ⎥
⎢ GJ ⎥
⎢ 0 0 0 ⎥
⎢ L ⎥
⎢ 6 EI y 4 EI y ⎥
⎢ 0 0 − 0 ⎥
L2 L
⎢ 6 EI z 4 EI z ⎥
⎢ 0 0 0 0 ⎥
⎢ L2 L ⎥
⎢− EA 0 0 0 0 0
EA ⎥
⎢ L L ⎥
⎢ 12 EI z 6 EI z 12 EI z ⎥
⎢ 0 − 0 0 0 − 0 ⎥
⎢ L3 L2 L3 ⎥
[K]L ⎢ 0 12 EI y 6 EI y 12 EI y ⎥
0 − 0 0 0 0
⎢ L3 L2 L3 ⎥
⎢ GL GJ ⎥
⎢ 0 0 0 − 0 0 0 0 0 ⎥
⎢ L L ⎥
⎢ 0 6 EI y 2 EI y 6 EI y 4 EI y ⎥
0 − 0 0 0 0 0
⎢ L2 L L2 L ⎥
⎢ 6 EI z 2 EI z 6 EI z 4 EI z ⎥
⎢ 0 0 0 0 0 − 0 0 0 ⎥
⎣ L2 L L2 L ⎦

13/08/2019 46
x’ → along centroidal axis of the member.
z’→ towards viewer.
y’→ can be ascertained by right hand system rule
(y’ and z’ should be along the principal axes of cross section)

kˆ = iˆxˆj
z
y
COORDINATE TRANSFORMATION x

⎡ x′ ⎤ ⎡ cos θ sin θ ⎤ ⎡ x ⎤ ⎡ l1 m1 ⎤ ⎡ x ⎤
For 2D ⎢ y′⎥ = ⎢− sin θ =
cosθ ⎥⎦ ⎢⎣ y ⎥⎦ ⎢⎣l2 m2 ⎥⎦ ⎢⎣ y ⎥⎦
⎣ ⎦ ⎣

l1,m1 : direction cosines of x’ axis w.r.t global system.

l2,m2 : direction cosines of y’ axis w.r.t. global system.

13/08/2019 47
 Direction cosines
⎡ x'⎤ ⎡ l1 m1 n1 ⎤ ⎡ x ⎤ l1, m1, n1 : x’
⎢ y '⎥ = ⎢l n2 ⎥⎥ ⎢⎢ y ⎥⎥
⎢ ⎥ ⎢2 m2 l2, m2, n2 : y’
⎢⎣ z ' ⎥⎦ ⎢⎣l3 m3 n3 ⎥⎦ ⎢⎣ z ⎥⎦ l3, m3, n3 : z’

R
R 0 0 0

[T] = 0
0
R
0
0
R
0
0
12 x 12
0 0 0 R
13/08/2019 48
 L = [(x2-x1)2 + (y2-y1)2 + (z2-z1)2]1/2
l1 = (x2-x1)/L
m1 = (y2-y1)/L
n1 = (z2-z1)/L

Unit vector along x’ = l1î +m1ĵ+n1k = iˆ'

We must specify the direction of y-axis

1. Unit vector along y’ = l2î + m2ĵ + n2k = ĵ'

OR

2. Two points along y’ : (x3,y3,z3) & (x4,y4,z4)so that we

can find: l2, m2, n2

Since x’y’z’ from right handed coordinate system, unit vector along z’

13/08/2019
kˆ' = iˆ'xˆj' 49
 z
y

13/08/2019 50
 ANALYSIS STEPS

• Fix member dimensions tentatively.

• Perform analysis.
• Check for adequacy of member sizes at
key locations.
• Revise dimensions if necessary

13/08/2019 51
 TEMPERATURE VARIATION
L L

Let there be uniform temperature change

∆T throughout the member
Determinate vs indeterminate structures,
any difference??
13/08/2019 52
 2 5 8
7
1 4
L L
6 9
3

L
11 14 17
10 13 16

12 15 18

Identify and mark the degrees of freedom

L
Unconstrained
length
L+∆L
13/08/2019 53
 2 5 8
7
1 4 ∆T
L L
6 9
3

L
11 14 17
10 13 16

12 15 18
Convert thermal effect into fixed ended forces
f
F ∆T

⎛ EA ⎞
13/08/2019 F =⎜
f
⎟ΔL ΔL = LαΔT F f = EAαΔT
⎝ L ⎠
54

APPLY OPPOSITE OF THE FEF ON THE STRUCTURE

 F f = EAαΔT

Solve the matrix equation as before

⎡
⎢
0 ⎤
⎥
and obtain displacements and
0
⎢
⎢ 0
⎥
⎥
member forces
⎢ ⎥
⎢ − EA α ΔT ⎥ Final member end forces can be
P=⎢ 0 ⎥
⎢ ⎥ obtained by superimposing the fixed
⎢ 0 ⎥
⎢ EAαΔT ⎥ ended condition with above solution
⎢ ⎥
⎢ 0 ⎥
⎢ ⎥
⎣ 0 ⎦ { f } = {K L }{d } + { f }F

13/08/2019
What would happen in case of 55
temperature fall??
EXTENSION TO LACK OF FIT
Longer member by
construction flaw

L L

Longer member : Analogous to temperature rise

Smaller member : Analogous to temperature fall
13/08/2019 56
HINGED/ GUIDED SUPPORTS

L L

13/08/2019 57
 Let us first consider all supports to
be fully rigid as treated so far……
8
2 5
7
1 4
L 6
9
KTS
3

L
11 14 P KPP
10 13 6x1 6x6
12 15
X KXX
9x1 9x9

13/08/2019 58
 Let us now altogether remove the
right support……
8
2 5
7
1 4
9
L 6
3

L
11 14 P KPP
10 13 9x1 9x9
12 15
X KXX
6x1 6x6

13/08/2019 P 7 = P8 = P9 = 0 59
 Let us now introduce the hinge

9
2 5
7
1 4
8
L 6
3

L
11 14 P KPP
10 13 8x1 8x8
12 15
X KXX
7x1 7x7
P7 = P8 = 0 (if no load acting at those points)
13/08/2019
u7 ≠ 0 and u8 ≠ 0 60
 WHAT HAPPENS WHEN THERE IS A
LOAD ACTING ALONG THE RELEASE?
Mo

P KPP KPX uP
8x1 8x8 8x1
=
X KXX ux
7x1 KXP 7x7 7x1

How would P change ? After solution, corresponding

P8 = Mo member end moment = Mo
 UDL IN ADJOINING SPAN
Convert the UDL into
equivalent joint loads

Any change in KTS ? YES/ NO

Any change in P, X? YES/ NO

P KPP KPX uP Upon solving,

8x1 8x8 8x1 we will get the
= values of u7 and
X KXX ux u8 P = ??
7x1 KXP 7x7 7x1 P2 = P5 = −
wL 3

2 P6 = ??
EXERCISE: Form the matrices P and X
62
 WHEN EQUIVALENT JOINT LOAD HAS
COMPONENT ALONG THE RELEASE
Convert UDL into
equivalent joint loads
wL
2

wL2
12
wL wL
2 2 P5 = -wL/2
W
2
wL wL2 P6 = -wL2/12
12 L 12 P8 = +wL2/12
In end, { f } = { K }{d } + { f }F
L
X9 also gets
Final moment at right end of member = 0 additional term
 ALTERNATE APPROACH
Convert UDL into equivalent
joint loads, but considering the
far end to be hinged.
3wL
5wL 8
8

wL2
5wL 3wL 8
8 8
W
2 P5 = -5wL/8
wL
P6 = -wL2/8
8 L
In end, { f } = {K L }{d } + { f }
F P8 = 0

Caution: Displacement correction needed in end. df to be added to

the displacement from the output
 INTERNAL HINGE

Hinge in beam only (not in column)

13/08/2019 65
 9
INTERNAL HINGE
12

8 11
10 13

2 6
1 3
1 2 5

3 4 Internal hinge 7
15 4 5 18
is tackled by
duplication of
14 17
the degree of
16 19
freedom

DOF (7) : Common for Members 2, 3, 5

DOF (3) : Common for Members 1, 4
13/08/2019 66
DOF (4) : Member 2 ONLY
 INTERNAL HINGE
Terms corresponding to DOF (4) will get contribution
from member 2 only.
Take care of the DOF in code number approach format

Association matrices:

⎡1 ⎤ ⎡14⎤
⎢ 2⎥ ⎢15⎥
⎢ ⎥ ⎢ ⎥
⎢ 4⎥ ⎢16⎥
A[2] = ⎢ ⎥ A[4] = ⎢ ⎥ P KPP KPX uP
⎢5 ⎥ ⎢1⎥ 7x1 7x7 7x1
⎢6 ⎥ ⎢2⎥ =
⎢ ⎥ ⎢ ⎥ X KXX ux
⎢⎣ ⎥⎦
7
13/08/2019
⎢⎣ 3 ⎥⎦
12x1 KXP 12x12 12x1 67
 INTERNAL HINGE 12WITH UDL
9
P1 = 0
8 11
10 13
P2 = -wL/2
2
1 6
3 P3 = 0
1 2 P4 = -wL2/12
7
5 P5 = 0
3 4 Influence of
15 4 5 18 P6 = -wL/2
UDL absorbed
by (4) only 19 P7 = +wL2/12
14 16 17

wL
wL W
wL2 2 2 wL2
13/08/2019
12 L 12 68
 INTERNAL HINGE 12WITH UDL
9
P1 = 0
8 11
10 13
P2 = -wL/2
2
1 6
3 P3 = 0
1 2 P4 = -wL2/12
7
5 P5 = 0
3 4 Influence of
15 4 5 18 P6 = -wL/2
UDL absorbed
by (4) only P7 = +wL2/12
14 17

16 19 ALTERNATE
APPROACH ONE
wL
2
wL W SIDE HINGED……
wL2 2 2 wL2
13/08/2019
12 L 12 69
 INTERNAL HINGE 12WITH UDL
9
P1 = 0
8 11
10 13
P2 = -3wL/8
2
1 6
3 P3 = 0
1 2 P4 = -0
7
Influence of 5 P5 = 0
3 4
15 4 UDL absorbed 5 18 P6 = -5wL/8
by (4) only
19
14 P7 = +wL2/8
{ f } = {K L }{d } + { f }F17
ALTERNATE
2
W 6 APPROACH ONE
1 SIDE HINGED……

4 7
L 5
13/08/2019 70
HINGE THROUGH COLUMN
AND BEAM

3
5

Independent DOF for all

members meeting at the joint,
rest of the procedure same. 71
 INTERNAL HINGE : ALTERNATE
APPROACH- TO MODIFY [K]L

2
5
1
4
3 6
Permanent hinge
This boundary condition is not to be altered
while deriving the member stiffness matrix
Hinge not to be fixed while deriving the
member stiffness matrix
72
 INTERNAL HINGE : ALTERNATE
APPROACH- TO MODIFY [K]L
EA
L
3EI
0
L3
[K]L = 0
EA
0 0
EA
− 0 0
L L
3EI 3EI
0 − 3 0 0
L L3
3EI 3EI 3EI
0 0 0 − 2
L2 L L

3rd row and 3rd column: All terms zero

13/08/2019 73
Let us derive second column
 ALTERNATE APPROACH:
INTERNAL HINGE

Use the code number approach, but

here there is no duplication of the
13/08/2019 DOF as in the earlier approach. 74
 8 11
7 10
9 12

2
1 5 3
1 2 4

3 4 KPP KPX 5 6
14 4 17
6x6
16
13
15 KXX 18
KXP 12x12
2
Automatically, member 2 will not make any contribution in the third
row or column of KTS. Members 1 and 4 will make contribution as
before
The displacement corresponding to DOF 3 remains
unknown for member 2. Corresponding
13/08/2019 displacement of the column can be obtained 75
 ALTERNATE APPROACH: HINGE
THROUGH BEAM AND COLUMNS BOTH

13/08/2019 76
 ALTERNATE APPROACH: HINGE
THROUGH BEAM AND COLUMNS BOTH
8 11
7 10
9 12

2
1 6 3
1 2
5 4
3
14 4 5 17
18
13 16
15

13/08/2019

No duplication of DOF as before…… 77

 IMPLICATIONS
All three members 1, 2 and 3 have modified [K]L
After KTS is formed, we will find the third row and third
column to be zero. WHY????
The diagonal element of KTS (3,3) shall be ZERO. This
would imply the matrix to be singular, |KTS | =0, hence, we
will encounter run time error.
To circumvent this situation, eliminate the DOF (3).
Renumber the DOFs and skip numbering this DOF.

13/08/2019 78
 ALTERNATE APPROACH: HINGE
THROUGH BEAM AND COLUMNS BOTH
7 10
6 9
8 11

2
1 4 3
1 2
3 5

13 4 5 16
17
12 15
14

13/08/2019

Need to skip the DOF corresponding to rotation 79

(displacement output will be devoid of the values of these)
PRACTICE EXERCISE
<1>

80
 PRACTICE EXERCISE
<2> 500 kNm
500 kN
L L

81
PRACTICE EXERCISE
<1>

82
 PRACTICE EXERCISE
<2> 500 kNm
500 kN
L L

83
PRACTICE EXERCISE
<3>

84
PRACTICE EXERCISE
<4>

85
 P KPP up
6x1 6x6 6x1
=
X KXX ux
6x1 6x6 6x1

− X 11 Alternate approach:
u11 =
k Add one more member (a
link element)
13/08/2019 86
 PRACTICE EXERCISE
<5>
200 kN
15 kN/m 15 kN/m
= =
4m 4m 4m

87
 TREATMENT OF
NON-PRISMATIC MEMEBERS
Case I: Determinate Structures
P

For a determinate structure, both

member end forces as well as
deflections can be directly
calculated…..
88
 SOLUTION BY PRINICIPLE OF
1
VIRTUAL WORK
P

δ
BMD for the
M (x) actual load

BMD for unit load at the

m(x) point of displacement

M ( x ) m( x )
δ= ∫
To take care of the non-prismatic
dx nature of the member
EI
Conclusion: For determinate structures, both member end forces and deflections can 89
be easily computed by incorporating the variation of EI
 TREATMENT OF
NON-PRISMATIC MEMEBERS
Case II: Indeterminate Structures

How to determine KL??

Any issue? 90
 TREATMENT OF
NON-PRISMATIC MEMEBERS
Case II: Indeterminate Structures
2 5
3
1 4

6
In order to derive stiffness matrix, as per first principles, we
need to apply unit displacement along a particular DOF
keeping all other displacements zero
Basic slope-deflection formulations
no longer valid. Use of classical
M Θ=1 force method too tedious..
???
Being indeterminate, principle of 91
Virtual Work cannot be applied
 TREATMENT OF
NON-PRISMATIC MEMEBERS
Case II: Indeterminate Structures

How to determine KL?? Any issue?

We have to use indirect approach,
employing flexibility method 92
 WE WILL UTILIZE THE
FLEXIBILITY METHOD
Fij =
Displacement along the line of action of the
ith force when we apply unit force along the
line of action of the jth force
(no force acting along the lines of action of
other designated forces=> no restraint)
Unlike the stiffness approach (which emphasizes on
locking remaining displacements), this process creates
a determinate structure…
How this is done?....see the next step.
13/08/2019 93
 USE OF FLEXIBILITY METHOD

F33 F63
1

Apply unit force along “3”, no force to be applied along

other force lines.

Hence, no other force is generated, except reactions. The

structure is determinate, so that we may easily apply the
principle of virtual work

13/08/2019 94
APPLY PRINCIPLE OF VIRTUAL WORK

F63 Real
1 F33 system

1 Virtual
system

EXTERNAL VIRTUAL WORK = INTERNAL VIRTUAL WORK

L L 2
M ( x ) m( x ) [m1 ( x)]
1.F33 = ∫0
EI
dx F33 = ∫
0
EI
dx

13/08/2019 95
 SIMILARLY….
L 2
[m2 ( x)]
F66 =
∫
0
EI
dx
L
m1 ( x)m2 ( x)
F36 = F63 =
∫
0
EI
dx

BUT THIS IS NOT WHAT

WE FINALLY WANT
13/08/2019 96
 WE USE THE PRINCIPLE OF
SUPERPOSITION TO GET FINAL SOLUTION
F66
1
F33 F63 F36
1

A B
Combine (A) and (B) in following fashion:

⎡ F63 ⎤
A− ⎢ ⎥ B
⎣ 66 ⎦
F
13/08/2019 97
 1

⎡ F36 F63 ⎤ ⎡ F63 ⎤

⎢ F33 − ⎥ ⎢− ⎥
⎣ F66 ⎦ ⎣ F66 ⎦

Hence, by definition

1 F66
K 33 = = 2
⎡ F36 F63 ⎤ F33 F66 − F36
⎢ F33 − ⎥
⎣ F66 ⎦
13/08/2019 98
Similarly,
F33
K 66 = 2
F33 F66 − F36
− F63
K 63 = 2
F33 F66 − F36

13/08/2019 99
 HOW TO DERIVE K22

BY APPRORIATE COMBINATION OF

1θ
θ

A B
Such that the net angle of rotation on the
13/08/2019
left end of the beam is ZERO 100
 HOW TO DERIVE K11
P
Δ
Apply the Principle of Virtual Work
External Virtual Work = External Virtual Work
L

1.Δ = ∫
0
FV ε R dx

L
⎛ FR ⎞
13/08/2019
0
∫
= FV ⎜
⎝
⎟dx
AE ⎠ 101
 NON-PRISMATIC MEMEBERS
How to obtain fixed ended forces?

Release the restrainsts and convert the structure

into a determinate system

θ1 θ2

102
 By superimposing the A, B, C:

θ1 θ2 A

θ1
K33θ1
K63θ1
B
θ2
K66θ2
C
103
K36θ2
 PRACTICE EXERCISE

2A
A
2I
I

Derive the stiffness matrix

13/08/2019 104
 INCLUSION OF SHEAR
DEFORMATION EFFECT
Necessary for deep sections L/D <= 6

13/08/2019 105
 INCLUSION OF SHEAR
DEFORMATION EFFECT

Treatment shall be restricted to prismatic

sections only

In any deformed member, strain energy is

given by

1
U= ∫∫∫
V
2
σεdV
13/08/2019 106
 1
U=
V
2 ∫∫∫
σεdV
⎛
1⎜ M ⎟ 2 ⎞ ⎛
1⎜ F ⎟2 ⎞
U=
L
∫
2 ⎜⎝ EI ⎟⎠
dx +
L
∫
2 ⎜⎝ EA ⎟⎠
dx

1⎛ V ⎞ 2
+ s ∫ ⎜⎜ ⎟⎟dx
2
L ⎝
GA ⎠
s = Form factor or shear correction factor
or shear deformation coefficient
13/08/2019 107
 The shear area of the member is defined
as the area of the section which is
effective in resisting shear deformation.

Form factor is defined as the ratio of the

gross area of the section to the shear area
of the section

1⎛ V ⎞ 2
U shear = s ∫ ⎜⎜ ⎟⎟dx S> 1

L
2 ⎝ GA ⎠
13/08/2019 108
 A*
VQ
y τ=
N A Ib

Q( y) = ∫ y dA
A*

First moment of part of

cross-section above the
13/08/2019
section considered 109
 A ⎛Q 2
⎞
s= 2
I ∫A
⎜⎜ 2
⎝b
⎟⎟dA
⎠

A*
s = 1.2
y s = 10/9
N A
s=2

Practice: Find ‘s’ for a rectangular section bx2D

13/08/2019
110
13/08/2019 111
 2 5
3
1 4
DEEP
6
1 F33

BMD m1 ( x) SFD 1/ L

F66
1

BMD m2 ( x ) SFD 1/ L 112

112
 Applying the principle of Virtual Work,

m1 ( x)m1 ( x) + s V1 ( x)V1 ( x) dx
F33 = ∫
EI
dx ∫
L
GA
L

⎛ x⎞ 1
m1 ( x) = ⎜1 − ⎟ V1 ( x) =
⎝ L⎠ L
Similarly,

m1 ( x)m2 ( x) V1 ( x)V2 ( x)
F63 = ∫ dx + s ∫ dx
L
EI L
GA
13/08/2019
113
 Similarly get F66

m2 ( x)m2 ( x) V2 ( x)V2 ( x)
F66 = ∫ dx + s ∫ dx
L
EI L
GA
x 1
m2 ( x) = V2 ( x) =
L L
and
m1 ( x)m2 ( x) V1 ( x)V2 ( x)
F63 = ∫ dx + s ∫ dx
L
EI L
GA
13/08/2019
114
 m2 ( x)m2 ( x) V2 ( x)V2 ( x)
F66 = ∫ dx + s ∫ dx
L
EI L
GA

1
K 33 =
⎡ F36 F63 ⎤
⎢ F33 − ⎥
⎣ F66 ⎦

13/08/2019 115
 From slide18

In computer program, we need not store entire [T],

we may simply store [R]

⎡ RT 0 ⎤ ⎡K A K B ⎤ ⎡R 0 ⎤ ⎡ RT K A R RT K B R ⎤
[k]G= [T]T[k]L[T] = ⎢ ⎥ ⎢ ⎢ ⎥ =⎢ T ⎥
⎣0 RT ⎦ ⎣ K B K C ⎥⎦ ⎣ 0 R ⎦ ⎣R K B R RT K C R ⎦

[KL]

13/08/2019 116
 DIRECT STIFFNESS METHOD: ASSUMPTIONS

1. Restricted to frame and truss structures (skeletal structures) only. Members

assumed as line elements (passing through neutral axis) with lumped
sectional properties. At first, we restrict analysis to prismatic members only.

2. Hooke’s law of elasticity holds.

3. Small deflections => no change in overall geometry of structure.

4. Plane sections remain plane after bending.

5. In bending mode, very small slope

d 2 y / dx 2 d2y
≈
[1 + (dy / dx ) ]
Curvature = 2 3/ 2
dx 2
6. If displacement takes place normal to member, no change in length of the member.
Change in length of an element due to flexural deformation (curvature effects) is
also negligible.
13/08/2019 117