Solution of Equilibrium

Equations in Dynamic Analysis

Kelompok 3 :
 Brian Widyan Hadi 18/434841/PTK/12404
 Ratih Dewanti 18/434868/PTK/12431
 Sanggabuana Satria K 18/434871/PTK/12434
 Tirana Novitri S 18/434874/PTK/12437
 Tri Mulyanto 18/434875/PTK/12438
 Ulul Azmy 18/434879/PTK/12442
 Yohanes Meinata 18/434883/PTK/12446

MAGISTER TEKNIK SIPIL

UNIVERSITAS GADJAH MADA
YOGYAKARTA
 Direct integration methods

System of equilibrium equations in linear dynamics

 M U   C U    K U   R

The direct integration consists of integrating the

equilibrium equations using a numerical step-by-step
procedure. The “direct” term means that prior to the
numerical integration, no transformation of the equations
into a different form is carried out.
 Direct integration methods

• Central difference method

• Houbolt method
• Wilson-Θ method
• Newmark method
The central difference
method
 The central difference method

System of equilibrium equations in linear dynamics

 M U t  C U t   K U t  Rt

Aproximation for aceleration and velocities

U   2 U t t  2 U t  U t t 

1
t t

U t

1
2t
  U t t  U t t 
 The central difference method

Aproximation for aceleration and velocities

2  t t
U    2 U t  U t t 
1
U
t t

U  t

1
2t
  U t t  U t t 

Substitution of the equations above into the equilibrium

equation leads to the following expression:
 1     1
C   U t t
1 2 1
 2  M    C   U t t  Rt    K   2  M   U t   2 M  
 t 2 t   t   t 2t 
 The central difference method
Alternative way of presenting the equilibrium equation
 1     1
C   U t t
1 2 1
 2  M    C   U t t  Rt    K   2  M   U t   2 M  
 t 2 t   t   t 2t 

Defining the auxiliary matrices,

 M̂   a0  M   a1 C 
 
 Rˆ    R    K   a2  M  U    a0  M   a1 C  U 
 t t t t t

with,
a0  1/ t 2 , a1  1/ 2t , a2  2a0 , a3  1/ a2
 The central difference method

Alternative way of presenting the equilibrium equation

 1     1
C   U t t
1 2 1
 2  M    C   U t t  Rt    K   2  M   U t   2 M  
 t 2 t   t   t 2t 

Using the definitions presented previously we obtain, the

following system of equations

Rˆ  solve for U 

1
U t t   Mˆ 
t t t
 The central difference method

Having {U}t+Δt determined, compute acelerations and

velocities at time t
U   a0 U t t  2 U t  U t t 
t

U   a1   U t t  U t t 
t

The calculation of {U}t+Δt involves {U}t & {U}t-Δt.

Therefore a starting procedure must be used to compute
the solution at time Δt. The equation below is used for
this purpose:
t 2
U t  U 0  t U   U
0 2 0
 The central difference method

Critical time step computation to ensure numerical

stability of the algorithm:

t 
2
H
    M   0 ,
K   2

H  max 1 , 2 ...n 

 CDM flowchart implementation
Input:
 K  ,  M  , R(t )
U 0 , U 0 , U 0 , n=1, t f

Compute t , U t

n  n 1 t  nt

NO

Rˆ
1
U t t   Mˆ  t
t  tf ?
U   a0 U t t  2 U t  U t t 
t

YES U   a1   U t t  U t t 

t

U t t  U t
STOP
U t  U t t
The central difference method - Example

k1 k2
M1 M2

U1(t), F1(t) U2(t), F2(t)

 The central difference method - Example

 2 0  U1   6 -2  U1  0 
 0 1      2 4  U   10
  U 2    2  

Time step computation:

 6 -2 2 2 0  1  0 1  1.4142 

               
2 4  0 1  2  0 2  2.2361

2
t 
2
The central difference method - Example
3.5
Central difference method
Exact solution
3

2.5

2
U1(t)

1.5

0.5

0
0 5 10 15 20 25
Number of time steps
The central difference method - Example
6

4
U2(t)

2
Central difference method
1 Exact solution

0
0 5 10 15 20 25
Number of time steps
The Houbolt method
 The Houbolt method

System of equilibrium equations in linear dynamics

 M U t t  C U t t   K U t t  Rt t

Aproximation for aceleration and velocities

U   2 2 U t t  5 U t  4 U t t  U t  2 t 
1
t t t

U 
t t

1
6t
11U t t  18U t  9 U t t  2 U t  2 t 
 The Houbolt method

Substitution of the approximations for velocity and

aceleration leads to the following form:
 2 11   5 3 
 2  M        t t  t t  2      U t
C  K U  R  M  C
 t 6t   t t 
 4   1
C   U t t   2  M   C   U t 2t
3 1
  2 M  
 t 2t   t 3t 
 The Houbolt method

Defining the auxiliary matrices,

 K̂    K   a0  M   a1 C 
 

  t t   R t t   M   a2 U t  a4 U t t  a6 U t  2 t 

 Rˆ 

 C   a3 U t  a5 U t t  a7 U t  2 t 

with,
a0  2 / t 2 , a1  11/ 6t , a2  5 / t 2 , a3  3 / t , a4  2a0
a5  a3 / 2, a6  a0 / 2, a7  a3 / 2
 The Houbolt method

Defining the auxiliary matrices,

 K̂    K   a0  M   a1 C 
 

  t t   R t t   M   a2 U t  a4 U t t  a6 U t  2 t 

 Rˆ 

 C   a3 U t  a5 U t t  a7 U t  2 t 

Resultant system of equations to be solved,

 Kˆ  U   Rˆ
  t t  t t
 solve for U t t
 The Houbolt method

Having {U}t+Δt determined, compute acelerations and

velocities at time t+Δt
U t t
 a0 U t t  a2 U t  a4 U t t  a6 U t  2 t

U t t
 a1 U t t  a3 U t  a5 U t t  a7 U t  2 t

The calculation of {U}t+Δt involves {U}t, {U}t-Δt & {U}t-2Δt

Therefore a starting procedure must be used to compute the
solution at time Δt and 2Δt. The CDM may be used for
this purpose
 The Houbolt method

Critical time step computation to ensure numerical

stability of the algorithm:

• There is no critical time-step limit to integrate the

equilibrium equations using the Houbolt method!
However, the starting procedure requires the same
critical time step used in the CDM!

t 
2
H
    M   0 ,
K   2

H  max 1 , 2 ...n 

 HM flowchart implementation
Rˆ
1
Input: U t t   Kˆ 
t t
 K  ,  M  , R(t )
U   a0 U t t  a2 U t  a4 U t t  a6 U t  2 t
U 0 , U 0 , U 0 , n=1, t f t t

U  t t
 a1 U t t  a3 U t  a5 U t t  a7 U t  2 t

Compute t , U t U t 2 t  U t t

U t t  U t
t  nt
U t  U t t

NO
n  2?
NO
n  n 1 t  tf ?
YES
YES
COMPUTE U t & U 2 t
using CDM STOP
 The Houbolt method - Example

 2 0  U1   6 -2  U1  0 
 0 1      2 4  U   10
  U 2    2  

Time step computation:

 6 -2 2 2 0  1  0 1  1.4142 

               
2 4  0 1  2  0 2  2.2361

2
t 
2
 The Houbolt method - Example
3.5
Exact solution
Houbolt method
3

2.5

2
U1(t)

1.5

0.5

0
0 5 10 15 20 25
Number of time steps
 The Houbolt method - Example
6

4
U2(t)

Exact solution
1 Houbolt method

0
0 5 10 15 20 25
Number of time steps
The Wilson-Θ method
The Wilson-Θ method (linear acceleration method)

System of equilibrium equations in linear dynamics

 M Ut t  C Ut t   K U t t  Rt t

U
t
U  t t
U 
t t

 t  t t  t
t
0    t
The Wilson-Θ method (linear acceleration method)

Assumed acceleration for the time interval t to t+ΘΔt:

U
t
U  t t
U 
t t

 t  t t  t
t

U t 
 U  
t t

U
t t
 U 
t

The Wilson-Θ method (linear acceleration method)

Assumed acceleration for the time interval t to t+ΘΔt:


U  t 
 U  
t t

U
t t
 U 
t

Integrating the expression above we obtain:

2
U  t 
 U   U   
t t 2t
U 
t  t
 U 
t


 
3
 U t  U    U   2  U  U 
1
U t  t 2 t 6t t t t
The Wilson-Θ method (linear acceleration method)

Using the previous expressions at time t+ΘΔt we obtain:

t
U t t
 U  
t 2
U 
t t
 U 
t

 2 t 2
U t t  U t  t U t  
6
U  t t
 2 U 
t

From which we can solve for acceleration and velocities
in terms of {U} t+ΘΔt:

U   2 2 U t t  U t   U   2 U 
6 6
t t  t t t t

t
U  t t

3
t
 U t t  U t   2 U t  U t
2
The Wilson-Θ method (linear acceleration method)

The expressions below are then substituted into the

dynamic equilibrium equation

U  2  t t  t  U   2 U 
6 6
 U  U 
t t  t
2
t t t

t
U 
t t

3
t
 U t t  U t   2 U t  U t
2

 M Ut t  C Ut t   K U t t  Rt t

The Wilson-Θ method (linear acceleration method)

Defining the auxiliary matrices,

 K̂    K   a0  M   a1 C 
 

 Rˆ    
  t t   R t    R t t   R t   M  a0 U t  a2 U t  2 U t 
 C   a U   2 U   a U  
1 t 3
t t

with,
a0  6 /(t )2 , a1  3/ t , a2  2a1 , a3  t / 2, a4  a0 / 
a5  a2 /  , a6  1  3/  , a7  t / 2, a8  t 2 / 6
The Wilson-Θ method (linear acceleration method)

Leads to the following expression for {U}t+ΘΔt

 Kˆ  U 
  t t
 
ˆ
R
t t
 solve for U t t

With {U}t+ΘΔt compute acceleration, velocity and

displacement at t+Δt:

U   a4 U t t  U t   a5 U   a6 U 
t t t t

U t t t

 U   a7 U 
t t
 U 
t

U t t  U t  t U t  a8 U t t  2 U t 
The Wilson-Θ method (linear acceleration method)

Critical time step computation to ensure numerical

stability of the algorithm:

t 
2
H
    M   0 ,
K   2

H  max 1 , 2 ...n 

• The Wilson-Θ algorithm is conditionally stable for Θ

values greater than 1.37, typically 1.40!
 W-Θ flowchart implementation
Input:
 K  ,  M  , R(t )
U 0 , U 0 , U 0 , n=1, t f

Rˆ
1
Compute t U t t   Kˆ 
t t

U   a4 U t t  U t   a5 U   a6 U 
n  n 1 t  nt t t t t

NO
U  t t
 U   a7 U 
t
 t t
 U 
t

t  tf ? U t t  U t  t U t  a8 U t t  2 U t 
U   U 
t t t

U   U 
YES
t t t

STOP U t t  U t
U t  U t t
 The Wilson-Θ method - Example

 2 0  U1   6 -2  U1  0 
 0 1      2 4  U   10
  U 2    2  

Time step computation:

 6 -2 2 2 0  1  0 1  1.4142 

               
2 4  0 1  2  0 2  2.2361

2
t 
2
 The Wilson-Θ method - Example
3.5

3 Wilson-theta method
Exact solution

2.5

2
U1(t)

1.5

0.5

0
0 5 10 15 20 25
Number of time steps
 The Wilson-Θ method - Example
6

4
U2(t)

2
Wilson-theta method
Exact solution
1

0
0 5 10 15 20 25
Number of time steps
The Newmark method
 The Newmark method

System of equilibrium equations in linear dynamics

 M U t t  C U t t   K U t t  Rt t

Aproximation for velocity and displacements

U  t t
 U   1    U    U   t
t  t t t 

 1   2
U t t  U t  U  t      U    U   t
t
 2  t t t

 The Newmark method

U  t t
 U   1    U    U   t
t  t t t 
(I)

 1   2
U t t  U t  U  t      U    U   t (II)
t
  2  t t t

Solving (II) for acceleration at t+Δt and then substituting
the acceleration at t+Δt into (I) and the result into the
equilibrium equation we obtain,

 Kˆ  U    Rˆ 
  t t   t t
 The Newmark method

 Kˆ  U    Rˆ   solve for U 
  t t   t t t t

Where,
 Kˆ    K   a0  M   a1 C 
 

Rˆ
t t

 Rt t   M  a0 U t  a2 U   a3 U  
t t

C   a1 U t  a4 U t  a5 U t 

 Rˆ    R    K   a2  M  U    a0  M   a1 C  U 
 t t t t t
 The Newmark method

with,

a0  1/ t 2 , a1   / t , a2  1/ t , a3  1/ 2  1, a4   /   1

a5   t / 2  /   2  , a6  t 1    , a7  t

  0.50,   0.25  0.5+   Values that ensure numerical stability

2
 The Newmark method

Having {U}t+Δt , velocity and acceleration are computed

as follow,

U   a0 U t t  U t   a2 U   a3 U 
t t t t

U t t
 U   a6 U   a7 U 
t t t t
 Newmark flowchart implementation
Input:
 K  ,  M  , R(t )
U 0 , U 0 , U 0 , n=1, t f

Rˆ
1
Compute t U t t   Kˆ 
t t

n  n 1 t  nt
U   a U   U    a U   a U 
t t 0 t t t 2 t 3 t

NO
U   U   a U   a U 
t t t 6 t 7 t t

t  tf ?
U   U 
t t t

U   U 
t t t
YES U t  U t t
STOP
 The Newmark method - Example

 2 0  U1   6 -2  U1  0 
 0 1      2 4  U   10
  U 2    2  

Time step computation:

 6 -2 2 2 0  1  0 1  1.4142 

               
2 4  0 1  2  0 2  2.2361

2
t 
2
 The Newmark method - Example
3.5

3
Newmark method
Exact solution
2.5

2
U1(t)

1.5

0.5

0
0 5 10 15 20 25
Number of time steps
 The Newmark method - Example
6

4
U2(t)

2
Newmark method
1 Exact solution

0
0 5 10 15 20 25
Number of time steps
Error analysis
 Error analysis - Example

 2 0  U1   6 -2  U1  0 
 0 1      2 4  U   10
  U 2    2  

Time step computation:

 6 -2 2 2 0  1  0 1  1.4142 

               
2 4  0 1  2  0 2  2.2361

2
t 
2
 Error analysis
3.5

3 Central difference
Exact solution
Houbolt
2.5 Wilson-theta
Newmark
2
U1(t)

1.5

0.5

0
0 5 10 15 20 25
Number of time steps
 Error analysis
6

4
U2(t)

2 Central difference
Exact solution
Houbolt
1 Wilson-theta
Newmark

0
0 5 10 15 20 25
Number of time steps
 Error analysis

Relative solution error definition:

U Exact  U Num
  %  100
U Exact
 Error analysis
3
Central difference
Houbolt
2 Wilson-theta
Newmark
Relative error (%)

-1

-2

-3
0 5 10 15 20 25
Number of time steps
Thank You