Dr.G.Kumaran, Professor, Dept.

of Structural Engg dept, FEAT AU

DIRECT INTEGRATION METHODS

Direct Integration methods provide an alternative to the mode superposition methods, and
is particularly suitable in the analysis of non-linear systems where super position is
generally not valid. In direct integration, the second-order differential equations of motion
are integrated by a numerical step-by-step procedure. Unlike mode superposition
methods, direct integration methods do not call for a transformation of the equations into
principle coordinates, hence the term ”direct’.

mx  cx  k x  F (t )

Direct Integration methods are based on two basic concepts: (i) The above equation
requires to be satisfied at certain discrete time instants (intervals  t apart), and not at
any time t. (ii) A suitable variation of displacements, velocities and accelerations is
assumed in each time intervals  t .

Direct Integration methods may be classified as either EXPLICIT or IMPLICIT. An

“explicit” method is defined as one in which the new response values calculated in each
step depend only on quantities obtained from the preceding steps; hence the analysis
proceeds from one step to the next.

In an “Implicit” method, on the contrary, the expressions giving the new response values
in each step include one or more unknown values pertaining to the same step; hence trial
values of these quantities need to be assumed and to be refined by successive iterations.
Alternatively, it is possible to convert the “implicit” formulation to an “explicit;
formulation for ease in computation.

Central difference method (explicit)

Implicit methods converted to explicit methods (Honbolt, Linear acceleration, Wilson” 

”, New marks  methods)
Linear acceleration method:

If the acceleration is assumed vary linearly in the interval  t . It is convenient to use the
incremental displacement as the basic variable, rather than incremental acceleration (this
brings to convert to explicit form).

Incremental equilibrium equations:

mx(t )  cx (t )  k x(t )  F (t ) (1)

By linear acceleration assumptions,
x(t )  6
x(t )  6 x(t )  3x(t ) (2)
(t ) 2
(t )

x(t )  3 x(t )  3x(t )  t x(t ) (3)

(t ) 2
Thus, the equilibrium equations can be reduced top the following simplified form.

1
Dr.G.Kumaran, Professor, Dept. of Structural Engg dept, FEAT AU

 
 k (t )  m  3 c(t )x(t )
6
 (t ) 2
(t )
 
 6
= F (t  m x(t )  3x(t )  c(t )3x(t )  t x(t ) (4)
 (t )   2 
which is in a standard static-stiffness from:

k eff (t ) x(t )  F 

eff (t )

The solution for the incremental displacements x(t ) is obtained by Gauss or Choleski
decomposition, which needs to be done at each time step. After determining x(t ) ,
velocities x (t ) and accelerations x(t ) are computed using equations (2) and (3) and
the initial conditions for the next step are obtained thus:

x(t  t )  x(t )  x(t )

x (t  t )  x (t )  x (t )
x(t  t )  x(t )  x(t )

Note: (1) At the first time step x(0) and x (0) are known and x(0) is obtained by
solving

mx(0)  cx (0)  k x(0)  F (0)

(2) for good accuracy, the choice of the time step should be such that the lowest period of
vibration should be at least 5 to 10 times  t

(3) The linear acceleration method is only conditionally stable, and will blow up the
results if  t is not sufficiently small.

WILSON “  ” METHOD

It is a modification of the linear acceleration method; it guarantees unconditionally

stability, provided the assumption of linear acceleration id extended beyond 0    t to
0    t , where   1.37 , i.e the time step is extended to t . Accordingly, the same
series of equations are acceptable here, except that  t shall in all cases be replaced by
 t .

x(t )  6
x(t )  6 x(t )  3x(t ) (5)
(t ) 2
(t )
k eff (t ) x(t )  F 
eff (t )

 
 k (t )  6
m  
3
c (t ) x(t )
   2
 (t ) 
 (t ) 

2
Dr.G.Kumaran, Professor, Dept. of Structural Engg dept, FEAT AU

 6
= F (t  m x(t )  3x(t )  c(t )3x(t )  t x(t ) (6)
 (t )   2 

Note: the acceleration at the next normal time step is given by linear interpolation as
x(t )  1 x(t )

and
x(t )  tx(t )  t x(t )
2
x(t )  tx (t )  t x(t )  t x(t )
2 3
Thus the staring values for the next step are
2 6
determined.

NEWMARK  METHOD

Newmark proposed the above equations for approximating the velocity and
displacements; by generally certain numerical integration methods that had been in use
up-to that time.

x (t  t )  x (t )  1   x(t )  x(t  t )t (6)

x(t  t )  x(t )  x(t )t  1 / 2  x(t )  x(t  t )t 2 (7)

The parameter  in equation (6) produces numerical damping within the time step  t . If
  1 / 2 , an artificial negative damping results; If   1 / 2 , an artificial positive damping
results; amd if   1 / 2 , numerical damping becomes equal to zero and equation (6)
reduces to the trapezoidal rule.
The parameter in equation (7) controls the variation of acceleration within the time step.
For this reason, the techniques is referred to a s Newmark’s generalised acceleration
method or Newmark  method. If   0 results in the constant acceleration method; If
  1 / 4 results in the average acceleration method; If   1 / 6 results in the well-known
linear acceleration method. In Newmark’s original proposal   1 / 2 and   1 / 4 were
recommended for un unconditionally stable scheme.
x(t  t )  x(t )  t x(t )  x(t  t ) (8)
2
x(t  t )  x(t )  x (t )t  t x(t )  x(t  t )
2
(9)
2
From (6) and (7),
x(t  t )  x(t )  t x(t )  x(t  t ) (8)
2
x(t  t )  x(t )  x (t )t  t x(t )  x(t  t )
2
(9)
2