4b - Implementation - Dynamics and Vibrations

You might also like

Download as ppt, pdf, or txt
Download as ppt, pdf, or txt
You are on page 1of 24

CEG8315

Advanced structural modelling

4. Finite elements in dynamics and vibrations

4b. Time history analysis

School of Engineering

School of Engineering
Natural Frequencies and Mode Shapes.

The Eigenvalue Problem. An undamped structure, with no external loads applied unrestrained d.o.f.,
undergoes harmonic motion (caused perhaps by initial conditions) in which each d.o.f. moves in phase with
all other d.o.f.. Thus,

where are the amplitudes of nodal d.o.f. vibration and  is the circular frequency ( radians per second).
The cyclic frequency ( in Hertz) is f =  / 2  and the period is T = 1 / f (seconds). Both  and f are called
simply “frequency” and relate to undamped motion unless otherwise stated.
   ext ext
Combining the above expressions for D and D with M D  C D  K D  R , and with C and R both
zero, we have,
K  M D  0, wher e    2
This is the basic statement of the vibration problem. The above equation a generalised eigenproblem or
simply an eigenproblem. When the matrix K  M is singular, then this equation has only the trivial solution
D 0
. We are interested in the non-trivial solutions and hence wish to determine the eigenvalues  that
satisfy: det K  M   0

Associated with each eigenvalue i is an eigenvector Di .


The lowest non-zero i is the fundamental vibration frequency.

School of Engineering
Natural Frequencies and Mode Shapes.

Example. We consider an elementary example of a matrix eigenproblem and its solution. Consider the
uniform one-dimensional unsupported bar with mass density  , elastic modulus, E, and cross-sectional area
A as illustrated below. With a consistent mass matrix, then K  M D  0, wher e    2 becomes:

 AE 1  1 2  AL 2 1   u1  0
            
 L 6 1
  1 1 2  u 2  0

For non-trivial amplitudes d  u1 u 2 T to exist, the determinant of the expression in parenthesis must
vanish. Thus,
 
 2  2  AL  12 AE / L  0
from which 1 = 0 and .
The easiest way to determine the eigenvector associated with any eigenvalue i is to set one d.o.f. in
the vector d to an arbitrary non-zero number ( , for example), substitute the known value of i and
d
solve for the remaining amplitudes in . (If, by coincidence, the d.o.f amplitude that was assumed to be
non-zero is in fact zero, then the resulting system of equations will be singular and no solution will exist for
the remaining amplitudes. Then the eigenvector can usually be found by assuming a non-zero value for one
of the other amplitudes.)

Fig. 13.5-1
(a) Unsupported two-d.o.f. uniform bar.
(b) Vibration modes 1 = 0 (rigid-body
translation) and 2 > 0 (axial straining
mode).

School of Engineering
Natural Frequencies and Mode Shapes.

Example. Thus, from these sets of equations, the eigenvalues for this problem are obtained as:

The first eigenvector describes rigid-body translation in the x-direction. The second describes an axial
straining mode (which the exact fundamental natural frequency of a continuous unsupported bar of length L
 / L  E / 
is
m   AL / 21 1
. Hence, the single-element consistent-mass model over predicts the exact fundamental
frequency by about 10%. If instead of the consistent mass  2  2 the
matrix, / L  lumped
E /  mass matrix
is used instead, the computed frequencies are 1 = 0 and .

(a) Unsupported two-d.o.f.


uniform bar. (b) Vibration
modes 1 = 0 (rigid-body
translation) and 2 > 0 (axial Fig. 13.5-1
straining mode).

School of Engineering
Natural Frequencies and Mode Shapes.

Further remarks. In many design situations, we wish to know if severe dynamic excitation of a structure is
likely. Therefore, we compare the frequency spectrum of the structure with that of the time-dependent
loading. If a natural frequency of the structure is close to the excitation frequency, then severe vibration and
resonance are likely. This usually necessitates alteration of the structure’s natural frequencies by resizing or
by adding members or dampers. If frequencies of the structure and the excitation are well separated, the
structure still vibrates, but the amplitudes of the response is likely to be tolerable.
In static analysis, symmetry may be exploited, for example, by analysing half the entire structure. In
vibration analysis, symmetry of the structure and supports does not imply symmetry of all vibration modes.
By imposing symmetry, all anti-symmetric modes of vibration would be excluded, which are probably as
important as the symmetric modes.
It should also be noted that stabilisation methods used to suppress mechanisms in under-integrated
elements may be associated with relatively low stiffness. The associated non-physical vibration modes may
contaminate part of the vibration spectrum of greatest interest.

School of Engineering
Time-History Analysis. Direct Integration Methods.

In direct integration methods or step-by-step methods, a finite difference approximation is used to replace
D and D
the time derivatives appearing in the equation below (i.e.  D
) by differences of displacement at
various instances of time.
MD  C D  K D  R ext

Finite difference methods for approximately solving initial value problems have been well studied.
For many structural dynamics and wave propagation problems, including those with complex non-linearities,
direct integration is more expedient.
Many methods of direct integration are popular and the choice of method is strongly problem-
dependent. In this section explicit and implicit methods are introduced.

School of Engineering
Time-History Analysis. Direct Integration Methods.

In direct integration, the approach is to write the equation of motion at a specific instant of time,

  C D  K D  R ext
MD 
n n n n

where subscript n denotes time n t and t is the size of the time increment or time step. The absence of
time step subscripts on the mass, damping, and stiffness matrices implies linearity. For problems with
material non-linearity, the stiffness matrix is a function of displacements and therefore of time as well.
Difference methods for direct integration of equation  above can be categorised as explicit or implicit.
Explicit methods have the form:
D n 1 
 f D , D , D
n
 , D ,
n n n 1 
and hence permit Dn 1 to be determined in terms of completely historical information consisting of
displacements and time derivatives of displacements at time n t and before.
Implicit methods have the form:

Dn 1  f D n 1, D

n 1, Dn , 
and hence computation of Dn 1 requires knowledge of the time derivatives of Dn 1 , which are
unknown.
Explicit and implicit method have markedly different properties. This has important practical
implications.

School of Engineering
Explicit Direct Integration Methods.

A popular method, which is characteristic of explicit methods in general, is the central-difference method. It
approximates velocity and acceleration by:

D n 
1
Dn 1  Dn 1 
2 t
  1 D
D n n 1  2Dn  Dn 1  
t 2
These equations are obtained by expanding Dn 1 and Dn 1 in Taylor series about time n t:

t 2  t 3 
Dn 1  Dn  t D  Dn  Dn   
2 6

 t 2  t 3 
Dn 1  Dn  t D  Dn  Dn   
2 6

Subtracting  from  and adding  and  yields  and . Combining  and  with  provides:

 1 
  2 M 
1
C  Dn 1  R ext
n  K Dn 
1
M  2Dn  Dn 1  
1
C Dn 1
 t 2 t  t 2 2 t

School of Engineering
Explicit Direct Integration Methods.
Remarks.
1. Equation  is a system of a linear algebraic equations. If the mass and damping matrices are
diagonal, then the equations are uncoupled
Dn 1 and can be obtained without solving simultaneous
equations.
2. R int nthe
For small finite element models,  Kstiffness
Dn matrix can be formed and stored in the computer’s core
memory and the internal force at each time step can be obtained by matrix multiplication.
However, it is more common, even for linear problems, to compute the internal force vector at each
time step by summation of element contributions. Because the element stiffness matrix need not be
formed or stored, explicit methods can treat large three-dimensional models with comparatively modest
computer storage requirements. D D
1 0
3. D 0 the method from n = 0 requires
Starting which can be computed from known initial conditions
 t 2 
and and : D1  D0  t D0  D0
2

D 0
where terms with t and higher powers

are omitted.

is obtained from the equation of motion at
3
1 ext
time zero: D0  M R 0  K D0  C D0
t  2 /  max
4.

Equation  is conditionally stable and requires t such that:
det K   2M  0 
where max is the highest natural frequency of .

If equation  is not satisfied, computations will be unstable. This is indicated by an obviously


erroneous time-history solution that grows unbounded, perhaps by orders of magnitude per time step.
(In non-linear problems, instabilities
 1 may be more difficult to detect.)
 2 M  2t C  Dn 1  R n  K Dn  2 M  2Dn  Dn 1   2t C Dn 1
1 ext 1 1
School of Engineering 
 t  t
Implicit Direct Integration Methods.
Most of the useful implicit methods are unconditionally stable and have no restriction on the time step size
other than as required for accuracy. A popular unconditionally stable method is the trapezoidal rule or
average acceleration method. The trapezoidal rule (or average acceleration method) relates
displacements, velocities, and accelerations (from a Taylor series) by: t 
Dn 1  Dn 
2

Dn  D n 1  
D n 1  D n 
2

t  
Dn  D n 1  

These equations can be solved for D n 1 and D



n 1 to provide:

D n 1 
2
Dn 1  Dn   D n
t

D n 1 
4
Dn 1  Dn   D n  D
4 
n
t 2 t
Combinations of these two equations with the equation of motion at time ( n + 1 ) t , yields:

K eff Dn 1  R eff n 1 
where the effective stiffness matrix and effective load vector are, respectively:
4 2
K eff  M  C K
t 2 t
 4 4    C  2 D  D 
R eff n 1  R ext n 1  M  Dn  D n  D   n 
 t 2 t n   t n

School of Engineering
Implicit Direct Integration Methods.

Remarks.

1. Equation  is a system of coupled linear algebraic equations even if the mass and damping matrices
are diagonal. For linear problems, the effective stiffness matrix needs be formed and factored once.
After the initial expense of factorisation, time stepping can be performed for only the cost of forward-
and back-ward substitution.
2. If the mass matrix is positive definite, then the effective stiffness matrix is non-singular even if the
structure stiffness matrix permits rigid-body displacements.
3.
  
1 ext
This method is easily started from initial conditions D0 and D0 and D0  M R 0  K D0  C D0
 
.
4. Dn 1
For problems having material non-linearity, the structure stiffness matrix and hence the effective
stiffness matrix are functions of Dn 1
, and possibly its time derivative, which are unknowns.
Dn 1 matrix must be predicted using an estimate for
Accordingly, the structure stiffness . Equation 
is then solved for improved ; hence, the prediction of the structure stiffness matrix is improved,
and so on. For severe non-linearity, convergence may be difficult and expensive.

School of Engineering
Implicit Direct Integration Methods.

Remarks.

Choice of time step t. Since the method is numerically stable for any t , time step selection is based on
accuracy considerations alone. Compared with explicit methods, the per-time-step cost of an implicit
method is high. Thus, implicit methods are economically attractive only when t can be much larger the
would be used in an explicit method. Unconditional stability (which emphatically does not imply
unconditional accuracy) coupled with the economic need for large t tempts many analysts into
using time steps that are too large.
To select a time step that will provide accurate results, the highest natural frequency of interest in the
loading or response of a structure must be identified. Let this frequency be called u. As an approximation,
structure modes with frequencies higher than about 3 u participate quasi-statically in the response, while
modes with frequency lower than 3u participate dynamically. With second-order accurate time integration
methods (most popular methods are second-order accurate), a minimum of 20 time steps per period of 3 u
should provide very good accuracy for modes that participate dynamically in the response; that is, use

t  2 / u  / 20  0.3 / u

unless a smaller t is required because of convergence difficulties in non-linear analysis.

School of Engineering
Other Implicit and Explicit Methods. Mixed Methods.

The central-difference and trapezoidal-rule methods do not provide automatic dissipation of high-frequency
numerical noise, as is sometimes desirable. Outlines of methods that often have dissipation (also called
artificial viscosity or numerical damping) are given in what follows.

The Houbolt method is obtained by cubic Lagrangian interpolation of D at times ( n – 2 ) t through to


( n + 2 ) t , with
D n 1 
1
11 Dn 1  18 Dn  9 Dn 1  2 Dn  2 
6t

D n 1 
1
2
2 Dn 1  5 Dn  4 Dn 1  Dn  2 
t

This method is implicit and unconditionally stable, but provides artificial damping that is too high for low-
frequency response. The Houbolt method was once common in general-purpose transient codes but has
been supplanted by methods with better algorithmic damping.

School of Engineering
Other Implicit and Explicit Methods. Mixed Methods.

The Newmark family of methods is very popular and is given by:

where  and  are chosen by the analyst to control stability and accuracy. Substitution of the above
equations into  at time ( n + 1 ) t yields equations similar to  for explicit Newmark methods (  = 0 ) and
to .

School of Engineering
Other Implicit and Explicit Methods. Mixed Methods.

Example Newmark family of methods

Table 13.12-1

Summary of Newmark methods: u = undamped, d = damped. crit = max tmax, .


Stability requires t  crit / max.

School of Engineering
Concluding Remarks on Time-History Analysis.

Choice of Method. The choice of method for time-history analysis is strongly problem-dependent. The
efficiency of a given method depends on whether the problem is of a wave propagation or a structural
dynamics type, the time span for which analysis is required, whether response is linear or non-linear, and
the topology of the finite element mesh.
In wave propagation problems the excitation is usually dominated by high-frequency components. Time
scales of interest are short and of the order of acoustic wave traversal time across a structure. Usually, we
are interested in observing the passage of stress waves through elements and the transients produced.
In structural dynamics problems, the excitation and response are characterised by low-frequency, long-
term-scale components. Analysis duration is usually long compared to that normally required for a wave
propagation problem.
If material response becomes non-linear or deformations become large, a structure’s eigenvalues and
eigenvectors change. Because of the expense of solving eigenproblems, it is prudent to continuously update
eigenpairs during non-linear response.

School of Engineering
Concluding Remarks on Time-History Analysis.

With explicit methods of direct integration, stability typically requires that the time step be small enough that
information does not propagate across more than one element per time step.
Explicit methods are ideal for wave propagation problems in which behaviour at the stress wave front is of
engineering importance. Here the stability restriction is not a serious disadvantage because a small time
step is necessary for accuracy. Other factors in favour of explicit time integration are ease of implementation,
accurate treatment of general non-linearities, and the capability of treating very large problems with only
modest computer storage requirements. For structural dynamics problems, time scales and analysis
durations are usually long and accuracy considerations alone would permit a much smaller time step than
the upper limit of time step for stable explicit integration. Although explicit methods are often used for
structural dynamics problems, they are not as well suited to this class of problems as they are to wave
propagation problems.
The only advantage of implicit methods over explicit methods is they allow a much larger time step
because they are unconditionally stable. Implicit methods are expensive for wave propagation problems
since accuracy requires a small time step. For long-duration structural dynamics problems, implicit methods
are usually more effective than explicit methods, although this depends on mesh topology and severity of
non-linearities. Compared with explicit methods, implicit methods are more difficult to implement, particularly
for nonlinear problems, and they require considerably more computer storage.

School of Engineering
Concluding Remarks on Time-History Analysis.

Choice of Element and Mesh. When discretising a structure or a medium, an analyst can choose from
simple, low-order elements such as the linear-displacement bar, quadratic beam, and bilinear quadrilateral,
or from higher-order elements such as quadratic-Lagrange and serendipity quadrilaterals.
In wave propagation problems, discontinuities of strain propagate throughout the model. Lower-order
displacement elements are more adept at modelling these discontinuities than are higher-order elements,
which tend to produce more numerical noise.
Structural dynamics problems tend to have strain fields that vary smoothly with time. Hence, higher-order
elements can be used to more advantage than in wave propagation problems. Higher-order elements can
also be used effectively in eigenvalue problems.
Element sizes should not change abruptly. If they do, the mass matrix will be a poor discrete representation
of the actual continuous mass distribution of the structure. This gives rise to artificial wave reflections and
additional numerical noise when wave cross boundaries between elements of markedly different size.
With explicit methods, a lumped mass matrix is preferred for reasons of economy and accuracy. With implicit
methods, a consistent mass matrix is preferred for accuracy and is only slightly detrimental to economy.

School of Engineering
School of Engineering
School of Engineering
School of Engineering
School of Engineering
School of Engineering
CEG8315
Advanced structural modelling

4. Finite elements in dynamics and vibrations

4b. Time history analysis

School of Engineering

School of Engineering

You might also like