Você está na página 1de 5

ME 6106 Project

Method to Select Master and Slave degrees of Freedom


Akash Padmane
150100034
April 2018

1 Abstract
Eigenvalue problems are widely used to solve various problems such as dynamic analysis of large, complex
structures in Finite Element Methods. For large problems with large degrees of freedom, the process of
solving an Eigenvalue problem becomes computationally expensive. The method of Guyan Condensation
identifies master degrees of freedom for a given Spring-Mass systems. The actual Mass and Stiffness matrices
are divided into master and slaves.
The problem with the method arises when we need to choose master and slave degrees of freedom. The
relations derived for Guyan Condensation relate the solution of an eigenvalue problem with master degrees
of freedom to that of the actual solution. Even though this helps to reduce the computational time and
resources to get a solution, the solution obtained is an approximation of the actual solution. The accuracy
of the obtained solution depends on the master and slaves degrees of freedom.
The following part focuses on implementing a method to select master and slave degrees of freedom. The
method gives masters and slaves degrees of freedom which can be used to reduce an existing eigenvalue
problem. Here the eigenvalues of the reduced matrix obtained is compared with the actual eigenvalues for
the original problem.

1
2 Introduction
In the field of Finite Element analysis, the problem of large, complex structures of spring and Mass are
common. These types of problems are of multiple DOFs and are solved by finding their eigenvalues. It is
computationally expensive to find all the eigenvalues of the components. One of the methods to gain an
approximate solution is by reducing the system to a new system that preserves the low frequency spectrum
of the initial problem. This method was implemented as a part of Guyan Condensation.
The method of Guyan condensation consists of selecting master and slave degrees of freedom. The next
step in Guyan Condensation method is to reduce the existing equations to get reduced equations. The second
step is demonstrated in the following theory section.
The major aim of the implemented method is to get most efficient and proper selection of master and slave
degrees of freedom. This selection should allow significant reduction of computation power as well as must
ensure all the significant lower frequencies are taken into account into the reduced system

3 Theory
A general nxn eigenvalue problem is of the following type.

[K]x = w2 [M]x
[K] and [M] are nxn stiffness and mass matrices of the original structure that need to be analysed. w is an
eigenvalue and x is a displacement vector.
We then divide x into xm and xs where the subscript denotes masters and slave displacement vectors.
According to the master and slaves, we also partition the mass and stiffness matrices.
        
Mmm Mms Kmm Kms Xm 0
−w2 + = 1)
Msm Mss Ksm Kss Xs 0

Here, Mxx and Kxx are sub matrices obtained after dividing the original Mass and Stiffness matrices
according to the master and slave degrees of freedoms.
On expanding the equation (1), we get the following equation.
−1 
{Xs } = − −w2 Mss + Kss −w2 Msm + Ksm {Xm }
 

Here, we neglect the terms containing Mss andKss .Hence, thenewrelationobtainedisasf ollows.

 −1 
Xs = − Kss Ksm Xm
The matrix -K−1
ss Ksm isrenamedasRg whichf ormsthef ollowingtransf ormationmatrix.
 
I
{X} = Xm
Rg
 
I
Tg =
Rg
This transforms the original equation to the following reduced equation.

Mg Ẍ + Kg X = 0

Mg = TgT M Tg

2
Kg = TgT KTg

TheMg and Kg matrices are obtained at the end of Guyan reduction which are then solved to obtain the
approximate solution to the original equation.

4 Methodology
This section focuses on the ways of selection of master and slave degrees of freedom. For the purposes of
calculation, a system of n degrees of freedom is constructed. This system consists of n Masses of random
values from 0 to 1 and n+1 springs of stiffness values in the range of 0 to 10000 and is a constrained system.
We also introduce a parameter fc and wc which is termed as the cutoff frequency. This frequency is signifi-
cantly higher than all the other frequencies of the original system. The steps to reduce the original system
are -

• We first find a degree of freedom for which ratio Kii /Mii is largest.
• If this ratio is greater than wc2 then that degree of freedom is eliminated.
• The above steps are repeated until all the degrees of freedom retained by the Mass and Stiffness matrices
have the ratio Kii /Mii less than wc2 . The above method is implemented by creating an array gamma. the
ith term of the array is given as -
Kii
gamma(i) = − wc2
Mii
For any ith DoF, the value of gamma(i) is positive, that corresponds to the slave degree of freedom and the
rest are taken as master degrees of freedom.

5 Results
For the problem of 100 DoFs, following two studies are done.

5.1 Variation of Master and Slave Degrees of Freedom with Fc


In this case, the parameter Fc is changed and the number of master and slave degrees are observed.

Value of fc Slaves DoFs Masters DoFs


10 97 3
20 64 36
30 31 69
40 16 84
50 12 88
60 9 91
70 5 95
80 2 98
90 0 100
100 0 100

3
5.2 Comparison of Eigenvalues
For the case of fc set as 20Hz, we will compare the eigenvalues obtained using Guyan Condensation and
ones obtained from original matrices.

The following are the graphs obtained on increasing the Initial Degrees of Freedom

• For 500 Degrees of Freedom

• DoFs = 1000

4
6 Discussion
Here, we can observe that the maximum of the New Frequencies is smaller than the maximum of the Original
Frequencies. Hence, all the small and significant degrees of freedom are taken into account by the method
that we’ve used.

7 Original Work
I have implemented the steps given in the reference paper in a different manner. Rather than moving
sequentially and eliminating the slaves one after another, i created the gamma array to remove all the
degrees which have values of Kii /Mii more than wc2 .
I also did the analysis of the data obtained. One way was to observe the change of number of masters
obtained with change in fc. Another was to compare the original eigenvalues obtained to the one obtained
from the reduction method.

8 Conclusion
After the Reduction of system using the method discussed above, we can conclude that all the significant
degrees of freedom will be taken into account(in the reduced system).
We also conclude that such a method would allow us to reduce the computation times of larges problems
significantly.

References
1. V. N. Shah and M. Raymund,’ Analytical selection of masters for the reduced eigenvalues problems

Você também pode gostar