Readme note for vma.v2001 alter package.
Revised January 22, 2002
M. A. Gockel

This alter does not interact well with changes made to the residual flexibility capability for 
V2001, and fixing the alter to do so would be a major project.  It is therefore being withdrawn.  
It is no longer on the web site, and its use with the SSSALTER library delivered with V2001 is 
not recommended.

Some of the content of the alter has been incorporated into V2001, but not all.  More updates 
are being discussed for V2002.  In the meantime, you can get most of the benefit of the alter 
by using CMS techniques without the need for DMAP alters.  The alter saves cost because the 
eigensolution modules use the mass matrix many times when orthogonalizing trial eigenvectors. 
 When the mass matrix is dense it greatly increases the cost and computer resource requirements 
of the eigensolvers.  VM tends to make very dense mass matrices.  The alter gets most of its 
efficiency by first finding the modes of the structure alone, then using these modes to make 
generalized coordinates that solve the problem with the VM added, in a much smaller basis.  
This basis tends to be fully coupled from structural effects.  There is no additional coupling
 due to VM, just the cost of reduction, which is relatively small.


CMS uses similar techniques.  All, or at least most, of the DOFs attached to the VM are put in 
the omitted set by user action.  The CMS modes are computed in the o-set without VM.  These 
shapes are used to reduce the o-set matrices to a-set size.  The VM effects are added before 
the reduction takes place.  The VM effects are present in the a-set (final) eigensolution.  If 
the o-set is much larger than the a-set cost savings and computer resource reductions comparable
 to what was achieved with the vma.2001 alter package can be achieved.

Test Problem

The old vma.v2001 demonstration input file vm01.dat has been modified in file vm01rfr to 
demonstrate how to convert a VM model from alter package usage to CMS usage.  The model has
 been converted to modal frequency response, and a shaker structure has been added to allow 
enforced acceleration input.  The enforced motion capability requires the presence of 
param,resvec, yes, the user interface to request residual flexibility effects.  A step-by-step
 process to convert the model is described here.

1. Remove the include statement that brings in the vma.v2001 alter package.
2. Place all the structural DOFs in the o-set.  This is done automatically by the input of step 3.
3. Add scalar points to serve as new variables, and qset1 entries to put these DOFs in the
 q-set, the generalized coordinate set for CMS.  The q-set is a subset of the a-set.  
Its presence causes all physical DOFs not otherwise constrained (MPC, SPC) to go to the 
o-set.
4. Add param, resvec, yes, and param,reviner, yes.  They provide increased accuracy for 
enforced motion.

You will observe in the results that there are one or several intermediate eigensolutions, 
followed by a last eigensolution.  The last eigensolution will have lower natural frequencies 
than some of the prior ones because it has VM effects, while some of the prior intermediate 
solutions do not.

Input file vm01x was made from the prior file.  It removes all CMS effects, and solves the 
problem with param, vmopt, 1, the most expensive way to solve the problem because it adds 
VM before the first eigensolution.  Both input files get acceleration results that agree 
within engineering accuracy.

Revision January 22, 2002

The slosh mode modeling technique described in the Advanced Dyanmics Guide and in the Sample 
Problem library does not produce good slosh modes when the alternate solution method involving 
an o-set.  Use the unmodified solution method, where all free DOFs are in the a-set.