$-----------------------------------------------------------------------
$ Rigid Format 103 - Normal modes analysis
$ MSC/NASTRAN Version 2001 or NX Nastran Version 1.0
$
$
$ ******************************************************
$ ***** COPYRIGHT (C) 2003 *****
$ ***** BY ATA ENGINEERING INC. *****
$ ***** ALL RIGHTS RESERVED *****
$ ******************************************************
$
$ 09-03- ATA/Paul Blelloch
$
$ Description:
$
$ This alter performs an mass weighted effective independence scheme to\
$ optimally select A-set DOF. It starts from the A-set defined by the user and
$ iteratively removes DOF until it reaches a user specified number. At
$ each iteration it eliminates the DOF with the highest ratio of diagonal
$ stiffness to mass. This will tend to do the best job at predicting
$ low frequency modes.
$
$ There are two ways to use this alter. The first is to not define any
$ ASET DOF in the model. In this case the alter starts from the F-set.
$ The alter always forces Nastran to eliminate massless DOF from the
$ problem before moving to the eigensolution. If the F-set is large
$ (> 5,000 DOF), run times and scratch space can be extensive. The
$ alternative is to define a large ASET. In this case the run time is
$ a function of the size of the starting ASET. This is the recommended
$ approach, with a starting ASET size < 5,000. A number of methods can
$ be used to generate this initial ASET including grid point kinetic
$ energies or residual kinetic energies.
$
$ There are two variations on the alter. The default multiplies the mode
$ shapes by the full mass matrix at each step, while the alternative
$ uses the square root of the diagonal of the mass matrix. Preliminary
$ experience doesn't show any clear preference between these methods, but
$ they will give different results.
$
$ The number of DOF eliminated at each step of the iterative process is
$ controlled by the parameter EFFILT. The default value of 1.0 will
$ eliminate one DOF at each step. Setting EFFILT to a value less than
$ 1.0 may result in a larger number of DOF eliminated at each iterative
$ step. All DOF with a stiffness/mass ratio greater than EFFILT times
$ the largest ratio are eliminated at each step.
$
$ If the value of EFFILT results in deleting enough DOF that the
$ remaining number is less than NEFYAN at any step it is internally
$ reset to 1.0. From this point on one DOF will be eliminated at all
$ following steps.
$
$ The alter will optionally calculate two error norms at each step in the
$ iteration. The error norm is based on the following orthogonality
$ error matrix:
$
$ EORT = [I] - PHIX'*MXX*PHIX
$
$ The first norm (NORT) is the RSS of all the terms in this matrix. The
$ second norm (MORT) is the maximum absolute value of all terms in
$ this matrix. The user must supply the modes to be used for this
$ calculation, and can optionally supply a range of modes (N1 to N2)
$ to be used for the calculation.
$
$ Currently this alter will calculate the eigenvalues for the reduced
$ mass and stiffness matrices, but will then fail. It will write the
$ selected DOF list to the PCH file. This can be used to create ASET
$ cards to perform a Guyan reduction.
$
$ Special instructions to use this alter:
$-----------------------------------------------------------------------
$ FILE MANAGEMENT SECTION (FMS)
$
$ If no ASET is defined, or the ASET is large (>1000 DOF) a considerable
$ amount of scratch space may be required. This may require the addition
$ of a card of the following form
$
$ INIT SCRATCH LOGICAL=(SCR1(20GB)),
$ SCR300=(SCR300(20GB))
$
$-----------------------------------------------------------------------
$ EXECUTIVE CONTROL DECK
$
$ SOL 103
$ Include this alter immediately before the CEND statement
$
$-----------------------------------------------------------------------
$ CASE CONTROL DECK
$
$ No special input is required.
$
$-----------------------------------------------------------------------
$ BULK DATA DECK
$
$ Optional parameters:
$
$ PARAM,ENORM,CHAR 'YES' : Calculate error norm
$ 'NO' : Do not calculate error norm (default)
$
$ PARAM,WRTMAT,CHAR 'YES' : Write reduced matrices at end of run
$ 'NO' : Do not write reduced matrices at end of run (default)
$
$ PARAM,DMASS,CHAR 'YES' : Multiply modes by square root of diagonal of mass
$ 'NO' : Multiply modes by full mass matrix (default)
$
$ PARAM,NDOF,I : Number of DOF in final set (default = 1000)
$
$ PARAM,EFFILT,RS : Filter to remove DOF at each step (default = 1.0)
$
$ PARAM,OMODES,I <0 : Read FEM modes from OUTPUT2 file UNIT = |OMODES|
$ 0 : Read FEM modes from DMIG cards
$ >0 : Read FEM modes from OUTPUT4 file UNIT = OMODES
$
$ PARAM,N1,I : 1st FEM mode to use in error norm calculation (default = 1)
$
$ PARAM,N2,I : Last FEM mode to use in error norm calculation (default = all)
$
$-----------------------------------------------------------------------
$ EXAMPLE NASTRAN DECKS
$
$
$ ASSIGN MASTER='gpsc_aset.MASTER'
$ ASSIGN DBALL ='gpsc_aset.DBALL'
$ $
$ SOL 103
$ INCLUDE iter_guyan.v2001
$ CEND
$ $
$ TITLE = GENERAL PURPOSE SPACECRAFT
$ SUBTITLE = SELECT DOF
$ LABEL = MASS WEIGHTED EFFECTIVE INDEPENDANCE METHOD
$ $
$ SPC = 1
$ METHOD = 50 $ Modes to 50 Hz
$ $
$ BEGIN BULK
$ $
$ PARAM,NDOF,50 $ Select best 50 DOF
$ PARAM,EFFILT,0.9 $ Eliminate DOF with > 90% of max ratio
$ $
$ EIGR,50,AHOU,1.0,50.0
$ $
$ INLCUDE 'gpsc.blk'
$ $
$ ENDDATA
$
$-----------------------------------------------------------------------