3.9. Modal Analysis Tools for Subsequent Mode-Superposition Analysis

3.9.1. Using the Residual Vector or the Residual Response Method to Improve Accuracy

A mode-superposition solution tends to be less accurate when the applied dynamic loads excite the higher resonant frequency modes of a structure. Many modes are often necessary to render an accurate mode-superposition solution.

The residual vector or residual response method can help in such cases. The method's improved convergence properties require fewer natural frequencies and modes from the eigensolution.

Both methods are similar except that just like the mode shapes, the residual vectors are orthonormalized while the residual responses are physical static responses.

The residual vector method improves the accuracy of a mode-superposition transient, mode-superposition harmonic, PSD, or spectrum analysis.

The residual response method improves the accuracy of a mode-superposition transient or mode-superposition harmonic analysis. It is particularly useful when the residual vector method does not apply, for example when the residual vector has zero frequency. Another example is when the equations are not symmetric and the unsymmetric eigensolver (MODOPT,UNSYM) is used.

3.9.1.1. Using the Residual Vector Method to Improve Accuracy

To use the residual vector method, you must first calculate residual vectors in the modal analysis. You can use any of these modal analysis mode-extraction methods:

  • Block Lanczos (MODOPT,LANB)

  • PCG Lanczos (MODOPT,LANPCG)

  • SNODE method (MODOPT,SNODE)

  • SUBSP method (MODOPT,SUBSP)

  • QR damped method with no complex solutions (MODOPT,QRDAMP,,,,OFF)

Mechanical APDL stores the calculated residual vectors in the Jobname.mode file and uses them in the subsequent mode-based analysis (for example, mode-superposition harmonic, mode-superposition transient, spectrum, or random vibration (PSD) analysis).


Note:  When using the residual vector method, issue MODOPT and set FREQB to zero or blank. Residual vectors cannot characterize the high-frequency range if significant modes in the low frequency range are not extracted.

Because the residual vectors are calculated before the modes are expanded, they are based on all the extracted modes. As a result, a partial expansion driven via MXPAND does not affect the residual vectors. If you do not expand the residual vectors, they will not be available in the results file for postprocessing.


The following process determines the residual vectors:

  1. Build the model.

  2. Specify the mode-extraction method (MODOPT,LANB or MODOPT,LANPCG, MODOPT,SNODE, MODOPT,SUBSP, or MODOPT,QRDAMP,,,,OFF).

  3. Activate residual vector calculation (RESVEC,ON).

  4. Specify pseudo-constraints (D,,,SUPPORT) if rigid body motion is present.

  5. Specify the load vectors (F, BF, SF, etc.).

  6. Solve the modal analysis. (Mechanical APDL generates a Jobname.mode file containing the residual vectors.)

  7. Issue a FINISH command.

    The residual vector calculation can be performed during a modal analysis restart. For more information, see Modal Analysis Restart in the Basic Analysis Guide.

  8. Set up a mode-superposition transient, mode-superposition harmonic, PSD, or spectrum analysis, and include the previously calculated residual vectors (RESVEC,ON).

    A load vector is also generated in Step 6. Ensure that you do not duplicate any loading.

  9. Solve the mode-superposition analysis. Mechanical APDL includes the residual vectors in those calculations.

Specifying Pseudo-Constraints

If rigid body motion exists, specify only the minimum number of displacement constraints necessary to prevent rigid body motion: three constraints (or fewer, depending on the element type) for 2D models and six (or fewer) for 3D models.

3.9.1.2. Using the Residual Response Method

To use the residual response method, you must first calculate residual responses in the modal analysis. You can use any of these modal analysis mode-extraction methods:

  • Block Lanczos (MODOPT,LANB)

  • PCG Lanczos (MODOPT,LANPCG)

  • SNODE method (MODOPT,SNODE)

  • SUBSP method (MODOPT,SUBSP)

  • Unsymmetric method (MODOPT,UNSYM) with real solutions (CpxMod = REAL) and both left and right eigenvectors requested (ModType = BOTH)

  • QR damped method with no complex solutions (MODOPT,QRDAMP,,,,OFF)

The program stores the calculated residual responses in the Jobname.rst file as load step 2 for postprocessing. They are also written to the Jobname.resf file and used in the expansion pass of the mode-superposition harmonic or mode-superposition transient analysis.

The notes from the residual vector method also apply to the residual response method.

Use the following procedure to calculate residual responses:

  1. Build the model.

  2. Specify the mode-extraction method - see above for supported methods. Make sure all significant modes are extracted.

  3. Activate residual vector calculation (RESVEC,,,,,ON).

  4. Specify the load vectors (F, BF, SF, etc.).

  5. Solve the modal analysis. (Mechanical APDL generates a Jobname.rst file containing the residual responses.)

  6. Issue FINISH.

    The residual response calculation can be performed during a modal analysis restart. Refer to Modal Analysis Restart in the Basic Analysis Guide for more details.

  7. Set up a mode-superposition transient or mode-superposition harmonic analysis and solve it.

    A load vector is also generated in Step 5. Ensure that you do not duplicate any loading.

  8. Perform the expansion pass of the reduced solution including residual responses (RESVEC,,,,,ON)

The residual response procedure does not support:

When using the residual response, all significant modes must be extracted. Unlike the residual vector, the residual response reduces to a static contribution.

3.9.2. Reusing Eigenmodes

Mechanical APDL analyses that require the eigenmodes from the modal analysis can reuse the modes from an earlier modal analysis solution. You can reuse the Jobname.mode (or Jobname.modesym) file that is created in a modal analysis in the following modal-based methods:

This section describes the process for saving and reusing the eigenmodes from an earlier modal analysis.

3.9.2.1. Spectrum Analysis (ANTYPE,SPECTRUM)

To run a spectrum analysis, first perform a modal analysis to generate the MODE file. For multiple spectrum analyses, a unique MODE file can be used when the mode-reuse key (modeReuseKey on SPOPT) is enabled. The mode-reuse key prepares the database and the necessary files for a new spectrum analysis that reuses an existing Jobname.mode.

3.9.2.2. Modal Transient Analysis/Harmonic Analysis

To use new load vectors, residual vector, and/or enforced motion in modal transient or modal harmonic analyses with existing modal analysis results, refer to Restarting a Modal Analysis

3.9.2.3. QR Damp Complex Modes Extraction

In QRDAMP eigensolver the solution occurs in two steps. First, an eigensolver is used to extract the symmetric matrix eigenmodes. These eigenmodes are then used in the second pass to build the modal subspace matrix of the non-symmetric eigensystem and calculate the complex eigenmodes.

When an existing Jobname.modesym containing the eigenmodes of the symmetric eigensolution of the model is available, it can be reused in the second pass by enabling the reuse flag (ReuseKey on QRDOPT).

3.9.3. Generating and Using Multiple Loads in Mode-Superposition Analyses

In a mode-superposition harmonic, transient, or PSD analysis, only nodal forces can be defined directly during the analysis. The other loads (SF, SFE, BF, ACEL, OMEGA, etc.) must be defined in the modal analysis. They can then be retrieved, scaled, and applied via LVSCALE in the mode-superposition analysis. Spring preloads, initial strains, pretension loads, and thermal loads are ignored and do not generate a load.

Nodal forces can also be applied more efficiently using this multiple loads generation method.

3.9.3.1. Understanding the Multiple Loads Method

To use the multiple loads method, you must first generate load vectors in the modal analysis. You can use any of these modal analysis mode-extraction methods:

  • Block Lanczos (MODOPT,LANB)

  • PCG Lanczos (MODOPT,LANPCG)

  • Supernode method (MODOPT,SNODE)

  • Subspace method (MODOPT,SUBSP)

  • Unsymmetric method (MODOPT,UNSYM)

  • QR damped method with no complex solutions (MODOPT,QRDAMP,,,,OFF)

Mechanical APDL stores the generated load vectors in the Jobname.mode file so that they can be used in the subsequent mode-superposition analysis.

3.9.3.2. Using the Multiple Loads Method

The procedure for generating multiple load vectors is as follows:

  1. Build the model.

  2. Specify the mode-extraction method (MODOPT,LANB; LANPCG; SNODE; SUBSP; UNSYM; or QRDAMP).

  3. Activate multiple load vectors generation (MODCONT,ON). Issue THEXPAND to ignore thermal strains in the generated load vector.

  4. Specify the first set of loads (F, BF, SF, etc.).

  5. Issue SOLVE. This first solution performs the modal extraction and generates the first load vector.

  6. Delete the first set of loads and apply the second set of loads.

  7. Issue SOLVE. This solution generates the second load vector.

  8. Repeat step 6 and 7 for any additional load vectors.

  9. Issue FINISH.

The load vectors calculation can be performed during a modal analysis restart. For more information, see Modal Analysis Restart in the Basic Analysis Guide.

The load vectors can then be used in a mode-superposition analysis as follows:

  1. Set up a mode-superposition transient, harmonic, or PSD analysis. Ensure that any loads used in the creation of the load vector in the modal analysis are deleted; otherwise, duplicate loads may result.

  2. Scale the load vectors (LVSCALE). All load vectors must be scaled before issuing the first SOLVE. Use a zero scale factor if load vectors are not actually used in this first load step.

  3. Solve the mode-superposition analysis. Ensure that load vectors unused at a given load step have a zero scale factor (LVSCALE).

3.9.4. Restarting a Modal Analysis

Reusing eigenmodes that have already been generated can save significant time in an analysis. Modal extraction typically requires more time than element loads generation, residual vector calculation, and enforced static modes calculation. The procedure for restarting the modal analysis to enrich the Jobname.mode file is described in Modal Analysis Restart.

3.9.5. Enforced Motion Method for Mode-Superposition Transient and Harmonic Analyses

In a mode-superposition transient or harmonic analysis, the enforced motion method can be used when acceleration or displacement base excitation is present.

3.9.5.1. Understanding the Enforced Motion Method

In a modal analysis, Mechanical APDL calculates pseudo-static modes and writes them to the Jobname.mode file. They can then be used in a subsequent mode-superposition analysis with base excitation.

3.9.5.2. Using the Enforced Motion Method

The procedure for calculating pseudo-static modes in a modal analysis is as follows:

  1. Specify modal analysis (ANTYPE, MODAL).

  2. Select one of the mode-extraction methods:

  3. Enable enforced motion calculation via MODCONT with EnforcedKey = ON.

  4. Specify the support points where imposed motion should be applied (node and degree of freedom label), as well as the enforced base-identification number (D). The ID number is input as VALUE, which must be an integer.

  5. Issue SOLVE.

  6. Issue FINISH.


Note:   The enforced motion pseudo-static modes calculation is only supported during the first load step of a base modal analysis or during the first load step of a restarted modal analysis. Refer to Modal Analysis Restart in the Basic Analysis Guide for more details about modal analysis restart.


To use the pseudo-static modes in the mode-superposition transient/harmonic analysis, issue DVAL.

3.9.5.3. Example: Enforced Motion Mode-Superposition Analysis

!  Build the Model
/FILNAM,...	     ! Jobname
/TITLE,…              ! Title
/PREP7                    
---
---			! Generate model
---
NSEL,...
CM, BASE1,NODE	! Define component BASE1 with selected support nodes undergoing the first enforced motion
NSEL,ALL
NSEL,...
CM, BASE2,NODE	! Define component BASE2 with selected support nodes undergoing the second enforced motion
FINISH


!  Generate the pseudo-static modes in Modal Analysis
/SOLU                      
ANTYPE, MODAL               
MODOPT, LANB,10             
MODCONT,, ON	! Activate enforced motion
              
---		! Define constraints
	
CMSEL,S,BASE1,NODE	! Select nodes of component BASE1
D, ALL, UX, 1		! Define enforced motion 1 to be applied on all nodes along UX
CMSEL,S,BASE2,NODE	! Select nodes of component BASE2
D, ALL, UY, 2		! Define enforced motion 2 to be applied on all nodes along UY
SOLVE
FINISH

!  Mode-Superposition Transient Analysis
/SOLU
ANTYPE, TRANSIENT
TRNOPT, MSUP
OUTRES, ALL,ALL
KBC,1
DELTIM,...
DVAL, 1, ACC, 100.0        ! Define constant acceleration for enforced motion 1
DVAL, 2, ACC, %ACCTAB%	! Define tabular acceleration for enforced motion 2
SOLVE

TIME,...
SOLVE

FINISH

3.9.6. Using Mode Selection

When performing a mode-superposition-based analysis (spectrum, PSD, transient, or harmonic), modes can be selected for mode expansion. The selected modes can be included in a downstream mode-superposition-based analysis. Consider using mode selection in the following cases:

  • When spurious modes have been extracted that may lead to erroneous mode-superposition results.

  • When the model and/or the number of extracted modes is large.

Modes are selected before mode expansion, such that the Jobname.mode file and the Jobname.rst file (from the modal analysis) contain only the selected modes. By reducing the total number of modes included in the superposition analysis, mode selection saves disk space and increases the performance of the postprocessing of the modal results. This means that mode selection can be useful even when a downstream mode-superposition analysis is not intended.

Mode selection is generally performed during a modal analysis restart. See Modal Analysis Restart in the Basic Analysis Guide for more details.

The following methods are available for mode selection:

Zero-frequency modes are ignored when mode selection is based on a criterion (Modal Effective Mass, Mode Coefficients, and DDAM methods). Selecting zero frequency modes is possible via the User Defined Array procedure only.

3.9.6.1. Mode Selection Based on a User-Defined Array

The mode selection is performed during a modal analysis or a modal analysis restart. It is applicable to real and complex eigensolvers.

The user-defined array is a *DIM array with a dimension equal to the number of extracted modes. The array values are 1 for selected modes and 0 otherwise.

Damped eigensolver solutions (MODOPT,DAMP) are pairs of complex conjugate modes. When filling in the table array, select the pairs (and not individual modes) to ensure that the MODE file remains consistent. (For example, modes #3 and #4 are a pair, so indices 3 and 4 in the array should have the same value, 0 or 1.)

Example 3.1: Mode Selection Based on a User-Defined Array

! Build the Model
/filnam,…
/title,…
/prep7
…
finish

! Obtain the Modal Solution
/solu
antype, MODAL
modopt, UNSYM, 10	! Unsymmetric eigensolver, 10 modes are requested
solve
finish
save

/post1
…		! Postprocess the modes to identify the modes of interest
finish

/clear, nostart	! Start a new session
resume,, db

*dim, tab, ARRAY, 10	! Create the array 
tab(1) = 1
tab(2) = 0
tab(3) = 0
tab(4) = 0
tab(5) = 1
tab(6) = 1
tab(7) = 1
tab(8) = 0
tab(9) = 1
tab(10) = 1

! Perform the modal selection and expansion
/solu
antype, MODAL, RESTART
mxpand, %tab%,,, yes	! Mode selection is based on user input array, and element results are 
		! requested
solve
finish

! Postprocess the modal results or perform a mode-superposition analysis

For the selection of complex modes from the QR damped eigensolver (MODOPT,QRDAMP), you must input MXPAND,-1 in the modal solution. Doing so forces the generation of a complex MODE file that enables complex mode selection.

3.9.6.2. Mode Selection Based on Modal Effective Mass

The mode selection is performed during a modal analysis restart and applies to real eigensolvers only.

You can select the modes per direction using either of the following methods:

  • Use a significance threshold (SIGNIF on the MXPAND command). The significance level of a mode is equal to the modal effective mass divided by the total mass. The threshold is the same for all directions, but you can define the specific global x, y, z directions (MODSELOPTION).

  • Set a minimum ratio for the total modal effective mass to the total mass (MODSELOPTION). The ratio can be input for each direction.

Example 3.2: Mode Selection Based on Modal Effective Mass

! Build the Model
/filnam,…
/title,…
/prep7
…
finish

! Obtain the Modal Solution
/solu
antype, MODAL
modopt, LANB, …             ! Block Lanczos
solve
finish

! Perform the modal selection and expansion
/solu
antype, MODAL, RESTART
mxpand,,,, yes,,, EFFM	! Mode selection is based on the modal effective mass, and element
						! results are requested
modseloption, .90, no,no, no,no,no     ! Only direction X is selected, a minimum of 90% of the total 
			! mass is requested in this direction
solve
finish

! Postprocess the modal results or perform a mode-superposition analysis

If the minimum ratio cannot be achieved, you may need to extract more modes and run the procedure again.

If you want to calculate residual vectors, do so during the modal restart analysis. Note that the residual vectors calculation is always based on all the extracted modes and not the selected modes.

If groups of repeated frequencies (Repeated Eigenvalues in the Theory Reference) are present, note the following. The sum of the effective masses of all the eigenmodes in a group is unique. However, the effective mass, and therefore, the modal effective mass ratio of each eigenmode in the group is not. It can lead to different mode selection results depending on the machine and SMP/DMP settings. When minimum ratios are defined (MODSELOPTION), set them in all the directions of the eigenmodes of a group with repeated frequencies so that all the eigenmodes of the group are selected.

3.9.6.3. Mode Selection Based on the Mode Coefficients

The mode selection is performed during a modal analysis restart and applies to real eigensolvers only. This mode-selection method is used when a Single Point Response Spectrum (SPRS) analysis is performed following the modal analysis.

You can select the modes using a significance threshold (SIGNIF on the MXPAND command). The significance level of a mode is equal to the mode coefficient divided by the maximum mode coefficients (of all modes per direction). The threshold is the same for all directions.

Example 3.3: Mode Selection Based on the Mode Coefficients

! Build the Model
/filnam, …
/title, …
/prep7
…
finish

! Obtain the Modal Solution
/solu
antype, MODAL
modopt, LANB, …	! Block Lanczos
solve
finish

! Obtain the SPRS Mode Coefficients
/solu
antype, SPECTRUM
spot, SPRS
svtyp, 2, 386.4	        ! Input spectrum type 
sed, 1, 0, 0		! Input spectrum direction
freq, 1, 2, 5, 15, 30	! Input spectrum frequencies
sv,,   0.1, 0.2, 0.2, 0.05, 0.05	! Input spectrum values
solve
finish

! Perform the modal selection and expansion
/solu
antype, MODAL, RESTART
mxpand,,,, yes, 1e-2,, MODC	! Mode selection is based on the mode coefficients, the significance
								! threshold is 0.01 and element results are requested
solve
finish

! Perform a spectrum analysis (combination step)
/solu
antype, SPECTRUM
spopt, SPRS		! Single Point Response Spectrum analysis
srss                                    ! Square Root of Sum of Square combination
solve
finish 

For multiple input spectra in a SPRS analysis, to obtain the mode coefficients in the spectrum analysis, the spectrum input definitions are requested, each one followed by SOLVE.

This method is not supported for Multiple Point Response Spectrum (MPRS) and Power Spectrum Density (PSD) analyses. However, you can use a dummy SPRS input spectrum to perform the mode selection for the subsequent MPRS and PSD analysis, or else use the mode selection based on the modal effective mass.

3.9.6.4. Mode Selection Based on the DDAM Method

The mode selection is performed during a modal analysis restart. It is applicable to real eigensolvers only. The intent of this mode selection method is to perform a Dynamic Design Analysis Method (DDAM) Spectrum and Single Point Response Spectrum (SPRS) analyses following the modal analysis. It is based on the procedure described in NAVSEA [408].

In this procedure, the modes are selected using a significance threshold equal to 0.01. The significance level of a mode is equal to the modal effective weight or mass divided by the total weight or mass. The threshold is the same for all directions.

The sum of the modal effective weights or masses of all calculated modes must be greater than 80% of the total weight or mass, otherwise you may need to extract more modes and run the procedure again.

After the significance calculation, the procedure can select additional modes based on the nodal acceleration check. This check accounts for the localized response of relatively light-weight sub-components of critical areas. To apply the acceleration check, you need to select all the nodes of interest and create a nodal component named “_DDAM_NODCHK” (see Creating Components)

The additional modes included correspond to those for which the nodal acceleration of any of the nodes of interest exceeds by 10 percent the maximum nodal acceleration calculated from the selected modes.

If you want to calculate residual vectors, do so during the modal restart analysis. Note that the residual vectors calculation is always based on all the extracted modes and not the selected modes.

This method is not supported for Multiple Point Response Spectrum (MPRS), Power Spectrum Density (PSD), Mode Superposition Harmonic, or Mode Superposition Transient analyses. However, you can use a dummy SPRS input spectrum to perform the mode selection for the subsequent MPRS and PSD analysis, or use the mode selection based on the modal effective mass.

Example 3.4: Mode Selection Based on the DDAM Method

! Build the Model
/filnam, …
/title, …
/prep7
…
esel,s,,,78,89
esel,a,,,96
esel,a,,,112
nsle				! Select nodes of interest
cm,_DDAM_NODCHK,node			! Create a component  for the
					! nodal acceleration check
allsel,all,all
…

finish

! Obtain the modal solution
/solu
antype, MODAL
modopt, LANB, …			! Block Lanczos

solve
finish

! Obtain the DDAM Mode Coefficients
/solu
antype, SPECTRUM
spopt, DDAM				! DDAM Spectrum analysis
addam,1.0,10.0,37.5,12.0,6.0	! Input spectral acceleration   
vddam,1.0,30.0,12.0,6.0		! Input spectral velocity           
sed,0,1,0				! Input spectrum direction
solve
finish

! Perform the modal selection and expansion
/solu
antype, MODAL, RESTART
mxpand,,,, yes,,, DDAM	! Mode selection is based on DDAM
				! procedure, the default significance 
				! threshold is 0.01 and element results are ! requested
solve
finish

! Perform a DDAM spectrum analysis (Combination Step)
/solu
antype, SPECTRUM
spopt, DDAM			! DDAM Spectrum analysis
nrls,…                     ! NRLSUM mode combination
dmprat,…			! Define damping ratio
solve
finish