Damping methods in OptiStruct ™

Introduction

In this article we want to clarify the different types of damping cards that can be used in frequency
response and transient response analysis. Many possibilities are available depending on the
solution type (frf/transient) and solving method (direct/modal).

Air

Intermolecular

Source: https://www.youtube.com/watch?v=-QJguYTWFsI

Contact
Figure 1: Damping mechanic

Damping can be produced by the interaction of the structure that vibrates with the surrounding
medium, like the air in the above image, through energy exchange, or can be the result of a
mechanical dissipation involving friction, with dampers and contacts.

The below images show some examples of real damping devices.

 Figure 2:Tuned mass damper for earthquakes

Figure 3: semi-active damping device for space launchers

Figure 4: friction damping in turbine blades

Damping formulation

From a mathematical point of view damping can be represented as:

proportional to velocity, usually called viscous damping, where the damping force is defined as

𝐹 = 𝑗𝜔𝐶𝑥 = 𝑗2𝜁(𝜔/𝜔0 )𝐾𝑥

where 𝜁 is the percent of critical damping.

Proportional to displacement, usually called structural damping, or hysteretic damping, where

the damping force is defined as

𝐹 = 𝑗𝐻𝑥 = 𝑗𝛾𝐾𝑥

where 𝛾 is the percent of stiffness.

Viscous and structural damping forces are not the same except at resonance where:

𝛾 = 2𝜁

Figure 5: Damping force

Some people describe structural damping as

𝐶
𝐺=2
𝐶0

with C0 being the critical damping. From the above image you can notice that structural damping
is constant but viscous damping is frequency dependent.

In the dynamic equation of motion, the damping (viscous) matrix is called B and is comprised of
several terms:

𝐵 = [𝐵1 ] + [𝐵2 ]
where:

➢ [𝐵1 ] is the contribution from damping elements (CBUSH, CVISC, CDAMPi) and B2GG.
 ➢ [𝐵2 ] is the contribution from B2PP direct input matrix and transfer functions.

In frequency response structural damping inserted with PARAM, G and GE on the MATi card do
not form a damping matrix but modify the stiffness matrix instead.

[𝐾] = (1 + 𝑖𝐺)[𝐾] + 𝑖 ∑ 𝐺𝐸 [𝐾𝐸 ]

where:

➢ [𝐾] is the global stiffness matrix.

➢ 𝐺 is the overall structural damping (PARAM, G).
➢ [𝐾𝐸 ] is the element stiffness matrix.
➢ 𝐺𝐸 is the element structural damping coefficient (GE on MATi entry).

Different cards to express damping are available based on the solution sequence

Figure 6: Damping types for frf/transient

Direct frequency response supports all damping types except SDAMPING. This modal damping
is applied to each mode separately with a tabular function TADBMP1 to preserve the
diagonalization of the dynamic matrix produced by the coordinates change from physical to
modal. This is the reason why SDAMPING is not supported by the direct approach.

SDAMPING can be of type “G”, “CRIT” or “Q”.

Type “G” is equivalent to PARAM G, type “CRIT” is equivalent to PARAM G if:

➢ the damping value gi in the TABDMP1 or TABDMP2 is set to half the value of param, G
➢ param, KDAMP is set to -1, that means the SDAMPING is inserted as structural damping
into the stiffness matrix.

SDAMPING of type “Q” is called quality factor, that is the ratio of the resonance’s center frequency
to its half-power bandwidth:
𝑓0 1
𝑄= =
∆𝑓 2𝜁
 Figure 7: Q factor definition

Half-power corresponds to a decay factor of 1/√2 and Q factor is inverse proportional to the
damping factor ζ, that means the higher its value the less damped the system. This method can
be used to estimate damping of real structures from the experimental response curves to be used
in fea computation.

Figure 8: Amplification factor effect

In the direct approach, SDAMPING can be substituted by HYBDAMP card.

Rayleigh damping

Rayleigh damping model is a form of viscous damping proportional to a linear combination of

stiffness and mass:

𝐵 = 𝛼𝑀 + 𝛽𝐾
where:

➢ α is the mass proportional damping coefficient;

➢ β is the stiffness proportional damping coefficient.
 Figure 9: Rayleigh damping definition

Relationship between the modal equations and orthogonality conditions allow this equation to
be simply rewritten as:
1 𝛼
𝜉𝑖 = ( + 𝛽𝜔𝑖 )
2 𝜔𝑖
where:

➢ 𝜉𝑖 is the critical damping ratio;

➢ 𝜔𝑖 is the circular frequency 𝜔𝑖 = 2𝜋𝑓𝑖 .

α and β can be defined using 2 known values of frequency and damping ratio:

𝜔1 𝜔2 (𝜔2 𝜉1 − 𝜔1 𝜉2 )
𝛼=2
𝜔22 − 𝜔12
𝜔2 𝜉2 − 𝜔1 𝜉1
𝛽=2
{ 𝜔22 − 𝜔12

In OptiStruct ™ α and β can be inserted with param, ALPHA1 and ALPHA2. With Rayleigh damping,
no more than 2 damping ratios can be specified.

Note from the shape of the damping curve that mass damping term will dampen out rigid body
motions. In the case of rotating blades for example, which undergo rigid body motions, this
means that using mass proportional damping will artificially dampen out the motion of the
blades, leading to lower power output than expected. For these cases, it is recommended not to
use mass proportional damping.
Damping in Transient Analysis

Transient analysis supports damping in a similar way to frequency response analysis. The only
exception is PARAM G.

Figure 10: Undamped transient response

For PARAM G to be effective in Transient you need to add W3, W4. The reason is that transient
analysis does not allow complex stiffness, so structural damping has to be converted to viscous
damping and added to the other sources as follows:
𝐺 1
[𝐵] = [𝐵1 ] + [𝐵2 ] + [𝐾] + ∑ 𝐺𝐸 [𝐾𝐸 ]
𝑊3 𝑊4
Where:

➢ [𝐵1 ] is the contribution from damping elements (CBUSH, CVISC, CDAMPi) and B2GG.
➢ [𝐵2 ] is the contribution from B2PP direct input matrix and transfer functions.
➢ [𝐾] is the global stiffness matrix.
➢ 𝑊3 is the frequency in radians per unit time (PARAM, W3) for the conversion of overall
structural damping into viscous damping
➢ 𝐺 is the overall structural damping (PARAM, G).
➢ [𝐾𝐸 ] is the element stiffness matrix.
➢ 𝑊4 is the frequency in radians per unit time (PARAM, W4) for conversion of element
structural damping into viscous damping
➢ 𝐺𝐸 is the element structural damping coefficient (GE on MATi entry).

The explanation for the conversion expression can be found in the identity between viscous and
structural damping force at resonance of Figure 5,

𝐺
𝑏𝑖 = 𝑘
𝜔𝑖
Two parameters are used in the conversion of structural damping to viscous damping. PARAM,
W3 converts the overall structural damping that arises from PARAM, G specification. The default
value for the W3 parameter is 0.0, which causes the structural damping from PARAM, G
specification to be ignored in the transient response analysis. PARAM, W4 converts element
structural damping that arises from specifying values in the GE field of material. The default value
for the W4 parameter is 0.0, which causes the element structural damping to be ignored in the
transient response analysis. The units for PARAM, W3 and PARAM, W4 are radians per unit time.

To best select values for W3 and W4, you should consider the modal content of the structure and
the harmonic content of the excitation. Typically, the dominant frequency at which the damping
is active is used to specify a value for W3 and W4. Often, the first natural frequency is chosen,
but isolated individual element damping can occur at different frequencies, which can be
handled by using a different value for W4.

A thorough approach to selecting values for W3 and W4 might be to first perform a Fourier
transform of the excitation to obtain its harmonic content. Then using the dominant frequency as
the value for W3 and W4.

As an alternative to param, W3 and W4 can be specified from the TSTEP card

Modal Transient

Modal transient approach has similar support to damping wrt. direct approach. In addition,
SDAMPING card is supported. In modal approach the viscous damping matrix [B] is converted to
a modal viscous damping matrix.
𝐺 1
[𝐵𝑀 ] = [Φ𝑇 ] ([𝐵1 ] + [𝐵2 ] + [𝐾] + ∑ 𝐺𝐸 [𝐾𝐸 ]) [Φ]
𝑊3 𝑊4
where:

➢ [Φ] is the eigenvector matrix extracted from the modal analysis

With modal approach OS (2023.1 or later version) offers the possibility to use PARAM, WMODAL
to avoid the need to specify W3 and W4. In this scenario structural damping is converted to
viscous damping using the solved modal frequencies, and the viscous modal matrix is computed
as:
1 1
⋯ 0 ⋯ 0
√𝜔1 √𝜔1
[𝐵𝑀 ] = [Φ𝑇 ]([𝐵1 ] + [𝐵2 ])[Φ] + ⋮ ⋱ ⋮ [Φ𝑇 ] (𝐺[𝐾] + ∑ 𝐺𝐸 [𝐾𝐸 ]) [Φ] ⋮ ⋱ ⋮
1 1
0 ⋯ 0 ⋯
[ √𝜔𝑛 ] [ √𝜔𝑛 ]

where ωi are the modal frequencies in rad/sec.

Damping term for rigid body modes is set to 0, depending on the value of PARAM, FZERO for
Lanczos or PARAM, RBMEIG for Amses.

In Figure 12 a comparison between different damping parameters is shown in modal transient

analysis of the bracket model of Figure 11.
 Figure 11: first 3 modeshapes of the bracket

In Figure 12:

➢ blue curve is computed with PARAM, G, GE of MAT1 and PARAM, W3, W4 set to the first
natural frequency of 14hz;
➢ green curve is computed with PARAM, G, GE of MAT1 and PARAM, WMODAL;
➢ pink curve is computed with constant SDAMPING of type “G”, GE of MAT1 and PARAM, W4
set to the first natural frequency.

Figure 12: Modal transient analysis of bracket with different damping parameters

You can see that the results of the 3 methods are equivalent. This can be explained with the period
of oscillation being 0.074sec for this model, that corresponds to the first natural frequency of
14hz, that is the dominant mode for the dynamic of the system. Mode #2 and #3 are much more
localized and less effective.

You can also notice that different methods of damping can be combined together.
Conclusion

This article has tried to shed some light on the different damping possibilities available for
dynamic analyses.