Menaka Konara Hello. I think the mode superposition approach is a better one and worked more in my case and research. Also, try analysing it with a predetermined load function in your analytical process and see if you get the results. In my opinion, this should work.

Menaka Konara Hello. I think the mode superposition approach is a better one and worked more in my case and research. Also, try analysing it with a predetermined load function in your analytical process and see if you get the results. In my opinion, this should work.

Shahida Anusha Siddiqui I would try that method.Is it necessary to have a predetermined load with frequency response analysis? Does it affect the frequency response? I thought mode shapes in a system independent of external forces(ex-driving force in a forced vibration). Thank You!

Menaka Konara Hello. Here, have a look:

Modeling Frequency Response in COMSOL Multiphysics®

Setting Up the Study

After adding a structural mechanics physics interface in the Model Wizard, you will be presented with a number of study types, four of which can be used for computing frequency response:

Look at Capture image here

Available study types for a Solid Mechanics interface.

Two of the studies use a direct solution approach and two use the mode superposition approach. In the prestressed types of analysis, the change in stiffness from a stationary preload is taken into account. Mode superposition is very well suited for frequency-domain analysis, since it it easy to select the appropriate eigenmodes based on the given frequencies.

In either case, you perform a frequency sweep by providing a list of frequencies in the study settings for which the response is computed. Often, you want to cluster the frequencies around the natural frequencies of the structure.

Look at Capture 1 image here

Entering frequencies for a frequency sweep.

Note that without damping, the response exactly at a natural frequency tends toward infinity. This means that it is not possible to solve an undamped frequency response problem at, or close to, a natural frequency. The numerical formulation will give a singular, or at least ill-conditioned, system matrix.

Perturbation or Not?

There is a very important setting in the Stationary node in the solver sequence for a frequency-domain study: Linearity.

Look at Capture 2 image here

Selecting the Linearity property.

In principle, any frequency-domain analysis can be considered to be a small perturbation, so using Linear perturbation is never wrong. The most common case, however, is that the vibrations are centered around zero. In that case, it does not really matter whether the problem is considered as Linear or Linear Perturbation. The setting does, however, always fundamentally change the interpretation of loads. A load can be tagged as Harmonic Perturbation. Such a load is only taken into account if Linearity is set to Linear perturbation. All loads not marked as Harmonic Perturbation are ignored in such a study. Conversely, if Linearity is not Linear perturbation, then all loads marked as Harmonic Perturbation are ignored, and other loads are considered as harmonic.

Look at Capture 3 image here

An edge load, designated as Harmonic Perturbation.

The purpose of this setting is to be able to discriminate between loads causing a possible prestress state and the harmonic excitation acting on top of that.

When you add a standard Frequency Domain study, the study is, by default, not set as perturbation. Thus, the Harmonic Perturbation tag should not be used for the loads in this case, unless you change the Linearity setting. When you add a Frequency Domain, Prestressed study, the frequency response study step is set up for perturbation analysis. If the study is of a mode superposition type, then the study is always of a linear perturbation type.

Interpreting the Results

The results of a frequency-domain analysis are complex-valued and the harmonic variation is implicit. The phase angle of the complex number describes the phase shift with respect to the reference phase (which can be chosen arbitrarily, but is often taken as the phase of the main load). It also provides information about the phase shift between different points in the structure. Note that since the displacement components within a single finite element can have different phase angles, it is also quite possible that the components of the stress tensor are not in phase with each other. This can be of importance in, for example, fatigue analysis.

In many cases, like in a color plot, it is only possible to display a real number. The convention during all results presentation is as follows: If you request a complex-valued variable v in a context where a real value is expected, then the real part is used.

Look at Capture 4 here

The phase angle Φ is a property of the dataset that you can modify.

Look at Capture 5 here

Adjusting the phase angle in the dataset.

In most frequency-response analyses, you are interested in the amplitude of a result quantity, v, as function of frequency. This means that you should investigate abs(v) rather than v itself. The difference between the two is shown in the figure below.

Look at Capture 6 here

Example of a frequency response graph. Note that the graph of “u” is identical to “real(u)”.

In order to see what happens in more detail, we can add the imaginary part and argument of the result quantity to the graph:

Look at Capture 7 here

Frequency response including phase shift.

For low frequencies, the real part is close to the absolute value. In the vicinity of the natural frequency, the imaginary part is dominant instead. This means that the response is almost out of phase with the excitation.

Now, let’s investigate what happens if we change the phase angle in the dataset to 45°.

Look at Capture 8 here

Frequency response when the phase angle in the dataset is 45°.

As expected, the amplitude graph does not change. However, the individual values of the real and imaginary parts do. The phase angle curve shifts π/4 upward. Actually, this is the same exact graph that we would obtain if a 45° phase angle was added to the load.

Look at Capture 9 here

Adding a phase angle to a load.

Instead of using the phase angle input, you can equivalently enter the load directly using complex notation:

Look at Capture 10 here

Complex representation of the same load as above.

The possibility to prescribe the phase angle is important when not all loads are in phase with each other. A rotating unbalanced mass can, for example, be described conveniently by giving the load in the y direction a 90° phase shift with respect to the load in the x direction.

Results from a Perturbation Study

If the study is of the perturbation type, there will actually be two sets of results: the prestress solution and the perturbation solution. In this case, you will, in the various result presentation features, get access to an extra selection: Expression evaluated for.

Look at Capture 11 here

Selecting the evaluation type for a perturbation analysis.

Here, you can choose to study the perturbation solution, the prestress solution, or combinations thereof. For the perturbation solution, you also get one more option: the Compute differential check box.

Look at Capture 12 here

Selecting Compute differential.

This setting affects how nonlinear expressions are treated. When Compute differential is not selected, then a nonlinear quantity is taken at face value. For example, the expression u^2 will simply take the square of the variable u from the perturbation solution. Since u is, in general, complex-valued, this will usually be a nonsensical operation.

When Compute differential is selected, then the nonlinear quantity will be linearized around the prestressed state. The expression u^2 will evaluate to 2*u0*u, where u0 is the value at the linearization point.

Converting Frequency-Response Results to the Time Domain

There are some situations in which you may want to actually visualize the harmonic response from a frequency-domain analysis in the time domain. In particular, this is true if you have multiple excitation frequencies.

Look at Capture 13 here

Response to the excitation of two loads with different frequencies.

You can transform frequency-response results to the time domain using the Frequency to Time FFT study step.

Look at Capture 14 here

Study sequence for transforming results from the frequency domain to the time domain.

This technique is used in the following tutorial models:

• Vibration Analysis of a Thick Beam
• Viscoelastic Structural Damper –Transient Analysis

Concluding Remarks

Frequency-domain analysis is a powerful tool for analyzing linear systems subjected to harmonic excitation. Actually, by performing an initial Fourier transform of the loading, any type of periodic excitation can be studied using frequency-response analysis.

There are many more examples of mechanical frequency-response analyses available in the Application Gallery, such as:

• Elbow Bracket (the full version)
• Bracket (the bracket_frequency.mph model)
• Forced Vibration Analysis of a Composite Laminate
• Composite Piezoelectric Transducer