Experimental modal analysis in DIRAC – Workflow

In this article, you will learn all the spets needed to perform Experimental Modal Analysis (EMA) in DIRAC. The algorithm used to calculate poles and modes is described in detail in another article, that you can find here.

To use EMA in DIRAC, you need either the dir-modal-base license or dir-ema and dir-prepare. You do not need all DIRAC modules, so you can also use DIRAC just for performing modal analysis.

To perform EMA in DIRAC, we recommend following these steps:

  • Step 1: Choose the dataset.
  • Step 2: Run first computation with smart settings.
  • Step 3: Fine-tune the settings and manual pole picking.
  • Step 4: Visualize results.
  • Step 5: Export results.

Step 1: Choose the dataset

The first step consists of defining the dataset to use to perform EMA. You can perform EMA to measured FRFs or to Virtual Point Transformed FRFs (or both). To perform EMA in DIRAC, you do not necessarily need to perform FRF testing in DIRAC. You can also do an import campaign and import FRFs measured with another software.

For good EMA results, it is vital to have good-quality FRFs, so make sure you perform all quality checks on your data before moving to the EMA module. You can read about all DIRAC quality checks in this article (link). When you have checked your data, you can move on to the EMA module.

The very first step consists of selecting the dataset for performing modal analysis. You can select the desired preset from the drop-down in the Modal fit settings card. Here, you can also decide to reduce the data, by creating a custom preset. A custom preset can contain any amount of excitation, channels and or VPs.

To reduce the computation time, create a preset that only includes the data you are interested in.

Step 2: Run first computation with smart settings

After defining the dataset to use, it is necessary to check the settings. In DIRAC we have implemented smart settings, that automatically select the best settings for your dataset. These smart settings are set as Auto in the Modal Fit Settings card. However, before computing you still need to define some general settings. Afterwards, you are ready to run the first computation. In the Modal fit settings card you have to:

  1. Choose the frequency range of interest, by typing the minimum and maximum frequencies.
  2. Choose, if desired, the drive point strategy. You can have:
    • Single point drive points: you can select a single Channel and Reference channel that form a drive-point together (make sure that the chosen DoF are true drive points or can be assumed as such).
    • Virtual Point drive points: you can select one or multiple VPs, that offer multiple drive-point measurements (you also have the option to run it reciprocally).
    • None.
  3. Set the residue fit settings:
    • Residue fit method to linear (best option for the first calculation).
    • Mode type: for undamped systems, use Real (Normal) Modes; for damped systems with interacting modes, select Complex Modes.

After defining the general settings, you can compute the results by clicking the  button in the Modes card. These first results are a preliminary fit, which will be improved with some setting fine-tuning and doing manual pole picking if needed.

Define some general settings, but use the smart settings for the first computation.

To speed up the process of finding poles, start with a linear residues fit.

Step 3: Fine-tune the settings and manual pole picking

You have now made the first computation using the smart settings and DIRAC has found some poles and modes. Looking at the graphs, you can clearly see that some peaks are missing. From the stabilization diagram, you can notice that DIRAC found some poles but did not select them, because they were considered either unstable or stable only in frequency. DIRAC only selects poles if they are stable both in frequency and damping.

To improve your modal analysis results, you need to fine-tune the settings, to make sure you find all modes and that the original and synthesized FRFs match.

You can do this in multiple ways:

  • Edit pole selection settings.
  • Manually pick poles.
  • Run a  logarithmic fit.

Edit pole selection settings

Editing the pole selection settings (stability threshold, damping and frequency tolerance) can help improve the results. For example, you can reduce the Stability threshold and increase the Damping and frequency tolerance. After changing the settings, DIRAC needs to re-run some parts of the algorithm and re-do the computation. You can see that the results are not up to date from the status bar, at the bottom left of the window. Click the button to update the results.

You might need to perform this action multiple times before all poles are selected. However, you must consider not to increase/decrease them too much, to avoid unreasonable results. For this reason, it can be better to manually pick the last missing poles.

Edit the pole selection settings if multiple poles are not automatically selected by DIRAC.

Manually pick poles

A manual pole selection can be done to fine-tune the results. Any pole on the stabilization diagram can be toggled as stable and added to the list. The selection works per “pole group” ensuring that no two poles are selected per group. The choice of pole in the group is important, as they might have different damping ratio’s. For a better visualization, you can change the stabilization diagram settings, by plotting against damping.

To add a pole, just click on it in the stabilization diagram. After selecting all poles, compute the algorithm again, to update the results.

Manual pick poles to include them in the calculations.

Run a logarithmic fit

A linear fit is good for a quick check of the results of the pole selection, however, the logarithmic fit offers better results. The logarithmic fit is not sensitive to errors caused by ill-conditioned amplitude weighting which normally affects the residue fit. This issue normally causes bad fits in the lower frequency regime. With the logarithmic fit, this error is largely mitigated.

You want to run the algorithm with a logarithmic fit when the measured and synthesized curves have discrepancies in the anti-resonances. Note that this fit is non-linear and takes more time to run. You can change it in the Residue fit method  in the Modal fit settings card.

Run a logarithmic fit to improve fitting at the anti-resonances.

Step 4: Visualize results

After computing the results, you can visualize them in multiple ways:

  • Original, synthesized and fitted curves.
  • Modes list.
  • Mode shapes.
  • MAC matrix.
  • Similarity matrix.
  • Nyquist plot.


The original and reconstructed FRF matrices are plotted together in the Graphing card. Here, you can also visualize the similarity of the different FRFs over frequency. By right-clicking on the graph, you can select different curves to be shown, such as the fitted FRF (before globalization) or the synthesized FRF (after globalization).

Modes list

In the Modes card, you can find a list of all the calculated modes. For each mode, you can find the frequency and damping values. You can rename modes, as well as activate and deactivate them. You can also edit frequency and damping values, by simply double clicking on them and typing the desired values. If you have complex modes, you can also find the Modal Phase Collinearity (MPC), which says something about the quality of the complex modes.

Mode shapes

After selecting a mode from the Modes card, you can animate its mode shape in the 3D Viewer, by toggling the Mode shape toggle.

MAC matrix

You can visualize the MAC (Modal Assurance Criterion) matrix in the Modal matrix card, after clicking the MAC button. The MAC matrix can be used to analyze the orthogonality of the modes. Since we are using a normal mode model, we expect the mode shapes of each mode to be orthogonal to the others. If this is indeed the case, the MAC matrix is a diagonal matrix. Large off-diagonal elements indicate non-orthogonal modes. Selecting one of the diagonal elements in the matrix viewer will highlight the contribution of that mode to the FRF that is currently displayed in the Graphing card.

Similarity matrix

The similarity matrix shows the average similarity of the reconstructed FRFs to the measured ones, comparing synthesized and original FRFs. Clicking on an element of the matrix will display both the measured and reconstructed version of that FRF in the Graphing card.

Step 5: Export results

You can export EMA results in MATLAB and UFF formats. In particular, you can export:

  • Mode models:
    • Residue Model.
    • Mode Model.
  • FRF matrices:
    • Fitted FRF.
    • Synthesized FRF.
  • Matrix:
    • MAC.

You can copy and paste all modes from the Modes list (including frequency, damping and MCP values), by selecting them in the Modes card and using the shortcuts CTRL+C and CTRL+V.

    Contact us for more VIBES

    Contact our support team or call us on +31 85 822 50 49