6 Checking coherence

The coherence is a quality indicator that allows you to check the quality of your signals. It checks how repeatable measurements of a particular measurement point were performed.

You should check the coherence of your data after measuring the excitations of one VP. This allows to spot and correct problems before moving to the next VP. Integrating this check into your measurement process ensures high-quality test-based modelling. (link)

How to: Visualize the coherence

In Analyze, there are some tools to investigate the coherence of the test-based models in every level of detail. The coherence matrix of the measured data is the easiest way to have an overview of the entire dataset to quickly verify its coherence.

1. The matrix viewer gives the possibility to get an overview of all the sensors and excitations in the system and allows to quickly spot coherence problems.
2. The matrix mode can be easily set as coherence mode.
3. The depicted coherence is dependent on the frequency range defined below the matrix viewer.
4. The Validate card shows all the measurements of the selected entry in the matrix viewer (grouped by excitation source). It shows which ones are set as valid and which not.
5. In the Graphing area, you can visualize all measurements for that excitation point, being orange the selected one and blue the average of the valid ones.
6. The coherence plot shows the coherence of the valid impacts.

How to: Read the coherence matrix

To read the coherence matrix correctly to evaluate the quality of your FRFs, you have to:

1. Select the Measured preset in the matrix viewer.
2. Set the matrix viewer to coherence mode.
3. Average over the frequency range of interest.
4. Increase the minimum value of the coherence until you see white areas appearing in the matrix.

The white areas correspond to the FRFs with the lowest signal quality. If the whole matrix turns white at the same time, you have very similar signal quality in your data. If you’re satisfied with the minimum coherence, you can move on to the next quality checks.

However, if there are clear white patterns in the matrix (as in the pictures above), you have signal problems in the data that you need to react to.

How to: Interpret white rows

If you spot a white row when zooming in, this means that there’s a problem with that sensor channel. You have to identify the sensor, and check if there’s a broken part in its measurement chain. If the matrix row is only partly white, this would be an indicator that the sensor fell off at some point during the tests. To go, you have to fix the sensor and measure all impacts of this VP again.

How to: Interpret white columns

If you identify white columns while zooming in into the matrix, it means that the repeatability of an impact was relatively bad. You have to identify the impact and make some more measurements. If you cannot improve the coherence, you should consider moving it to a position that you can excite more precisely. It is also important to use the Automatic Measurement Selection, which will choose the most similar impacts for you.

How to: Interpret white blocks in symmetric structures

Symmetric VPs on a component should have similar coherence values. If the blocks in the matrix corresponding to these symmetrical VPs have different coherence, improve the impacts of the worse one.

How to: Interpret frequency-related problems

In the end, you should check whether the coherence is good for the entire frequency range you are interested in. If the coherence is low for the high or low frequencies, you should consider adding an input source with a different hammer tip.

When you use multiple input sources, also remember to check the coherence in the merging range. If the coherence drops in this frequency range, either the impacts with the two hammers are not similar and need to be repeated, or the cut-off and roll-off frequencies need to be adjusted.