Simcenter Testing Solutions Cepstrum in Simcenter Testlab
2024-11-30T17:03:50.000-0500
Simcenter Testing Solutions Marketing
Simcenter SCADAS
Simcenter Testlab
Simcenter Testlab RT
Teamcenter Share
Simcenter Testlab Apps
Simcenter Testxpress
Simcenter Tecware
Simcenter Sound Camera
Other Hardware
Simcenter Soundbrush
Simcenter IQT
Summary
Details
This article covers how to apply the principles of cepstrum to real data using Simcenter Testlab.
The basic steps for calculating a cepstrum and the needed Simcenter Testlab software modules are shown in Figure 1 below.
Figure 1: Top - Steps to perform cepstral analysis, Bottom - Required Simcenter Testlab software modules.
The different steps are:
1.Getting Started with Simcenter Testlab
1.1 Requirements
1.2 Starting from Time History Data
1.3 Starting from Frequency Data
2. Interpreting Results
2.1 Simcenter Testlab Gear Example
2.2 Anovis Gear Example
1.Getting Started with Simcenter Testlab
To perform a cepstral analysis, a time file and specific software is required.
1.1 Requirements
To use Simcenter Testlab, a license for "Simcenter Testlab Desktop Standard & Advanced" is required. If only Desktop Standard is present, the Data Calculator is not available which is needed for the log and inverse Fourier Transform. Data calculator does not require additional tokens when using Simcenter Testlab Advanced Desktop.
If starting with time data, it needs to be converted to the frequency domain. "Signature Throughput Processing" add-in (36 tokens) is used for this.
As of release 2406, Simcenter Testlab Neo does not support cepstrum processing.
1.2 Starting from Time History Data
After recording the time data of the test object during operation, start by opening up Simcenter Testlab Desktop. If the data is already a frequency spectrum, skip to the next section.
Turn on the "Signature Throughput Processing" add-in by going to "Tools -> Add-ins" in the main Simcenter Testlab menu (Figure 2).
Figure 2: Pop up window of Simcenter Testlab add-ins with the “Signature Throughput Processing” add-in selected.
Figure 3: Right click on the folder or data files to be analyzed and select “Add to Input Basket”.
The input basket separates the data wanted for processing from all other data in the project, and it is convenient way to select what will get processed.
Select the tab at the bottom labeled "Time Data Selection", and then click the "Add" button to add the input basket data to the processor. If this step is skipped, no data will be processed (Figure 4).
Figure 4: Adding time data to the “Time Data Selection” tab.
Optionally select the check boxes by the data to see a display of the time data in the graph.
Next, move to the next tab at the bottom labeled "Time Data Processing". This is the tab where the settings for the frequency plot will be selected. Select "Change Settings" under the "Acquisition Parameters" menu (Figure 5).
Figure 6: Fixed sampling acquisition parameters. Selecting a resolution for the frequency data.
Selecting a proper frequency resolution is important. The right frequency resolution will have an impact on the quality of data seen in the cepstrum. Since the cepstrum is inversely proportional to the frequency domain, an 8 Hz resolution in the frequency domain will allow for a 1/8 quefrency resolution. If finer resolution is chosen for the frequency calculation (such as 0.5 Hz, 1 Hz, etc) then, the cepstrum will have a low resolution and peaks in the data may be difficult to see. It is recommended that the coarsest frequency resolution possible is chosen. In this instance, 8 Hz resolution will be selected for the example.
Then select "Channel processing" settings as shown in Figure 7.
Figure 7: "Channel Processing" settings for a vibration channel, with “Autopowers Linear” and “Hanning” parameters selected.
If using acoustic data, change any parameters desired on the acoustic tab. In this example, vibration data was taken, so the vibration tab is used to alter parameters. The important part in this selection is make sure that "Function" is set to "Autopowers Linear". This will take the magnitude of the Fourier Transform which is important for cepstrum processing. A good rule of thumb is to use a Hanning window to prevent leakage.
Lastly, select the check box next to all channels that are desired for processing, and click the "Calculate" button (Figure 8).
Figure 8: Calculating the frequency data from the time history by clicking “Calculate”.
The data will now appear in the navigator as frequency data.
More information in the knowledge articles:
If the data is already processed in the frequency domain, then the cepstrum calculations can begin here. First, find the frequency data in the "Navigator" tab. It doesn’t need to be added to the input basket this time, but the location should be available for clicking and dragging. Then, at the top of the navigator pane is a tab called "Data Calculator" (Figure 9).
"Data Calculator" is the tab where the processing will be done. Click on the Data Calculator tab, and then add the frequency data to the "Data Set" list, by highlighting the data, and clicking “Add". The “Add" function is the first button located in the "Data Set" list.
Recall that the way to calculate a cepstrum is to take the log, and then the inverse Fourier transform of a frequency spectrum (Figure 1). These two formulas will need to be entered. The Fourier transform was already performed to get the data into the frequency domain, so only the last two parts of the algorithm are needed here.
Similar to Time Signal Calculator, formulas can be created using the data in the "Data Set" list as variables. Click into a formula cell and select the "New Formula" button. In the new formula tab pop-up screen, find the "LOG10" function, either in the list or by narrowing the list down by clicking "Math" (Figure 10).
Figure 10: Selecting the "LOG10" function in the “Select Function” pop-up menu.
For function type in the variable ID from the Data Set list that you wish to process. In this example, the ID is F1.
Then enter the formula. In this instance, the result is output as variable ID, "F2", which is the log of the frequency data. The Point Id is "log".
The next step is to take the inverse Fourier transform, or IFFT, of the log data. Click on the cell below the previous one in the formula set, and select "New Formula" again (Figure 11).
This time, find the function in the pop-up menu called "FFT_INVERSE". Apply this to the ID that matches the log data. In this example, that was "F2". You can name this last cell in the “Point Id” something meaningful, such as "Cepstrum Calculation".
Lastly, after the formulas have been entered, click the "Calculate" button in the lower right corner (Figure 12).
The data will not be calculated and should appear as new rows in the “Data Set” list. In Simcenter Testlab, the cepstrum data will appear as an “AutoCorrelation”.
As shown in Figure 13, click on the row with the cepstrum calculation, and drag it into a new picture.
Figure 13: Drag and drop from the Dataset area to a FrontBack display to view the cepstrum.
When the data is first dragged into a blank Front/Back display from the Dataset area, the mouse turns to a four-way arrow when the data can be dragged.
The cepstrum will appear as a sharp peak at zero, and a mirror spectrum about zero. This is because the IFFT function in Simcenter Testlab reports data as a 2-sided FFT. This is a mathematical artifact, and is expected.
However, since the negative frequencies are not useful to look at in the cepstrum, the following display changes are recommended:
Questions? Email steff.nelson@siemens.com.
Related Links:
The basic steps for calculating a cepstrum and the needed Simcenter Testlab software modules are shown in Figure 1 below.
Figure 1: Top - Steps to perform cepstral analysis, Bottom - Required Simcenter Testlab software modules.
The different steps are:
- Fourier Transform: Simcenter Testlab Throughput Processing
- Log: Simcenter Testlab Data Calculator
- Inverse Fourier Transfrom: Simcenter Testlab Data Calculator
1.Getting Started with Simcenter Testlab
1.1 Requirements
1.2 Starting from Time History Data
1.3 Starting from Frequency Data
2. Interpreting Results
2.1 Simcenter Testlab Gear Example
2.2 Anovis Gear Example
1.Getting Started with Simcenter Testlab
To perform a cepstral analysis, a time file and specific software is required.
1.1 Requirements
To use Simcenter Testlab, a license for "Simcenter Testlab Desktop Standard & Advanced" is required. If only Desktop Standard is present, the Data Calculator is not available which is needed for the log and inverse Fourier Transform. Data calculator does not require additional tokens when using Simcenter Testlab Advanced Desktop.
If starting with time data, it needs to be converted to the frequency domain. "Signature Throughput Processing" add-in (36 tokens) is used for this.
As of release 2406, Simcenter Testlab Neo does not support cepstrum processing.
1.2 Starting from Time History Data
After recording the time data of the test object during operation, start by opening up Simcenter Testlab Desktop. If the data is already a frequency spectrum, skip to the next section.
Turn on the "Signature Throughput Processing" add-in by going to "Tools -> Add-ins" in the main Simcenter Testlab menu (Figure 2).
Figure 2: Pop up window of Simcenter Testlab add-ins with the “Signature Throughput Processing” add-in selected.
In this example, we will be looking at two sets of data, one set that contains a gear defect, and one set that does not. In the "Navigator" pane, locate the data, right click, and select “Add to Input Basket" (Figure 3).
Figure 3: Right click on the folder or data files to be analyzed and select “Add to Input Basket”.
The input basket separates the data wanted for processing from all other data in the project, and it is convenient way to select what will get processed.
Select the tab at the bottom labeled "Time Data Selection", and then click the "Add" button to add the input basket data to the processor. If this step is skipped, no data will be processed (Figure 4).
Figure 4: Adding time data to the “Time Data Selection” tab.
Optionally select the check boxes by the data to see a display of the time data in the graph.
Next, move to the next tab at the bottom labeled "Time Data Processing". This is the tab where the settings for the frequency plot will be selected. Select "Change Settings" under the "Acquisition Parameters" menu (Figure 5).
Figure 5: Using the "Time Data Processing" tab to transform the time data into frequency data.
This should bring up a pop up window of the Fourier transform settings applied to the data. In this tab, the number of averages and type of averaging can be selected and will depend on the type of data taken. To change the frequency resolution, select the "FS Acquisition" tab at the top of the "Acquisition Parameters" pop up window (Figure 6).
Figure 6: Fixed sampling acquisition parameters. Selecting a resolution for the frequency data.
Selecting a proper frequency resolution is important. The right frequency resolution will have an impact on the quality of data seen in the cepstrum. Since the cepstrum is inversely proportional to the frequency domain, an 8 Hz resolution in the frequency domain will allow for a 1/8 quefrency resolution. If finer resolution is chosen for the frequency calculation (such as 0.5 Hz, 1 Hz, etc) then, the cepstrum will have a low resolution and peaks in the data may be difficult to see. It is recommended that the coarsest frequency resolution possible is chosen. In this instance, 8 Hz resolution will be selected for the example.
Then select "Channel processing" settings as shown in Figure 7.
Figure 7: "Channel Processing" settings for a vibration channel, with “Autopowers Linear” and “Hanning” parameters selected.
If using acoustic data, change any parameters desired on the acoustic tab. In this example, vibration data was taken, so the vibration tab is used to alter parameters. The important part in this selection is make sure that "Function" is set to "Autopowers Linear". This will take the magnitude of the Fourier Transform which is important for cepstrum processing. A good rule of thumb is to use a Hanning window to prevent leakage.
Lastly, select the check box next to all channels that are desired for processing, and click the "Calculate" button (Figure 8).
Figure 8: Calculating the frequency data from the time history by clicking “Calculate”.
The data will now appear in the navigator as frequency data.
More information in the knowledge articles:
- Digital Signal Processing: Sampling Rates, Bandwidth, Spectral Lines, and more…
- Window Types: Hanning, Flattop, Uniform, Tukey, and Exponential
- Simcenter Testlab Throughput Processing Tips
If the data is already processed in the frequency domain, then the cepstrum calculations can begin here. First, find the frequency data in the "Navigator" tab. It doesn’t need to be added to the input basket this time, but the location should be available for clicking and dragging. Then, at the top of the navigator pane is a tab called "Data Calculator" (Figure 9).
Figure 9: The "Data Calculator" (top) is located in the Navigator tab (bottom).
"Data Calculator" is the tab where the processing will be done. Click on the Data Calculator tab, and then add the frequency data to the "Data Set" list, by highlighting the data, and clicking “Add". The “Add" function is the first button located in the "Data Set" list.
Recall that the way to calculate a cepstrum is to take the log, and then the inverse Fourier transform of a frequency spectrum (Figure 1). These two formulas will need to be entered. The Fourier transform was already performed to get the data into the frequency domain, so only the last two parts of the algorithm are needed here.
Similar to Time Signal Calculator, formulas can be created using the data in the "Data Set" list as variables. Click into a formula cell and select the "New Formula" button. In the new formula tab pop-up screen, find the "LOG10" function, either in the list or by narrowing the list down by clicking "Math" (Figure 10).
Figure 10: Selecting the "LOG10" function in the “Select Function” pop-up menu.
For function type in the variable ID from the Data Set list that you wish to process. In this example, the ID is F1.
Then enter the formula. In this instance, the result is output as variable ID, "F2", which is the log of the frequency data. The Point Id is "log".
The next step is to take the inverse Fourier transform, or IFFT, of the log data. Click on the cell below the previous one in the formula set, and select "New Formula" again (Figure 11).
Figure 11: Entering the second formula, the Inverse Fourier Transform equation into the formula set.
This time, find the function in the pop-up menu called "FFT_INVERSE". Apply this to the ID that matches the log data. In this example, that was "F2". You can name this last cell in the “Point Id” something meaningful, such as "Cepstrum Calculation".
Lastly, after the formulas have been entered, click the "Calculate" button in the lower right corner (Figure 12).
Figure 12: Calculating the cepstrum in “Data Calculator”.
The data will not be calculated and should appear as new rows in the “Data Set” list. In Simcenter Testlab, the cepstrum data will appear as an “AutoCorrelation”.
As shown in Figure 13, click on the row with the cepstrum calculation, and drag it into a new picture.
Figure 13: Drag and drop from the Dataset area to a FrontBack display to view the cepstrum.
When the data is first dragged into a blank Front/Back display from the Dataset area, the mouse turns to a four-way arrow when the data can be dragged.
The cepstrum will appear as a sharp peak at zero, and a mirror spectrum about zero. This is because the IFFT function in Simcenter Testlab reports data as a 2-sided FFT. This is a mathematical artifact, and is expected.
However, since the negative frequencies are not useful to look at in the cepstrum, the following display changes are recommended:
- Lower Limit Zero: Change the X-axis to start at zero.
- Zoom Y-Axis: The initial peaks will be hard to see, so you may need to zoom in, to find the peaks. This can be done in a few different ways but the easiest is holding and dragging over the area you want to zoom in on.
Questions? Email steff.nelson@siemens.com.
Related Links:
- Index of Simcenter Testing Knowledge Articles
- Cepstrum Analysis
- Simcenter Testlab
- Simcenter Testlab Throughput Processing Tips
- Simcenter Testlab Data Calculator
- What is the Fourier Transform?
- Spectrum versus Autopower
- Digital Signal Processing: Sampling Rates, Bandwidth, Spectral Lines, and more…
- Windows and Spectral Leakage
- Window Types: Hanning, Flattop, Uniform, Tukey, and Exponential