Speed Sweep Data Processing: RPM Increment, Framesize, Sweep Rate, and Overlap
This article explains some of the trade-offs encountered doing a spectral analysis of data from a speed sweep in Simcenter Testlab.
When processing data from a speed sweep, some important parameters include:
The RPM Increment and Framesize are available settings in Simcenter Testlab while the Sweep Rate is determined by the object being tested. Combined with the RPM Increment and Framesize, this determines the amount of overlap in the data being processed.
Overlap is desired to ensure no critical information is missed in the processed results. This article will explain overlap and illustrate the impact on the processed data results.
Simcenter Testlab Settings
Figure 1 shows some of the available processing settings a user can assign in Simcenter Testlab Throughput Processing. Similar settings can be done in Simcenter Testlab Signature for online acquisition.
In the “Tracking and Triggering” tab, a speed sweep analysis is to be performed:
Figure 2 shows the FS Acquisition tab.
The following spectral processing settings are shown:
With these settings, three different sets of data will be collected and processed, each at a different sweep rate:
The impact on the overlap will be shown in each scenario.
Scenario One: Processing Medium Speed Sweep
Using the settings from above, the first segment of time selected for an FFT calculation will occur at 1500 rpm, the second segment of time selected will occur at 1650 rpm, the third at 1800 rpm, and so on until the maximum rpm of 3000 rpm is reached. The first three acquisition frames are illustrated in Figure 3.
The processed result will create a colormap/waterfall display with 11 FFTs – one created at each 150 rpm increment over the entire sweep range. (There are 11 total FFTs calculated at 150 rpm increments over the minimum and maximum rpm settings).
In this scenario, the sweep rate was controlled such that a 150 rpm increment per second. The fixed sampling settings specify a frame size of 2 seconds. With these settings, and this sweep rate, a 50% overlap in the time data will occur for each spectrum.
Often Hanning windows are used in the spectral calculations. The Hanning window multiplies the beginning and end of the frame by zero, eliminating some data. With overlap, some of the data that is lost in one frame (due to the multiplication by zero) is picked up by the next frame.
Scenario Two: Processing Slow Speed Sweep
Suppose that the speed sweep was slowed down from 150 rpm per second to 37.5 rpm per second.
The processing settings are still the same. The data will be processed every 150 rpm. However, there will now be gaps in the data that is processed from the time history as shown in Figure 4.
With a 37.5 rpm per second increase, it takes four seconds to increase 150 rpm. With a two second frame, this leaves 2 seconds of unprocessed data each rpm increment. If important events happen between frames, they will not be processed.
In Figure 5, an from a spectral speed sweep analysis shows the importance of overlap.
Without a good amount of overlap, the processed results will miss information.
Scenario Three: Processing a Quick Sweep
In Scenario Three, the rpm sweep rate has been increased to 300 rpm per second.
Keeping the same acquisition settings and processing settings, there is a 75% overlap of the processed data as shown in Figure 6.
The time between each 150 rpm increment is 0.5 seconds. With the 150 rpm increment and two second frame size, there is a 75% overlap for each calculated spectrum.
While more overlap does not hurt the data quality, the computing speed and data size may be significantly higher, for little additional gain.
When tracking at a given rpm increment, the sweep rate and frame size become factors in setting an appropriate overlap for data processing. Depending on how these settings are chosen, order cuts, waterfalls, and, colormaps can convey different information, possibly leading to incorrect decisions regarding the results.
Questions? Email email@example.com or post a reply!