Main Content

Process Data and Explore Features in Diagnostic Feature Designer

This example shows how to process your data in the app in preparation for feature extraction. If you want to follow along with the steps interactively, use the data you imported in Import and Visualize Ensemble Data in Diagnostic Feature Designer. Use Open Session to reload your session data using the file name you provided.

The Open Session button is the second icon from the left.

If you have no session data, execute the steps for loading and importing data in Import and Visualize Ensemble Data in Diagnostic Feature Designer.

A key step in predictive maintenance algorithm development is identifying condition indicators. Condition indicators are features in your system data whose behavior changes in a predictable way as the system degrades. A condition indicator can be any feature that is useful for distinguishing normal from faulty operation or for predicting remaining useful life. A useful feature clusters similar system statuses together and sets different statuses apart.

Diagnostic Feature Designer lets you design features that provide these diagnostics.

  • For some features, you can generate features directly using signals you imported.

  • For other features, you must perform additional signal processing, such as filtering and averaging, to obtain meaningful results.

The processing you perform depends both on the computational requirements of the feature and the characteristics of your systems and your system data. This example shows how to:

  • Process your data in preparation for feature extraction

  • Generate various types of feature

  • Interpret the effectiveness of your features in histograms

Perform Time-Synchronous Averaging

The data for this system represents a transmission system with rotating parts. The variables include tachometer outputs that precisely mark the completion of each shaft revolution. The data, therefore, is an ideal candidate for time-synchronous averaging.

Time-synchronous averaging (TSA) is a common technique for analyzing data from rotating machinery. TSA averages rotation by rotation, and filters out any disturbances or noise that is not coherent with the rotation.

TSA is useful for isolating fault signatures that repeat each rotation, such as perturbations from gear-tooth defects. Features generated from a TSA signal rather than the original vibration signal provide clearer differentiation for rotational fault conditions. This advantage holds true even for features that are not specifically for rotating machinery.

To compute the TSA of the vibration data, first select the signal to average, Vibration/Data, in the data browser. Then, select Filtering & Averaging > Time-Synchronous Averaging.

The signal Vibration/Data is selected in the column on the left. Time-Synchronous Averaging is the top option in the column on the right.

A new Time-Synchronous Averaging tab appears. The app title bar above the tab section displays Vibration/Data, the signal you are processing.

The Tacho Signal selection and the Compute Nominal Speed option are the second and third items on the left.

Since you have a tacho signal, select Tacho signal and Tacho/Data. You can leave Compute Nominal Speed (rpm) selected, but this tutorial does not use the nominal rpm information.

Beneath the tab, the plot tab Data Processing: Vibration/Data displays the source signal for the TSA signal.

Signal trace plot of vibration/data signal with two colors, similar to the signal trace plot in Part 1 of the tutorial

Click Apply to start the TSA computation for each of the 16 members of the ensemble. A progress bar shows the status while the computation progresses.

The Vibration_tsa/Data signal is the third item from the top in the column on the left. The TSA plot is in the main pane.

When the computation concludes:

  • The app adds a new signal variable Vibration_tsa to the imported Ensemble1 data set.

  • The signal trace plots Vibration_tsa. The time axis of this trace is less than four seconds long. The original vibration data was 30 seconds long. The shorter timespan reflects the duration of a single rotation for each member.

  • The member shaft rates diverge. This divergence is evident in the increasing misalignment of the peaks during the rotation, and the fact that the member traces stop at different times.

Once the TSA computation is complete, both the Time-Synchronous Averaging tab and the Data Processing tab disappear. If you want to return to the Time-Synchronous Averaging tab, click the plot tab Data Processing: Vibration/Data. The app restores both toolstrip tabs, with the TSA tab active and the data processing tab inactive.

Data Processing plot of Vibration/Data signals, wit Time-Synchronous Averaging tab above

Inactive Data Processing tab

If you want to perform similar processing on another variable, click Close TSA. The data processing tab activates. From that tab, you can change the signal to process. Then, from the data processing gallery, you can select TSA processing or any other processing that is compatible with the signal selection. The process and tab that you select retains any settings that you specified previously in the session.

Active data processing tab

Compute Power Spectrum

The TSA signal gives you enough information to start generating time-domain features, but you must provide a spectrum to explore spectral features. To generate a power spectrum, select the new TSA signal Vibration_tsa/Data in the data browser. Then, click Spectral Estimation to bring up the spectrum options. From these options, select Autoregressive Model.

Signals & Spectra pane with the Vibration-tsa signal selected on the left, and Spectral Estimation list on the right. The Autoregressive model is the second option on the list.

The Autoregressive Model tab provides parameters that you can modify. Accept the default values by clicking Apply.

The title bar at the top identifies Vibration_tsa/Data as the source signal. The Autoregressive tab shows four sections, from left to right the parameters for the frequency grid, the model parameters, the plot results option, and the Apply and Close AR Model buttons.

The power spectrum processing results in a new variable, Vibration_ps/SpectrumData. The associated icon represents a frequency response.

You can determine the source of the new spectrum (the original signal from which new spectrum was derived) by pausing on the spectrum name in the data browser. The following figure shows the resulting tooltip, which contains the information that the signal source is Vibration_tsa/Data, which in turn has the source Vibration/Data.

A plot of the spectrum appears in the plot area. As with Signal Trace, a Power Spectrum tab provides options for plotting. These options are similar to Signal Trace. The plot has no Panner option because Panner works only with time and not frequency.

The new spectrum variable is the last item in the list on the left. A tooltip describing the variable sources is below. The large pane on the right contains the power spectrum plot.

Generate Features

Signal Features

Generate features based on general data statistics, using the TSA signal as your source. Select Time Domain Features > Signal Features.

Signals & Spectra pane with the Vibration-tsa signal selected on the left, and Time-Domain Features list on the right. Signal Features is the first option on the list

As with data processing, preselect your source signal before choosing a feature option. Select Vibration_tsa/Data and then click Signal Features to bring up the Signal Features tab. Above the tab, the app displays the selected source signal, Vibration_tsa/Data. By default, all features are selected. Clear the selections for Shape Factor and Signal Processing Metrics.

Signal Features tab, with Select All and Unselect All buttons on the left. The next three tab sections are, from left to right, statistical features, impulsive features, and harmonic features. The rightmost two sections contain the plot and close buttons. The Shape Factor option is cleared in the statistical features section. The harmonic feature options are all cleared.

For every selected feature, the app computes a value for each ensemble member and displays the results in a histogram. Each histogram contains bins containing the number of feature values that fall within the bin range. The Histogram tab displays parameters that determine the content and resolution of the histograms.

The histogram groups, or color codes, the data according to the condition variable faultCode in Group By. Blue data is healthy (faultCode = 0) and orange data is degraded (faultCode = 1), as indicated by the legend (color coding might appear different in your session). For feature values where the healthy and degraded labels overlap, the color appears brown because of the overlap between blue and orange.

The contents of the feature table are shown in the middle of the pane on the left. The main pane shows the histograms

You can get a rough idea of which features are effective by assessing which ones clearly segregate blue data from orange data. RMS and CrestFactor appear effective, as they have only small areas of overlap. Conversely, Skewness and Kurtosis have large amounts of overlap. These features appear ineffective for this data and this condition variable.

By default, the app plots the histograms for all the features in the feature table. You can focus on a subset of histograms by using Select Features. Use Select Features to limit the histogram plots to the first four in the feature table.

The feature list contains nine items. The first four items are selected.

The histogram view now includes only the features you select.

The histogram pane contains four histograms.

Control the appearance of the histograms using the parameters in the Histogram tab, which activates when you generate the histograms. The CrestFactor feature appears to separate healthy and unhealthy data almost completely. Investigate whether this result is sensitive to resolution. In the Histogram tab, the auto setting of bin width results in a resolution of 0.1 for CrestFactor. Enter a bin width 0.05, and click Apply.

The Options and Apply sections of the histogram are shown at the top. Bin Width is at the top of the left column of the Options section. The Apply button is on the right. The histograms are below the tab.

At this resolution, both CrestFactor and ImpulseFactor appear to completely separate healthy from degraded data. ClearanceFactor still has some mixed data, but to a lesser degree than with the larger bin width. Kurtosis has a smaller bin width of 0.002 with the auto bin width setting. Changing the bin width to 0.05 results in a single bin that contains all the Kurtosis data.

Histograms visualize the ability of features to separate healthy from unhealthy data. You can also get a numerical assessment using Group Distance. The group distance represents the separation between the healthy and unhealthy data distributions. Click Group Distance. In the dialog box, select CrestFactor in Show grouping for feature.

The Histogram tab is at the top and consists of View, Grouping, and Options sections from left to right. Show Group Distance is the lower option in the Grouping section. Beneath the tab is the dialog box that displays the group distance information. The title bar displays the condition variable faultCode. The top item within the dialog box displays the feature name. The table beneath the name shows, from left to right, the label for each group and the KS statistic.

The group distance, represented by the KS Statistic, is 1. This probability value represents complete separation.

Next, select Kurtosis. The Kurtosis histogram shows substantial intermixing.

The dialog box for Kurtosis has the same layout as for Crest Factor.

The KS Statistic in this case is about 0.6, reflecting the intermixing in the histogram.

Restore Bin Width to auto.

Rotating Machinery Features

Since you have rotating machinery, compute rotating machinery features. In the data browser, select your TSA signal. Then, select Time-Domain Features > Rotating Machinery Features.

Signals & Spectra pane with the Vibration-tsa signal selected on the left, and Time-Domain Features list on the right. Rotating Machinery Features is the second option on the list

The Rotating Machinery Features tab can create features from TSA signals as well as from TSA difference signals and TSA regular signals. Since you have only a TSA signal, the app disables the selections that do require a different signal type.

The Rotating Machinery tab contains the following sections from left to right: Signals to use, Metrics using TSA Signal, metrics using Difference Signal, Metrics Using Mixed Signals, Plot, and Apply.

Accept the default selections by clicking Apply.

The middle pane of the data browser on the left contains the new feature names. The top row of the histogram array contains the new feature histograms.

The app automatically adds the new features to the feature table and the Select Features list, and plots the new histograms at the top of the histogram display. CrestFactor and Kurtosis histograms are essentially the same whether they are computed as signal features or rotating machinery features, since both computations use the TSA signal as a source.

Spectral Features

Compute spectral features from the power spectrum you generated earlier. Select Vibration_ps/SpectrumData. Then, select Frequency-Domain Features > Spectral Features.

Signals & Spectra pane with the Vibration-ps/SpectrumData signal selected on the left, and Frequency-Domain Features list on the right. Spectral Features is the first option on the list.

Specify the frequency band to use by setting the minimum and maximum band values. To capture the power spectrum peaks efficiently, limit the frequency range to a maximum of 10 Hz, starting at 0.001 Hz. The plot displays this band as an orange rectangle underlaying the frequency plot.

The Spectral Features tab contains the following sections from left to right: Frequency Band, Spectral Peaks, Modal Coefficients, Band Power, Plot, Apply, and Close. The maximum frequency is the lower item in the Frequency Band section.

The histograms show substantial intermixing of healthy and unhealthy data in one or more of the bins for all three features.

You now have a diverse set of features.

Save your session data. You need this data to run the Rank and Export Features in Diagnostic Feature Designer example.

The Save Session button is the rightmost one in the File portion of the Feature Designer tab.

Next Steps

The next step is to rank those features to determine which ones provide the best indication of system condition. For more information, see Rank and Export Features in Diagnostic Feature Designer.

See Also

Related Topics