Import and Visualize Ensemble Data in Diagnostic Feature Designer

Step 1 of 3 in Identify Condition Indicators for Predictive Maintenance Algorithm Design

Diagnostic Feature Designer is an app that allows you to develop features and evaluate potential condition indicators using a multifunction graphical interface.

The app operates on data ensembles. An ensemble is a collection of data sets, created by measuring or simulating a system under varying conditions. An individual dataset representing one system under one set of conditions is a member. Diagnostic Feature Designer processes all ensemble members when executing one operation.

This example shows how to import data into Diagnostic Feature Designer and visualize your imported data.

Load Transmission Model Data

This example uses data generated from a transmission system model in Using Simulink to Generate Fault Data. Outputs of the model include:

• Vibration measurements from a sensor monitoring casing vibrations

• Data from the tachometer, which issues a pulse every time the shaft completes a rotation

• Fault codes indicating the presence of a modeled fault

Load the data. The data is a table containing variables logged during multiple simulations of the model under varying conditions. Sixteen members have been extracted from the transmission model logs to form an ensemble. Four of these members represent healthy data, and the remaining 12 members exhibit varying levels of sensor drift.

load dfd_Tutorial dataTable

View this table in your MATLAB^® command window.

dataTable =

  16×3 table

        Vibration               Tacho           faultCode
    __________________    __________________    _________

    [6000×1 timetable]    [6000×1 timetable]        0    
    [6000×1 timetable]    [6000×1 timetable]        1    
    [6000×1 timetable]    [6000×1 timetable]        1    
    [6000×1 timetable]    [6000×1 timetable]        1    
    [6000×1 timetable]    [6000×1 timetable]        1    
    [6000×1 timetable]    [6000×1 timetable]        1    
    [6000×1 timetable]    [6000×1 timetable]        1    
    [6000×1 timetable]    [6000×1 timetable]        1    
    [6000×1 timetable]    [6000×1 timetable]        0    
    [6000×1 timetable]    [6000×1 timetable]        0    
    [6000×1 timetable]    [6000×1 timetable]        0    
    [6000×1 timetable]    [6000×1 timetable]        1    
    [6000×1 timetable]    [6000×1 timetable]        1    
    [6000×1 timetable]    [6000×1 timetable]        1    
    [6000×1 timetable]    [6000×1 timetable]        1    
    [6000×1 timetable]    [6000×1 timetable]        1  

The table contains 16 rows, each representing one member. Each column has a variable name. The data variables Vibration and Tacho are each represented by a timetable, and all timetables have the same length. The third variable, faultCode, is the condition variable. faultCode has a value of 0 for healthy and 1 for degraded.

Open Diagnostic Feature Designer

To open Diagnostic Feature Designer, type the following in your command window:

diagnosticFeatureDesigner

Import Data

Import the dataset that you previously loaded into your MATLAB workspace. To initiate the import process, in the Feature Designer tab, click New Session.

The New Session dialog box opens. From the Source Choose Variable list in the Select more variables pane, select dataTable.

The Source variablespane of the dialog box now displays the variables within dataTable. By default, the app initially selects all source variables for import.

The app has extracted the variable names from your member tables and embedded timetables. The icon next to the Vibration and Tacho variable names indicates that the app interprets the variables as time-based signals that each contain Time and Data variables. You can verify this interpretation in the Summary pane at the bottom, which displays the variable name, type, and independent variable for each source-level variable.

A third variable, Sample (Virtual), also appears in the list, but is not selected and does not appear in Summary. The import dialog always includes this variable as an option to allow you to generate virtual independent variables within the app. Because Vibration and Tacho both contain this unselected variable, the selection boxes contain fill rather than a check mark.

View the properties of Vibration by selecting the Vibration row.

The Source variable properties pane displays the Vibration variable name and the Signal variable type. For Vibration, Signal is the only Variable type option because the vibration data is packaged in a timetable. Source variable properties also displays a preview of Vibration data.

Now examine the variable type of faultCode. The icon next to faultCode, which illustrates a histogram, represents a feature. Features and condition variables can both be represented by scalars, and the app cannot distinguish between the two unless the condition variable is a categorical. To change the variable type, click on the faultCode to open the its properties, and, in Variable type, change Feature to Condition Variable.

The icon for faultCode now illustrates a label, which represents a condition variable.

Confirm the Ensemble Specification in Summary and click Import.

Your imported signals are now in the Signals & Spectra area, and your imported ensemble Ensemble1 is in the Datasets area.

The color code next to a signal represents that signal in plots. The icon to the left of the signal indicates the variable type, which, for the variables you imported, is Signal.

Show information about your dataset by selecting its name in the Datasets area.

Visualize Data

After you load your signals, plot them and view all your ensemble members together. To view your vibration signal, in the Signals & Spectra pane, select Vibration/Data. Selecting a signal variable enables the Signal Trace option in the plot gallery. Click Signal Trace.

The plotting area displays a signal trace plot of all 16 members. As you move the cursor over the data, an indicator in the lower right corner identifies the member your cursor is on. A second indicator provides the fault code value for that member.

Interact with the trace plot using standard MATLAB plot tools, such as zoom and pan. Access these tools by pointing to the top right edge of the plot. You can also use the specialized options on the Signal Trace tab, which appears when you select the Signal Trace plot.

Explore Your Data Using Signal Trace Options

Explore the data in your plot using options in the Signal Trace tab.

Measure the distance between peaks for the one of the members with the high peaks.

1. Zoom in on the second clusters of peaks. In the panner strip, move the right handle to about 8. Then, move the panner window so that the left handle is at about 4. You now have the second set of peaks within the window.

2. Pause on the first high peak, and note the member number. The second high peak is a continuation of the same member trace.

3. Click Data Cursors, and select Vertical Cursor. Place the left cursor over the first high peak and the right cursor over the second peak for that member. The lower right corner of the plot displays the separation dX.

4. Select Lock Horizontal Spacing. Shift the cursor pair to the right by one peak for the same member. Note that the right cursor is now aligned with the third member peak.

5. Restore the full window by moving the handles back to the edges of the panner.

Show which members have matching faultCode values by using color coding. Select Ensemble View Preferences > Group by "faultCode".

The resulting signal trace shows you that all the highest vibration peaks are associated with data from faulty systems. However, not all the faulty systems have higher peaks.

Save your session data. You need this data to run the Process Data and Explore Features in Diagnostic Feature Designer example.

Next Steps

The next step is to explore different ways to characterize your data through features. The example Process Data and Explore Features in Diagnostic Feature Designer guides you through the feature exploration process.