Wavelet Homepage                               Table of Contents

5 Analysis of Test Data from Q Group Spring 1998        

In the spring of 1998, an ASE 463Q design group performed a variety of experiments on a Star-Lite home-built aircraft under ideal lab conditions as in Figure 14.  Four strategic locations were selected to detect the signal from a variety of mechanical inputs.  One wire was routed around the nose of the plane, just behind the engine (channel 1).  Another was run in the same manner roughly two feet behind the first wire (ch2).  A bundle of wires was run along the starboard side of the fuselage (ch3), and a bundle of wires was run down the port side (ch4).  The wires were held in place by Velcro strips with one side taped along the plane and the other wrapped around the wire bundle.  An audio channel was also recorded the acoustic signal (ch5).  For a complete description of the test setup please see reference 6.

Figure 14: Star-Lite Aircraft Used in Testing

 

.  The sampling rate of the machine used by the Q group was 16384 samples/sec.  The frequency bands for the resulting levels of wavelet decomposition are given in Table 2.  Note that different codes often produce plots with the levels labeled in reverse order.  In that case, simply reverse the order of column 1.

Table 2: Frequency Bands for given Decomposition Levels

 

 

Many of the tests performed by this Q group duplicate the mechanical events thought to occur in the last 30 minutes of the Beech 1900C flight.  The four main tests of interest to us are an impulse to the main support strut, the right wing, and the tail, also the break of the main landing gear support strut.  These tests will serve as examples of the ability of the wavelet technique to isolate events, as well as provide references for comparison to actual data.

  5.1 Impulse Hammer Strike to Right Wing

            The total recorded time for the impulse hammer strike to the right wing test was one second.  The wavelet analysis results and original time-series for each channel are given below.

Figure 15: Wavelet Analysis of Hammer Strike to Wing in Channel 1

 

 

Figure 16: Wavelet Analysis of Hammer Strike to Wing in Channel 2

Figure 17: Wavelet Analysis of Hammer Strike to Wing in Channel 3

 

Figure 18: Wavelet Analysis of Hammer Strike to Wing in Channel 4

 

Figure 19: Wavelet Analysis of Hammer Strike to Wing in Channel 5

Inspection of Figures 15 through 19 shows that the acoustic signal was much stronger than the observed inputs to the wires.  In fact, only the starboard side wires pick up the signal at all.  This is not too surprising given that the vibrations caused by the impulse may not have extended past the fuselage of the Star-Lite.

  5.2 Impulse Hammer Strike to Tail of Plane

            This test setup is the same as that for the hammer strike on the plane’s starboard wing, except that two impulses were recorded.  The full duration of this test was two seconds.   However, both impulses occurred during the first second of the recording and both the time-series and wavelet analyses results for that first second are given below.

 

Figure 20: Wavelet Analysis of Hammer Strike to Tail in Channel 1

 

Figure 21: Wavelet Analysis of Hammer Strike to Tail in Channel 2

 

Figure 22: Wavelet Analysis of Hammer Strike to Wing in Channel 3

 

 

Figure 23: Wavelet Analysis of Hammer Strike to Wing in Channel 4

 

 

Figure 24: Wavelet Analysis of Hammer Strike to Tail in Channel 5

 

 

            The wavelet analysis shows that both wires running down the length of the plane picked up both impulses, but the port side wire picked up the signal a full order of magnitude stronger than the acoustic and starboard wires.  The high frequency portions of the impacts can be seen in channels one and two provided you look for them, but the strength of the signal in those channels at all frequencies is significantly less than the strength in channels three and four.  The strength difference in the side wires may be related to the side of the tail the hammer was struck on, but that information could not be validated.

  5.3 Impulse Hammer Strike to Front Landing Gear Support Strut

The final impulse hammer test was performed on the main support strut of the front landing gear as picture in Figure 25.  The full length of this test was one second.  The full second is given below in wavelet decomposition form and time-series representation for all 5 channels.

Figure 25: Depiction of Hammer Strike to Main Landing Gear Support Strut

 

Figure 26: Wavelet Analysis of Hammer Strike to Main Support Strut in Channel 1

 

Figure 27: Wavelet Analysis of Hammer Strike to Main Support Strut in Channel 2

 

Figure 28: Wavelet Analysis of Hammer Strike to Main Support Strut in Channel 3

 

Figure 29: Wavelet Analysis of Hammer Strike to Main Support Strut in Channel 4

 

 

Figure 30: Wavelet Analysis of Hammer Strike to Main Support Strut in Channel 5

 

 

Channel 1, the wire wrapped around the nose of the plane closest to the engine does record the impact, and the strength of the signal is comparable to the strength in channel 3.  Channel four appears to not be connected to an input wire and still registers a signal.  This “signal” is most likely an artifact of the recorder.  The impact is visible in channel 2, but at only half the strength of channel 1.  This is unexpected given the strength of channel 3 and their relative distances to the impulse origin. 

  5.4 Break of Main Landing Gear Support Strut

The final test that directly relates to our analysis was the break of the main support strut of the front landing gear.  In attempting to break the strut, the previous Q group placed a shaker at the tail of the plane to repeatedly increase and decrease the load on the strut.  When this did not result in a complete fracture, the group decided to wrap a towel around the strut and impart impulses to it in the form of sudden jerks while the plane was also rocked from the tail.  This finally resulted in a successful break of the main support strut.  The data we analyzed was the last 30 seconds of the test which included several pulls leading up to the break, and the break itself.  It is impractical to show the wavelet decomposition of all 30 seconds of data, but the full 30 seconds of time-series data for each channel is given below.

 

Figure 31: Time-Series of Main Support Strut Test Channel 1

 

 

Figure 32: Time-Series of Main Support Strut Test Channel 2

 

Figure 33: Time-Series of Main Strut Test Channel 3

 

 

Figure 34: Time-Series of Main Strut Test Channel 4

 

Figure 35: Time-Series of Main Strut Test Channel 5

            Each one of the jerks on the strut is a significant event.  However, the most important aspect of each channel is how it recorded the break.  Given below is a plot of the second containing the break in channels 1 through 4.  The bottom subplot is the entire 30 seconds of data, the middle subplot is the second being analyzed, and the top subplot is the results of the wavelet analysis for that specific second.  The green lines in the bottom subplot mark the second being analyzed with regards to the full length of the signal.

Figure 36: Wavelet Analysis of Main Support Strut Break in Channel 1

 

 

Figure 37: Wavelet Analysis of Main Support Strut Break in Channel 2

 

Figure 38: Wavelet Analysis of Main Support Strut Break in Channel 3

 

Figure 39: Wavelet Analysis of Main Support Strut Break in Channel 4

 

            An opportunity to analyze a break like this is a significant step forward in our understanding of how the wires capture a physical event and how it appears through a wavelet analysis.  We now have a basis by which to judge events on other tapes that may indeed be breaks.  It is particularly significant is that all 4 channels have roughly the same time-frequency pattern.  This data set is especially helpful in that it also includes the pulling on the strut before it breaks.  A few of these pulls are shown below.  The green lines that border the second of data being analyzed show the relation of the pull within the data set.

Figure 40: Jerks Leading to Main Support Failure in Channel 4

 

Figure 41: Jerks Leading to Main Support Strut Failure in Channel 4

 

            These jerks to the strut provide valuable insight into how a mechanical impulse internal to the frame (e.g. a member slowly fracturing) will appear in comparison to a break.  An interesting facet of the jerks is that they are much more band isolated than what is normally expected for an impulse.  By band isolated we mean that the impulse exists over a limited frequency band.  This may be a result of the towel wrapped around the strut, or the short length of the strut.  The second noticeable element of the jerks is the double impulse they all seem to posses.  This is most likely the result of the reflection of the impulse wave through the plane.

  5.5 Conclusions From Wavelet Analysis

            The results gained by analyzing the previous groups data was very helpful.  The data provided examples of signatures that may resemble a member buckling, or an actual break in a support beam.  Along with these examples, the actual placement of the wires provides a great deal of insight into how the wires respond to given inputs.  Even though the wires at the nose of the plane are closer to the impulse in the main strut than the wires running parallel to the plane, the strength of the signal in the parallel wires is much greater than that in the wires on the nose of the plane.  Also, the strength of the signal coming through the wires mounted on the starboard side of the plane during the wing impulse test is significantly stronger than the signal of received by the port side wire.  These facts are counter to each other in stating that the triboelectric effect caused the signal in the wires.  If the movement of the right wing compared to the left wing caused the disparity in the signal strength, then the strength of the signal closest to the impulse in the main support strut should also be the strongest.  This is not true.  There are a few possible reasons for this irregularity. 

            There is a consequence of the Nyquist theorem that may cause the vibration of the break to be missed by the wires in the nose of the plane yet picked up by the wires running down the plane.  The consequence is that for a wire to pick up a vibration at a given frequency, the traveling wave the causes the vibration must complete one full wavelength while it is in contact with the wire.  Two possibilities arise from this consequence.  The first is that the front wire may have missed the lower frequencies of the vibrations because of its perpendicular orientation to the traveling wave.  Only frequencies with wavelengths shorter than the diameter of the wire would not have been damped out.  If this were true the high propagation speeds in the aluminum would mean that any frequency lower than roughly 10000 Hz would have been damped out.  What is more likely is that the shorter length of the wire in the nose caused the damping of the lower frequencies and the subsequent loss of the “jerks”.  By the same consequence of the Nyquist theorem as before, the shorter wires in the nose would cause a damping of any frequency lower than 405 Hz.  This was calculated by noting that the propagation speed is roughly 5100m/s and the wire length over the nose was roughly 2m.  The longer wire (5m) would have been able to record all frequencies above 162 Hz.

            Another possibility results from the sampling rate of the data acquisition system used by the group caused the highest frequency visible to be 8192 Hz.  When the short support rod was broken in the middle it resonated at a frequency too high for the sampling rate of the recorder to capture.  The broken rod imparted its momentum to the shell of the plane.  The larger size of the nose of the plane was not enough to make the resonant frequencies visible to the recorder, but once the vibrations entered the main body of the plane, the caused vibrations at frequencies visible to the recorder.

            It may also be possible that the fact that the bundle wires are able to capture signals that the signal wires around the nose cannot.   This type of effect is not in agreement with the triboelectric effect, and if the bundles do create higher readings, it is due to electromagnetic effects.

            Further testing is required to conclusively prove or disprove any of these theories.

 

Wavelet Homepage                               Table of Contents