TEST RESULTS
The preliminary compliant panel tests in the RC-19 wind tunnel used only the strain gage and the single vibrometer point to monitor and record panel response. The lack of spatial resolution this created, combined with the varying temperature difference between the frame and panel and little-understood dynamic pressure across the face of the plate, proved overwhelming when plate mode identification under these conditions was first attempted. The spatial resolution achievable with DIC, although never before attempted under these conditions, ultimately proved invaluable to understanding the complex plate response.
As mentioned previously, the most difficult hurdle in the implementation of 3D high-speed DIC measurement of a compliant panel subjected to such an extreme environment is the combination of resolution, sample rate, and length of time record required. Traditional DIC only requires perhaps a few hundred images for processing fracture tests which is achievable with low-speed cameras. Transient dynamic analysis requires high-speed imaging, but is typically limited to several thousand images over the span of milliseconds. Until recently, the sheer volume of images required for the current application was simply insurmountable. As an example a 1-second snapshot, 2-second snapshot, and full 20.8 second time record of panel displacement measured with DIC for a steady-state run condition in the RC-19 tunnel is presented in figure 8. At 5000 frames per second and a resolution of 640 x 352 pixels, the 1-second snapshot required the storage, download, conversion, and analysis of 5000 image files. This volume of images is already at or even more likely beyond what is typically attempted with DIC analysis software. As can be seen in figure 8(a) the lack of sufficient averages, needed due to the stochastic nature of the panel response, results in little-to-no useful information obtained from the power spectral density (PSD).
Until just recently the limit of a high-end, high-speed camera memory was about 4 GBs of storage capacity. Therefore, the length of the time record snapshot can be increased to 2-seconds and the result for this approximate maximum time record for a 4 GB camera is shown in figure 8(b). There is a slight improvement in the resolution of panel response; however, the amount of information that can be gleaned from these results is still marginal. Major peaks can be seen and are assumed to be the first 4 modes of the panel, but the room for error in mode identification is still well above acceptable levels. Figure 8(c) shows the full 20.8 second time record achieved with 32 GBs of memory per camera available with the SA5. Distinctive peaks are now visible and lower amplitude response not seen before is now able to be evaluated.
Fig. 8 DIC displacement PSD from (a) 1 second 5000 frame time record (b) 2 second 10,000 frame time record and (c) 20.8 second 101,661 frame time record
Armed with an appropriately averaged PSD of panel displacement response at several points across the plate center region, the question of DIC data reliability must be addressed. The response from the strain gage along the plate edge for a 60-second time record can be seen in figure 9. Unfortunately, the DIC speckle pattern over the strain gage was below the noise floor when attempts were made to evaluate it for either displacement or strain. An attempt was then made to evaluate strain at panel point 4 and the results are shown in figure 10. Even at this point where displacement at each facet is able to be evaluated and remains well above the noise floor, the difference in displacement is below the noise floor once again. Strictly for comparison of general trends to verify the measured response with DIC, the displacement at point 4 is shown in figure 11. The trend of the data between the strain gage in figure 9 and DIC displacement in figure 11 is clearly very similar despite being almost 2 cm apart across a high gradient region. The strong similarity between the two readings supports the validity of the DIC displacement results.
The laser vibrometer measurement at the retroreflective marker and the DIC displacement from point 10 can be seen in figure 12. The vibrometer velocity has been integrated and is presented as displacement. Again bearing in mind these data are at slightly different points on the panel, the similarity in general trends and even amplitudes further increases the confidence in the DIC measurements. The amplitude of the largest peak in the response of the DIC data is only slightly higher than that of the vibrometer data. This slight increase would be expected since the DIC measurement point is about 1 cm further inward from the panel edge where displacement goes to zero. As with previous attempts at direct comparison, the DIC analysis software was unable to track the photogrammetry target with the vibrometer laser reflection making a direct comparison unsuccessful. It is believed in the future the laser light could be filtered out of the DIC image to overcome the interference.
Fig. 9 Uni-axial strain gage measurement PSD at center of plate, long edge
Fig. 10 DIC strain measurement PSD at speckle point 4
Fig. 11 DIC displacement measurement PSD at speckle point 4
Fig. 12 PSD of integrated vibrometer velocity (-) and DIC displacement measurement at point 10 (-)
With confidence that the DIC measurements have at least been indirectly supported by both the laser vibrometer and strain gage measurements, the next task was to actually interpret the panel response. With all 21 points of displacement response data across the center region of the panel successfully evaluated for all test runs, adequate and consistent spatial resolution was achieved. Since at the time of this publication the varying pressure load on the face of the plate was not yet known, modal identification of the panel while under the various wind tunnel conditions had to be attempted with only the panel response. One means to shed light on the spatial distribution of the various peaks of the DIC displacement PSDs is to systematically march along one of the line of points on the panel and note the amplitude of the response at that point with respect to the points surrounding it. Beginning with a few points along the plate centerline in the x direction in figure 13 there are several characteristics to note. First, the amplitude of the largest peak at 265 Hz increases from point 8, closest to the short edge on the left, to the plate center at point 11. The response then decreases again at point 14 nearest the short edge on the right side. This general trend of increasing and decreasing response towards a peak at the center, coupled with the panel modal results before installation in the RC-19 test section in figure 6, would indicate that this is the panel first mode. The increase of about 20 Hz between the initial modal results and the RC-19 test results can most likely be attributed to the aforementioned 6.5 oC temperature difference between the panel and frame. This negative temperature delta would effectively pre-stress the panel and increase the frequencies. Also of note is the response peak at approximately 360 Hz. The response at the panel center point drops out almost completely at this frequency indicating that it is likely located at a node line. Furthermore, the points on either side of the node line have nearly identical amplitude. Again, looking at the initial modal results in figure 6 and the center node line behavior represented in the displacement PSD, this is clearly the panel second mode. The third major peak at about 515 Hz in figure 13 would then be the third panel mode as each measurement point is predictably at each of the three regions of highest displacement for this mode. The fourth panel mode should not be seen in this figure as all points fall on the node line. At just over 700 Hz, a mode with a node line along the center can once again be seen indicating this is likely the fifth panel mode.
Fig. 13 DIC displacement measurements PSD along the panel center line, x direction
Fig. 14 DIC displacement measurement PSD along panel centerline, y direction
Figure 14, representing the three points along the panel centerline in the y-direction, further supports the conclusions drawn from the previous figure. The first peak at 265 Hz once again has peak amplitude at the center and slightly lower and equal amplitude above and below. The peak at 360 Hz is essential gone representing the node line of the second mode. The peak at 515 Hz is once again seen at nearly equal amplitude at all three points with the highest at the center due to the center anti-node of the third mode. At 600 Hz the fourth mode can now be clearly seen and the expected node line at the center point where the amplitude disappears. The fifth mode at 715 Hz is now no longer visible due to the vertical node line at the center. To add further spatial information, in figure 15 are three points along the bottom edge in the x direction. The center node line is once again seen in the second (360 Hz) and fifth (715 Hz) modes as the response amplitude greatly decreases further supporting the conclusions drawn from the previous two figures. Nearly all the points along the centerline in the x direction are shown in figure 16 to add yet more support to the mode identification with even greater spatial resolution. Amplitudes increase and then decrease as you move along the length of the panel as expected and mode four (600 Hz) is predictably not present at every point.
The importance of the spatial resolution generated by DIC for panel mode identification while subjected to the RC-19 running conditions can also be noted by the various smaller peaks around 100 Hz, 450 Hz, and 850 Hz. Any one of these peaks could have potentially been mistaken for panel mode response due to the fluctuating pressures from the turbulent flow when in fact they are largely artifacts of the chaotic environment created by a continuous flow Mach 2 wind tunnel. Understandably RC-19 was not originally designed with vibration testing in mind. For instance the peak just below 100 Hz is the tunnels first bending mode as seen with the array of accelerometers attached to the outer tunnel walls. The sharp peak at about 120 Hz is actually flicker in the LCD lights from the AC power source. Attempts will be made in future tests to eliminate this noise through the use of battery powered LEDs. The peak at 850 Hz is actually a mode in the apparatus used to suspend the two cameras above the test section as seen in the acceleration data from a tri-axial accelerometer mounted to the camera crossbar.
Fig. 15 DIC displacement measurement PSD along panel bottom edge, x direction
Fig. 16 DIC displacement measurement PSD along panel center line, x direction
Finally, with confidence in the identification of the plate modes while under fluctuation pressure loads from the Mach 2 flow of the wind tunnel, some insight can now be gained into the effect of a shock impinging on the plate with the use of the variable-angle shock generating wedge. Looking at figure 17 the blue line is the baseline condition with the wedge at 0 degrees. The first mode peak at 265 Hz is relatively sharp and could be considered generally linear. The peak displacement for the 0 degree case is 0.29 mm which is just under half a plate thickness. The green line is the plate response when the shock generating wedge is raised to 10 degrees. Using Schlieren shock visualization techniques in preliminary tests it was seen that this shock wedge angle produced a shock impingement at about the panel quarter point in the x-direction. The most notable conclusion from these results is the significant change in panel response induced by the shock impingement. Consider the first mode response in particular. The displacement response was linear, or nearly so for the 0 degree case, while at the 10 degree shock generator angle the first mode peak has both increased and broadened in frequency. This response is the hallmark of acoustic fatigue, and is associated with a transition to a hard-spring Duffing-type nonlinear response. Furthermore, the peak panel displacement for the 10 degree case increased from 0.29 mm to 0.63 mm, which is nearly a plate thickness. A rule of thumb employed to determine nonlinearity, is displacement response equal to or above one plate thickness. These initial experimental results, intended to determine the validity of DIC for these test conditions, are encouraging. This measurement technique will be very useful in studying shock boundary layer interaction (SBLI) on a compliant panel investigation for future RC-19 experiments.
Fig. 17 DIC displacement measurement PSD at panel center (-) 0 degree shock wedge angle and (-) 10 degree wedge angle
CONCLUSIONS
Extreme environment experimentation produces many challenges in the measurement of representative thin-gauge aircraft panel dynamic response. Validating computational response prediction tools is difficult when only one or two discrete measurement points are available. The expected nonlinear structural response adds complexity to the problem, further necessitating full-field measurement techniques. Abundant and reliable information describing panel response due to shock boundary layer interaction will be required to design durable yet lightweight hypersonic aircraft structure. This information could be realized given robust full-field measurement techniques. DIC offers intriguing measurement possibilities assuming the numerous technical challenges can be overcome.
Determining which DIC specific parameters best suit the current application is critical. Tradeoffs between resolution, frame rate, and length of required time record must be considered. Various measurement "targets" have both advantages and disadvantages for different conditions. For the current study, the time record length and measurement location preparation proved to be the most critical for the refinement of random displacement response and noise floor considerations respectively. Both were successfully achieved for simultaneous displacement measurement of 21 points on a compliant metallic panel specimen, which replaced a portion of a supersonic wind tunnel test section wall during continuous Mach 2 flow operation. DIC trends compared well to both discrete vibrometry and strain gage measurements. Spatial resolution was achieved to a degree that the identification of panel vibratory modes was possible at both linear and nonlinear response levels. With the capability to identify panel modes under various conditions, a shock boundary layer interaction investigation can successfully proceed. However, to date DIC strain measurement at or near the panel root was not successful for current test conditions, and further noise floor refinement will be required.
