A Detailed Examination of Waveforms from Multiple Sensors on a Composite Pressure Vessel (COPV)

Transcription

1 A Detailed Examination of Waveforms from Multiple Sensors on a Composite Pressure Vessel (COPV) By M. A. Hamstad University of Denver, Department of Mechanical and Materials Engineering Denver, CO USA 9/7/212 Granada, Spain AE Meeting; September 212 1

2 Study was initiated to examine if there was relevant information in signal waveforms, not generally provided in automatic AE signal processing software Potential uses of such information: 1. Overcome these listed key facts that complicate accurate location of AE sources in complex composites: i) Velocity varies by propagation direction ii) High attenuation and varying source amplitudes result in threshold-based arrival times being inconsistent iii) Actual initial signal arrival may be down in the electronic noise iv) Intense signal modes/frequencies vary with source type, orientation and depth through thickness. Also, typically a hybrid composite due to metal or polymer liner Accurate location very useful for a weakest link model 9/7/212 Granada, Spain AE Meeting; September 212 2

3 Commercial filament-wound vessel monitored with eight wideband sensors on the cylindrical part of the vessel; interspersed wraps Bandpass 2 khz to 1 MHz; 248 pts at 5 MHz; pretrigger 25 pts; full scale 1 volt pp 12 bit system had low frequency noise (± 3.9 mv) at about 3 khz in waveforms filtered prior to current work; then noise (± 1.3 mv) Threshold ± 17 mv Test information: Carbon fiber, polymer matrix, nominal cylindrical section thicknesses composite 3 mm; aluminum alloy liner 1.3 mm; Cylindrical nominal length 29 mm, nominal outside diameter 125 mm Typ. size used in COPV research due to costs Manufacturer had done hydro-proof/ autofrettage to 63 % of eventual failure prior to hydro-ae cycle to same level AE sensor DW B125T, aperture about 6 mm; sufficiently wideband to allow easy identification of fundamental modes; simultaneous waveforms 9/7/212 Granada, Spain AE Meeting; September 212 3

4 For visualization of sensor locations, the cylindrical part of vessel was unwrapped so multiple paths to hoop sensors could be determined. Grid lines are 1 in. (25.4 mm) apart PLB point PLB point Coordinate origin 9/7/212 Granada, Spain AE Meeting; September 212 4

5 Prior to a pressure cycle to 63% of eventual failure, a PLB (a mechanical fixture) was done near sensor # 3 at a pressure of.69 MPa Amplitude, mv Pencil Break at + position Different known distances and angles of propagation were present # # # # # # # # Time, µs Sequence of sensor signals from top to bottom by apparent arrival of flexural mode 9/7/212 Granada, Spain AE Meeting; September 212 5

6 Due to the complex layup of the winding pattern, the group velocities of the fundamental Lamb modes were expected to vary by direction of propagation from the PLB to the sensors Initial approach selected to allow an estimation of the initialarrival velocities of each fundamental mode as a function of propagation directions and distances Initial arrival time of extensional mode determined from the time at the peak of the first positive half cycle above the background electronic noise. If this peak was questionable, then the arrival time was taken at one half the time between the questionable first positive peak and the subsequent first negative peak A similar approach used for the flexural mode, except its arrival time was determined by the peak of the first half cycle of either polarity that had an amplitude above the extensional mode signals. Again if there was some question, the time was taken at one half the time to the subsequent half cycle peak of opposite polarity 9/7/212 Granada, Spain AE Meeting; September 212 6

7 Key assumptions were made for this approach Assumed, for the propagation distance 37.6 mm from the PLB to sensor # 3, the wave-fronts of the modes in all directions traveled with the same velocities Assumed the Lamb waves were fully developed upon reaching sensor # 3, which was at 8.8 times the total nominal wall thickness. Both modes were clearly observed in this signal. Wave-front radius, 37.6 mm Sensor # 3 Sensor # n Velocity = Δ Distance*/Δ Arrival Time* PLB point *Difference in values for sensor # n minus sensor # 3 9/7/212 Granada, Spain AE Meeting; September 212 7

8 Using the relative (to sensor # 3 signal) propagation distances and relative arrival times, the modal initial-arrival propagation velocities from sensor # 3 to the other sensors were calculated Sensor number; angle of propagation Ext. velocity, mm/µs Ext. % difference vs. ave. w/o # 4 Flex. velocity, mm/µs 3; ; ; ; ; ; ; ; Ave. w/o # Flex. % difference vs. ave. w/o # 4 Propagation angles vary from 1 to 85 degrees from the hoop direction ( deg.) Large differences! 9/7/212 Granada, Spain AE Meeting; September 212 8

9 Since there was a wide variety of propagation angles for the nearest path from the PLB to the sensors, some questions arose relative to the much higher velocity results for the path to sensor # 4 Sensor # 4 PLB position O O Propagation from PLB to sensor # 4 is nearly in the hoop direction; off by about 1 degrees Question: Is this the reason? Y axial direction Propagation direction from PLB to # 4 is nearly in the hoop direction X hoop direction 9/7/212 Granada, Spain AE Meeting; September 212 9

10 Observations relative to the velocity of propagation of the first arrivals at sensor # 4 of the fundamental modes excited by the PLB signal COPV design places about twice as many fibers aligned in the hoop direction. Hence, as an approximation, the inplane stiffness is about twice as stiff in that direction compared to the axial direction. Thus, it is expected that the velocity of the fundamental longitudinal mode will be considerably higher in that direction than in other directions. The flexural mode depends strongly on the flexural modulus, which is influenced by the low modulus of the polymer. Thus, it is expected that the increase in the flexural mode velocity in that direction will not be large compared to the velocity in alternate directions. Based on these observations, an alternate analysis of the apparent initial arrival of the flexural mode was undertaken 9/7/212 Granada, Spain AE Meeting; September 212 1

11 Amplitude, mv At the selected arrival times (solid red arrows) for the initial flexural velocities, the signals at distances less than and more than that of # 4 show a different character Dist mm Sensor # 5 Dist mm Sensor # 4 Dist mm Sensor # Time, µs In the region of the flexural mode arrival time for # 4 signal compared to the others: A sharp pulse of higher frequency, without the following decreasing frequency in the other two sensor signals Later in signal # 4, at the red dashed arrow, the arrival shows a character much more similar to that from the other sensor signals at the flexural mode arrival 9/7/212 Granada, Spain AE Meeting; September

12 The previous slide observations indicated potentially a later arrival for the flexural mode initial arrival at sensor # 4 Hence, using the arrival time at the dashed arrow, the velocity of propagation to sensor # 4 was recalculated The resulting velocity was determined to be 1.77 mm/µs, which is only some 9 % faster than the average velocity for all the other sensor signals as shown in the previous table This result provided a strong indication the originally assumed flexural arrival at # 4 was due to another aspect 9/7/212 Granada, Spain AE Meeting; September

13 Noting that sensor # 4 is one-half way around the vessel from # 3 on the same circumferential circle an alternate path through the fluid (water) used to pressurize the vessel was considered The total distance of the water path was about 127 mm (composite, aluminum, water) propagation. By use of bulk velocity values and distances for water (118.4 mm), aluminum (total 2.6 mm) and composite (6 mm) a water path transit time was calculated of about 82 µs from the PLB to sensor # 4. To estimate where this transit time appeared in the signal from sensor # 4, the average velocity for the extensional wave traveling to sensor # 3 was used to calculate an approximate propagation time from the PLB to # 3. The approx. time for this distance (37.6 mm) was 7.1 µs. This located the time of the PLB at about 41 µs in the recorded signals 9/7/212 Granada, Spain AE Meeting; September

14 Amplitude, mv A water path provides an explanation for the early arrival in the signal from sensor # # 3 # 4 Arrival of water path signal at µs PLB time at approx. 41 µs Time, µs Experimental difference from the time of the PLB to the water path arrival = 84.8 µs Calculated propagation time for water path of the vessel was about 82 µs Conclusion: the water path accounts for the early arrival originally mistaken for the initial flexural wave arrival 9/7/212 Granada, Spain AE Meeting; September

15 CWDs of the signals for the three distances to sensors # s 5, 4 and 7 show a higher frequency content of the early arrival water path for sensor # 4; such a path is expected to better preserve high frequencies Higher frequencies water path signal; CWD peak at 25 khz is at µs vs µs on waveform # 5 at mm # 4 at mm # 7 at mm Note the broken nature of the lower frequency arrival at # 4. As will be shown, later the second path to sensor # 4 results in a initial flexural mode arrival only about 14 µs after the first path arrival of this mode 9/7/212 Granada, Spain AE Meeting; September

16 Due to the much larger velocity of the extensional wave arrival for # 4, the assumption of the wave-front being a uniform circle when it reached sensor # 3 was not correct. Hence, an iterative procedure was used to recalculate an estimate of the initial arrival extensional velocity for the path from the # 3 arrival to sensor # 4. Details are in the paper. The process converged at 7.52 mm/µs ( 1 % less than the original value) A similar procedure was used for the flexural mode initial arrival velocity; converged at 1.73 mm/µs ( 2 % less than the original value) 9/7/212 Granada, Spain AE Meeting; September

17 Distance, mm Distance, mm Observing, except for the path from the PLB to sensor # 4 (for the ext. mode), the calculated velocities of the initial arrivals of each mode were similar, distance versus arrival time correlations were done 3 25 Extensional initial arrival; w/o # 4 y = 5.231x R 2 = y = 1.631x R 2 =.978 Flexural initial arrival Arrival time, µs 9 1 Calculated ave. velocity was 5.33 mm/µs; 2 % higher than the slope velocity of 5.23 mm/µs Arrival time, µs Calculated ave. velocity was 1.63 mm/µs; same as the slope velocity of 1.63 mm/µs Granada, Spain AE Meeting; September 212 9/7/212 17

18 Amplitude, mv The 2 nd path waves may interact with 1 st path waves, particularly for the extensional mode. Estimated by the time from the PLB plus the time for propagation at the calculated velocities (average w/o # 4, or if path close to hoop direction, # 4 velocities). First four sensors hit; top to bottom E2 Second path initial arrival extensional mode F2 Second path initial arrival flexural mode E2 E2 E2 E Time, µs 9/7/212 Granada, Spain AE Meeting; September F2 Second shortest path was always closer to the hoop direction F2 F2 F2 # 3 # 2 # 1 # 6

19 Amplitude, mv The 2 nd path waves may interact with 1 st path signals, particularly for the extensional mode; estimated by the time from the PLB plus the time for propagation at the calculated velocities (average w/o # 4, or if path close to hoop direction # 4 velocities). Second four sensors hit; top to bottom E1 and E2 1 st and 2 nd path arrivals of modes F1 and F2 1 st and 2 nd path arrivals of mode E1 E2 E2 Water path E2 E2 F1 Close arrivals F1 and F2 F Time, µs 9/7/212 Granada, Spain AE Meeting; September F1 Second shortest path was always closer to the hoop direction F2 F2 Close arrivals F1 and F2 # 5 # 4 # 7 F2 # 8

20 Frequency, khz Frequencies of CWD peak intensity of extensional mode and flexural mode varied as a function of the propagation distance from the PLB Extensional mode Average frequencies: Ext. mode 337 khz Flex. mode 92 khz Peak CWD magnitudes at these frequencies used to determine arrival times Flexural mode Signals truncated to isolate the extensional mode and in one case for the flexural mode to obtain first path signal results Distance, mm Granada, Spain AE Meeting; September 212 9/7/212 2

21 Propagation distance, mm The initial arrival times of the extensional mode fit a straight line much better than for the CWD arrival time at the average peak frequency (results without using sensor # 4 data) 35 y = 4.38x Linear (CWD at 337 khz) R 2 =.951 Fit comparison: CWD at 337 khz, R 2 = Initial arrival slope, R 2 = Time, µs Granada, Spain AE Meeting; September 212 9/7/212 21

22 Propagation distance, mm The CWD arrival times at the average peak frequency of the flexural mode fit a straight line better than the initial arrival time 3 25 y = 1.468x R 2 =.997 Linear CWD at 92 khz Fit comparison: 2 15 Sensor # 4 CWD at 92 khz, R 2 =.997; w/o # 4, due to its broken CWD Time, µs Initial arrival slope, R 2 =.981; with # 4, no reason to exclude Granada, Spain AE Meeting; September 212 9/7/212 22

23 Using the superior S/N approach, CWD-based arrival times with the group velocity (1.468 mm/µs) at the average frequency of the flexural mode region, as a check, the location of the PLB was calculated with planar location software Sensors used X, mm Y, mm Radius error, mm # s 3, 2, 1, # s 1, 2, 7, Actual location Consistent results due to close fit of the data in the plot of propagation distance versus CWD-determined arrival time at the peak magnitude of the average frequency of 92 khz

24 Magnitude, mv x 1 3 Examination of CWD magnitudes versus time at the ave. frequency of 92 khz, indicates some issues for automatic implementation # 3 # 2 # 1 # 6 # 5 # 4 # 7 # Time, µs Red arrows show the peak magnitude where the arrival time was obtained in cases where there was no clear initial large peak of the CWD magnitude

25 Details of issues for automatic implementation are demonstrated considering signals # s 4, 7 and 8 farthest sensors mv mv x 1 3 mv mv x 1 3 mv mv x # 4 Signal Water path # 7 Signal # 7 CWD at 92 khz # 8 Signal # 8 CWD at 92 khz Time, µs # 4 CWD at 92 khz Multiple peak cases showed the features that created issues as to the selection of the correct arrival time : water path (# 4) close second path arrivals (# s 4, 8) later time of second path arrival also results in a need to truncate to eliminate a late absolute maximum (# 7)

26 Using the superior S/N approach with CWD-based arrival times at an average frequency of the flexural mode region, the location of a real AE event (with signal character similar to that of the PLB) was calculated with planar location software Sensors used # s 3, 6, 5, 2, 1 # s 3, 6, 5, 2, 1, 7 X, mm Y, mm Iterated velocity, mm/µs Without use of broken CWD results (# s 4 & 8), the average frequency of the flexural mode peak intensity was 85 khz, and since the velocity was not known, the location calculation velocity was iterated to minimize the uncertainty in the location result Event # 314 at 62 % of the eventual burst pressure # 4 # 8

27 Amplitude, mv Compared to the broken CWD cases # s 4 and 8, the CWD peaks for arrival times would be clearly defined by the first large increase in the peak values of the magnitude at 85 khz # 3 # 6 # 5 # 7 # 2 # 1 # 4 Water # 8 Signals event # Time, µs Magnitude, mv x # 3 # 6 # 5 # 7 # 2 # 1 # 4 # 8 CWD magnitude at 85 khz Broken Broken Time, µs The water path for # 4 was verified by calculations as before and this as well as the close second arrival of the flexural mode at # s 4 and 8 and lead to the broken CWD result and late time of the absolute maximums.

28 Several important conclusions can be drawn from this research: The group velocities (based on the initial arrival times and the times of the peak magnitude of CWD at the average frequency ) of the flexural mode were essentially the same for signals propagating in different directions in the COPV The group velocity of the extensional mode increased sharply for propagation near the hoop direction of the COPV Using the average frequency of the CWD for the flexural mode region, provides a means to obtain reliable arrival times for source location; need to ignore signals with broken CWD results. This approach provides arrival times from the signal region with the best S/N ratio for sources that result in signals with a dominate amplitude of the flexural region An automatic method to obtain such CWD-based arrival times could be implemented; requires more research examining real AE signals With small composite pressure vessels, a potential water path cannot be ignored 9/7/212 Granada, Spain AE Meeting; September

29 Thanks for your attention Questions?? 9/7/212 Granada, Spain AE Meeting; September