DECI NEWSLETTERS AND REPORTS

THE DECI REPORT - September 1998

H.L. Dunegan

MEASURING CRACK DEPTH OF GROWING CRACKS BY ACOUSTIC EMISSION TECHNIQUES

INTRODUCTION

Prosser and Gorman reported (ref. 1), that extensional waves could be generated in a plate by coupling a small thin aluminum plate to the surface of the plate being measured, and breaking a pencil lead on the edge of the attached plate. This idea was further expanded in a report by the author (ref. 2) which showed that a thick Compact Tension fracture toughness specimen (CT) coupled to a bar with petroleum jelly, cloned its response to out-of-plane (OOP) and in-plane(IP) pencil lead breaks to the steel bar on which it was coupled (figure 1). Pencil lead breaks were made at different depths in the CT specimen to represent where along the 19mm thickness of the specimen the crack front is propagating. These experiments showed that the extensional wave generated is very weak, but a strong high frequency shear wave is generated by the in-plane(IP) source, and it is the ratio of the amplitude of this high frequency shear wave to the amplitude of the low frequency flexure wave that was utilized for correlating the data. An SE9125-M dual mode transducer placed at approximately 24 inches from the specimen detected the plate waves from the pencil lead breaks and the high frequency HF to low frequency LF ratio of the peak amplitudes were plotted as a function of percent depth. It was discovered in these experiments that the ratios of HF/LF for the combined CT specimen and steel bar agreed well with pencil lead breaks made at different depths in the bar alone. Figure 7 in the above referenced report (ref. 2) shows these results. Reference 2 can be downloaded from our web page.

There has been some controversy in the AE community concerning how closely pencil lead breaks simulate actual crack growth and whether on not HF/LF ratios as implemented by the AESMART 2000 can be used to estimate crack depth. The experiment in this report was designed to provide an answer to both questions.

EXPERIMENTAL PROCEDURE

A CT specimen similar to the one shown in figure 1 was hydrogen charged and bolt loaded via the threaded hole shown in the specimen in figure 1. A large aluminum plate 600mm X 700mm X 5mm thick was utilized in the experiment to eliminate edge reflections. The bolt was tightened to put stress at the notch tip, and the specimen was coupled to the surface of the aluminum plate near one edge with petroleum jelly. An SE9125-M transducer was placed at 10 inches from the edge of the plate in a line perpendicular to the notch in the specimen (similar to figure 1). An AESMART 2000 instrument was used to gather the data. After completion of the test the AESMART transferred the data to an Excel spread sheet for analysis and graphical presentation.

Hydrogen embrittlement cracking occurs on a microscopic level. Therefore the crack front in this 19mm thick specimen will not advance uniformly through the thickness, it will advance in small distinct steps all along the crack front as stresses are redistributed due to the microscopic crack advance. An AE signal due to a microcrack propagating at the exact center of the specimen will provide the highest HF/LF ratio, while a microcrack advancing near either the top or bottom portion of the specimen will provide a ratio of approximately 1. Therefore for a given extension of the crack all along the crack front, the lower ratios will predominate.

EXPERIMENTAL RESULTS

Figure 2 shows the summation of counts from the high frequency channel as a function of time due to the advance of the crack front. This graphic was created in Excel following the test. It was also observed in real time on the AESMART screen. The decreasing slope of the counts curve indicated that a crack arrest should occur at approximately 6000 seconds. Arrest did not occur and cracking continued and increased in activity until a fairly large crack jump occurred that unloaded the specimen enough to prevent further cracking. The test was terminated at this time after approximately 2mm of crack growth was visible at the tip of the notch on both sides of the specimen. If one assumes that the total amount of crack growth is rectangular in shape, the total amount of crack area generated during the test was (2 mm X 19mm)= 38 square mm.

Figure 3 shows the counts per event plotted as a function of ratio for the event. Three events gave an infinite ratio and were defaulted to a value of 30. Note for this data that only one event gave a ratio of less than one. Note the thinning of the distribution at the higher ratios. This is consistent with the fact that the higher ratios only occur for crack growth at and near the center of the specimen depth. The counts per event are grouped rather tightly, indicating that the total area generated by each increment of crack growth due to the hydrogen cracking was approximately the same size.

Figure 4 is a scatter plot of the HF/LF ratio from the AESMART for each event as a function of time. Several interesting observations can be made of this data. First note that all of the ratios are greater than one with the exception of one event. Second note that there is a higher density of low ratios than high due to the fact that as the crack approaches either side the ratio approaches one, and only is high when crack growth occurs near the center of the specimen. The lack of data near 6000 seconds corresponds to the flattening of the counts curve in figure 1. The other areas without data correspond to the other flat regions of the curve in figure 1. Apparently the specimen is dormant until enough hydrogen diffuses to the crack vicinity to start cracking. The wide range of ratios present once cracking starts indicates that possibly the whole crack front advances during this active period and then waits for more hydrogen to diffuse to the area. Note beginning at 10,000 seconds the activity picks up and a higher proportion of high ratios are present indicating a lot of cracking near the center of the specimen. Again a dead period is present and the specimen becomes very active prior to the large crack jump that terminated the test. Again a large percentage of the ratios were high indicating crack propagation near the center of the specimen.

Figure 5 is a plot of the cumulative average ratio of all of the events as they occurred. Note that the average fluctuates between 5 and 7 for the first 625 events and then begins to climb for the next 100 events and finally begins to decrease just prior to the large crack jump in the specimen which terminated the test. The decreasing average ratio for the last 80 events indicates that during the final period leading up to the large crack jump, that crack growth was progressing from the center toward the edge or edges of the specimen. This is confirmed by the last group of data in figure 4 which shows a higher density of medium and lower ratios. The high amplitude event in figure 3 which resulted in 13,000 counts was the signal produced from the final crack jump. It produced a ratio of approximately 2 which indicates that the final crack jump occurred near the top or bottom surface of the specimen.

Figure 5 also shows the cumulative average ratio for 800 events produced by rubbing the aluminum plate with 400 grit size sandpaper. The average ratio of this OOP frictional noise is seen to be less than one.

DISCUSSION OF RESULTS

Approximately 38 square millimeters of crack area was created by 800 events. If one assumes that each event was produced by the same amount of new area created, the area corresponding to each event would be 0.048 square millimeters. This area is comparable to the contact area of a 0.3mm pencil lead break used for calibration purposes. With the sensitivity used in the test it was impossible to create an infinite ratio by breaking a 0.3mm pencil lead break at mid thickness in the CT specimen. About the highest ratio achievable was 16. A few infinite ratios were observed due to the hydrogen cracking, and ratios of above 20 were not uncommon. This indicates that on the average the area created by the hydrogen cracking was probably smaller than the contact area of the pencil lead. In addition the crack is created by a dipole force as opposed to a monopole force generated by the pencil lead. This seems to produce more favorable results as shown by the wider range of ratios than those produced by the pencil lead break.

During the test the ratio filter of the AESMART which was set at 1 was turned on for a short period. During this period 400 grit sandpaper could be rubbed on the surface of the aluminum plate without affecting the data. The extraneous noise produced by this friction was rejected by the AESMART. The reason for this rejection is obvious from figure 5 which shows that the ratio produced by the sandpaper was well below 1. During the majority of the test the filter was turned off in order to see the total distribution of ratios. The data shows that only one event out of 800 had a ratio of less than one.

CONCLUSIONS

These results show that the acoustic emission signals produced by crack growth due to hydrogen embrittlement cracking agree well with the results produced by IP 0.3mm pencil lead breaks. It is further shown that the HF/LF ratios of the acoustic emission signals as recorded by the SE9125-M transducer and AESMART 2000 instrument can be correlated by inference with the depth in the specimen where the crack growth is occurring. Future experiments will involve interruption of the test after a finite crack growth has occurred, heat tinting of the specimen, and fracture of the specimen in a mechanical test machine in order to get a direct correlation between HF/LF ratios and the amount and shape of the crack surface created.

REFERENCES

Prosser, W.H.and Gorman, M.R.,îAccurate Simulation of Acoustic Emission Sources in Composite Plates,î Proceedings of the 1994 ASNT Spring Conference, New Orleans, March 1994,pp. 152-154.

Dunegan H.L. ìUse of Plate Wave Analysis in Acoustic Emission Testing to Detect and Measure Crack Growth in Noisy Environments.î Proceedings of the Structural Materials Technology-An NDT Conference San Diego California, February 20-23 1996.