Sizing flaws by ultrasonics

TWI Bulletin, March/April 1993

George Georgiou graduated from Imperial College in Mathematics in 1972, and stayed on a further year as a research student studying theoretical fluid dynamics. He was a full-time mathematics lecturer at Tottenham College of Technology until 1983, gaining his PhD in 1982. Since 1990 he has worked at TWI on a variety of NDT related problems. In particular he is working on NDT of plastics and adhesives and is currently involved with BSI and CEN committees which are drafting standards for ultrasonic inspection of welds.

Kim Hayward joined TWI as a trainee in 1977 and studied part-time for an ONC in Mechanical Engineering whilst working in the NDT Research Department. He gained his first CSWIP in ultrasonics (plate and pipe) in 1979 and currently holds a PCN Level 2 in ultrasonics. As a Senior Research Technician he is a leading operator of P-scan and Zipscan equipment, presently involved in NDT procedure development.

In ultrasonic non-destructive testing there are a number of available techniques for measuring flaw lengths. Each has particular advantages and disadvantages. George Georgiou and Kim Hayward examine the importance of selecting the appropriate technique and the implications for CEN standards.

There are also other methods, these are called fixed amplitude techniques and include '50%DAC' and 'DAC-12dB'. The differences between all these techniques can influence the measured flaw length. In turn, this can have a significant effect on the accept/reject criteria. In many standards the flaw length parameter is regarded as being more relevant than the through-thickness height which is often not required to be measured anyway.

The advent of CEN (Comité Européen de Normalization) standards makes this a highly topical issue, as different European countries favour different techniques.

This assesses various techniques and makes comparisons with radiographic results. Some interesting correlations are concluded between flaw length, using the different techniques, and the corresponding peak echo amplitude envelopes of the A-scans.

The data set

Thirty-seven common flaws were recorded by both tests - the specimen thicknesses ranged between 10 and 40mm.

Ultrasonic terminology for flaw shape follows BS 3923, ^[1] Table 31. The methods used to characterise the flaws also follow the guidelines of BS 3923, but with some modifcations. It is sufficient here to classify the flaws briefly (see Table 1).

Table 1 Classification of flaw shapes

Ultrasonic length sizing methods

'6dB drop' method from the maximum

This method involves finding the peak A-scan echo amplitude from the flaw and adjusting the gain until the echo amplitude reaches a convenient level on the screen, such as 80% full screen height. The probe is then moved, while at the same range, until the echo amplitude has dropped by a half to 40% full screen height ( i.e. by 6dB). This gives a length measurement L ₁ as illustrated ( Fig.1).

'6dB drop' method from the sides

This follows the same principle as above, except that the 6dB drop is performed on the last peak from each end of the flaw that appears before the signal drops off dramatically. This gives a length measurement of L ₂ as illustrated ( Fig.2).

50%DAC

If a distance amplitude correction (DAC) curve is used, the length of the flaw can be measured from where the A-scan signal falls to 50% of the reference DAC curve ( i.e. DAC-6dB), giving a length measurement of L ₃ as illustrated ( Fig.3).

DAC-12dB

This follows the same principle as above. Thus, using the DAC curve, the length in this case is measured from where the A-scan signal falls to 12dB below the curve ( i.e. 25%DAC), giving a length measurement of L ₄ as illustrated ( Fig.4). A real case example of the variation of length values obtained by these four methods is illustrated ( Fig.5). These are peak echo amplitude envelopes of A-scans at the same range ('echo views') produced by the P-scan system. The part of the echo view showing above the horizontal white line (display level) corresponds to the length that is recorded with each method. The middle figure on the left hand side is the gain level of the line, where 0dB = 100%DAC. Here the maximum A-scan echo amplitude is 17dB.

Fig. 5. P-scan echo view showing lengths obtained by:
a) 6dB max;
b) 6dB sides;
c) 50%DAC;
d) DAC-12dB

Results

Each of the techniques described was applied to the 37 flaws common to radiography and ultrasonics. The results are given here in Table 2. The 'Defect No.' column in Table 2 ( e.g. 70/1) identifes first the specimen and second the particular flaw recorded. The second column identifies the flaw shape according to Table 1. The remaining columns represent the flaw lengths according to the various sizing methods. Figure 5 illustrates the particular case for flaw number 19 ( i.e. 60/2 Th). A graphical representation of the results in Table 2 has also been produced, and the case for the multiple type flaw (M) is included here as an example ( Fig.6).

Table 2 Flaw lengths, mm

In the analysis that follows radiography was used as a benchmark.

Mean and standard deviation

These have been calculated from the values in Table 2. This is quite revealing in itself, although it is recognised that perhaps a larger data set is required to draw more meaningful conclusions based on just these two parameters. Nevertheless, it can be seen that in general the 6dB drop from the maximum technique ('6dB max') is likely to undersize flaws compared with radiography. This also appears to be true for the 50%DAC technique, but less so. The 6dB drop from the sides ('6dB sides') and the DAC-12dB techniques are likely to be closer to radiography, with DAC-12dB tending to oversize slightly.

The standard deviation indicates the spread of the results from their respective mean values. However, a fuller analysis is required as some individual cases reveal precisely the opposite. For example, the last two points on each of the graphs ( Fig.6) illustrate this very well. The ultrasonic lengths are in fair agreement with each other but all disagree with radiography.

Deviations from radiographic results

Another small exercise was carried out to see how many times the various ultrasonic length measurements were within a specified deviation ( e.g. 10 and 15mm) from the radiographic length measurements. The results are given in Table 3 and this tends to support the tendency concluded from the results of Table 2. Naturally, the larger one chooses the specified deviation the more likely the agreement between techniques.

Table 3 Number of times lengths are within a certain range to radiography

Analysis of echo views

Each of the flaws defined in Table 2 had associated with it a unique echo view. The echo view for flaw number 19 ( i.e. 60/2Th) is illustrated ( Fig.5). A closer study of all echo views was made in order to see whether there was any correlation between the echo view shapes and the results given in Table 2. In fact, from the 37 unique echo views studied, it emerged that they could be mapped into five broad categories ( Fig.7). The unique echo view ( Fig.5), would fall under the category of 'peak with long decay'. These five categories also appeared in general to provide valid and consistent explanations of the results in Table 2. These are summarised:

• peak with long decay
  6dB sides and DAC-12dB similar to radiography
  6dB max undersized in all cases compared with radiography
  50%DAC undersized half of the time, otherwise like radiography
  
  
• peak with short decay
  ultrasonic measurements undersized compared with radiography
  ultrasonic measurements similar to each other. Note that the case 56/1 Al ( Fig.6) fell in this category.
  
  
• single peak, high amplitude
  with exception of DAC-12dB the rest were similar to radiography
  DAC-12dB oversized in some cases compared with radiography, this is a classic case where 6dB sides, 6dB max and 50%DAC could be considered equivalent techniques.
  
  
• single peak, low amplitude, fast decay
  ultrasonic measurements undersized compared with radiography
  ultrasonic measurements similar to each other
  50%DAC and DAC-12dB can give zero lengths in some cases (note that the case 52/3 M ( Fig.6) fell in this category)
  
  
• double peak
  not enough cases to draw consistent conclusions.

a) Peak with long decay;
b) Peak with short decay;
c) Single peak, high amplitude;
d) Single peak, low amplitude, fast decay;
e) Double peak

Conclusions

The main conclusions that were drawn from this study, if it is assumed valid to use radiographic flaw length measurements for common cases as a benchmark are:

• 6dB sides and DAC-12dB flaw length sizing techniques are more likely to give closer values to radiography;
• in some cases DAC-12dB can grossly oversize;
• in somes cases 6dB max can grossly undersize;
• fixed amplitude techniques (50%DAC and DAC-12dB) are less prone to operator variability, but this has to be weighed against large deviations (gross undersizing and oversizing);
• 6dB sides is an acceptable compromise, even with operator variability.

References

Since the presentation of this work to the CEN committee (CEN/TC121/SC5B/WG2), the fixed amplitude technique of DAC-10 has been proposed. As one might expect, length measurements are shorter than those using the DAC-12 technique, and in some cases avoid gross oversizing, but this can still be a problem.