I am looking for info on the relationship of EDM notches and natural cracks when performing Ultrasonic Inspection. Since EDM notches would seem to be more reflective than a natural crack, what ratio, if any do you use. For example, for a .025 by .500 critical crack, what size EMD notch would you use? Are there any studies comparing the UT response from EDM notches and natural cracks? Is there a standard ratio used in your industry, such as 2 to 1? Any help would be appreciated.

01:54 Mar-16-2004 Uli Mletzko R & D, Retired Germany, Joined Nov 1998 ^{89}

Re: EDM notches vs natural cracks in UT ----------- Start Original Message ----------- : I am looking for info on the relationship of EDM notches and natural cracks when performing Ultrasonic Inspection. Since EDM notches would seem to be more reflective than a natural crack, what ratio, if any do you use. For example, for a .025 by .500 critical crack, what size EMD notch would you use? Are there any studies comparing the UT response from EDM notches and natural cracks? Is there a standard ratio used in your industry, such as 2 to 1? Any help would be appreciated. ------------ End Original Message ------------

Hallo Dave, there is no general relationship between the signal height (amplitude) of artifical calibration reflectors or virtual modelling reflectors on one side and the signal height of real discontinuities (flaws or defects) on the other side. US response from flaws in isotropic materials is depending on size, shape, location, orientation, surface roughness, transparency (crack closure effect), wavelength etc. E.g., if you have a rough crack, the signal from a perpendicular wave might be lower than from a smooth notch, but the signal from a inclined wave might be higher. In general, you should not trust on such relationships to calculate safety margins etc. Only for very special applications, were you have always similar testing parameters (statistical high number of specimens, similar degradation mechanism or flaw type), it might be possible to find a valid relationship. IMHO, the only way to solve the problem is, not to use signal heights and recording levels, but to record the signals down to noise and then use imaging techniques like SAFT, echo tomography or TOFDT, and to try to find the real border lines of the flaws.

Regards, Uli Mletzko, MPA, University of Stuttgart, Germany

04:27 Mar-17-2004 N.Kuppusamy Consultant, NDT Level-III Engineer Advanced Inspection & Testing (S) Pte Ltd, Singapore, Joined Dec 2003 ^{34}

Re: EDM notches vs natural cracks in UT Hi Dave Katz,

EDM notches are definitely more reflective than natural cracks because they are more perfect planar defects. The Natural carcks, on the other hand, are less perfect planar defects. The ramifications and orientation of cracks are different from EMD notches that no body can predict.

Moreover there can not be two identical natural cracks (i.e., the reflectivity of two natural cracks indications are not the same). That is why we use the term Equivalent Defect Size.

When the Flaw detector is standardised using EDM notch or natural crack (say, the echo is kept at 80% of Full Screen Height)and you received a maximised defect indication equavalent to the reference indication (80% Full Screen Height, then it can be described as the Equivalent Defect Size of the defect is that of reference.

It can not be termed as absolute or actual defect size because What we do is a qualitative measurement. Hence no ratio shall be applicable between EDM and Natural cracks. This quality level of decribingindications are acceptable to the industry.

Best Regards,

N.Kuppusamy

----------- Start Original Message ----------- : I am looking for info on the relationship of EDM notches and natural cracks when performing Ultrasonic Inspection. Since EDM notches would seem to be more reflective than a natural crack, what ratio, if any do you use. For example, for a .025 by .500 critical crack, what size EMD notch would you use? Are there any studies comparing the UT response from EDM notches and natural cracks? Is there a standard ratio used in your industry, such as 2 to 1? Any help would be appreciated. ------------ End Original Message ------------

09:46 Apr-21-2004 Terry Oldberg Engineering, Mechanical Electrical Nuclear Software Consultant, USA, Joined Oct 1999 ^{42}

Re: EDM notches vs natural cracks in UT You've raised an excellent question. Unfortunately, it lacks an answer.

To understand why this is so, focus on the fact that an inspection system comprises methods for A) generating data from observations of material B) extracting patterns from training data and C) predicting outcomes for material by recognizing patterns in field data. By "pattern," I mean a condition on the data from an observation; an example is "a peak amplitude at the output from my instrument of more than 1.23 volts or an average amplitude of more than 0.65 volts." "Extracting patterns" is the process of setting those conditions on the data that predict outcomes in "training data" best. "Training data" are data from observations for which the outcomes are known.

In the scenario that you describe, an outcome of interest is whether material contains a crack of critical size. Pattern extraction is performed on training data from observations of material containing EDM notches of known depth but material in the field contains natural cracks. "Do patterns extracted from training data from observations of EDM notches predict whether material in the field contains a critical crack" is a way of restating your question with respect to the outcome of whether or not there is a critical crack.

Your question lacks an answer because the intellectual apparatus for answering it is framed in terms of probabilities and, as I have pointed out elsewhere in this forum, all of the various methods for detecting discontinuities embody an empirical violation of probability theory. In particular, the intellectual apparatus contains Shannon and Weaver's Mathematical Theory of Communications, wherein concepts such as "information," "signal" and "noise" are defined in terms of probabilities. You might be surprised to learn that, while "signal" and "noise" are often bandied about in NDT, they are actually undefined concepts in reference to discontinuity detection.

----------- Start Original Message ----------- : I am looking for info on the relationship of EDM notches and natural cracks when performing Ultrasonic Inspection. Since EDM notches would seem to be more reflective than a natural crack, what ratio, if any do you use. For example, for a .025 by .500 critical crack, what size EMD notch would you use? Are there any studies comparing the UT response from EDM notches and natural cracks? Is there a standard ratio used in your industry, such as 2 to 1? Any help would be appreciated. ------------ End Original Message ------------

Re: EDM notches vs natural cracks in UT Hello Dave

The notches are not the only way to make a cracklike reflector. We had a study where we studien relations of warious reflectors including real craks. If you want a expensive way of makeing cracks that is thermal cracking. There you have some possibilities to control lenght and height. Other a bit cheaper way is to drill holes in a row. Use a different diameters of drills and some misalignment. Try these holes you´ll be suprised. EMD notches appears a bit too much mirrorlike reflectors.

Re: EDM notches vs natural cracks in UT My colleague Charles Schneider has supplied me with the following information regarding this question.

Unfortunately, there is no ratio that can be universally applied. The ratio varies widely according to the inspection situation. It is particularly sensitive to: a) angle of misorientation (i.e. angle of incidence for pulse-echo testing) b) the ratio sigma/lambda (s/l) of the roughness of the crack face to the ultrasonic wavelength. See J M Coffey 'Quantitative assessment of the reliability of ultrasonics for detecting and measuring defects in thick-section welds' in Proc Conf 'Tolerance of flaws in pressurised components' (London, 16-18 May 1978, IMechE) for typical examples at 0 degree and 20 degree incident angles. (These examples were based on formulae in M V Berry 'The statistical properties of echoes diffracted from rough surfaces', Proc Royal Society Vol 273A, 611-654, 1973.) A more general discussion of the problem can be found in J A Ogilvy 'Theory of wave scattering from random rough surfaces' (Bristol, Adam Hilger, 1991). This book includes some useful approximations to the ratio you want for the following limiting cases: - s/l ¤ 0.3 (where diffuse scatter dominates) - very small s/l, under near-specular conditions (where perturbation theory gives a reasonable approximation). However, for many inspection situations where the roughness falls between the limiting cases above, I think it is best to estimate the ratio using established modelling software. TWI operates a number of such codes under licence. All the above presupposes that you know the roughness s of the crack face. Charles Schneider may the best source of such information for many of the typical cracking mechanisms in steel.

----------- Start Original Message ----------- : You've raised an excellent question. Unfortunately, it lacks an answer. : To understand why this is so, focus on the fact that an inspection system comprises methods for A) generating data from observations of material B) extracting patterns from training data and C) predicting outcomes for material by recognizing patterns in field data. By "pattern," I mean a condition on the data from an observation; an example is "a peak amplitude at the output from my instrument of more than 1.23 volts or an average amplitude of more than 0.65 volts." "Extracting patterns" is the process of setting those conditions on the data that predict outcomes in "training data" best. "Training data" are data from observations for which the outcomes are known. : In the scenario that you describe, an outcome of interest is whether material contains a crack of critical size. Pattern extraction is performed on training data from observations of material containing EDM notches of known depth but material in the field contains natural cracks. "Do patterns extracted from training data from observations of EDM notches predict whether material in the field contains a critical crack" is a way of restating your question with respect to the outcome of whether or not there is a critical crack. : Your question lacks an answer because the intellectual apparatus for answering it is framed in terms of probabilities and, as I have pointed out elsewhere in this forum, all of the various methods for detecting discontinuities embody an empirical violation of probability theory. In particular, the intellectual apparatus contains Shannon and Weaver's Mathematical Theory of Communications, wherein concepts such as "information," "signal" and "noise" are defined in terms of probabilities. You might be surprised to learn that, while "signal" and "noise" are often bandied about in NDT, they are actually undefined concepts in reference to discontinuity detection. : : I am looking for info on the relationship of EDM notches and natural cracks when performing Ultrasonic Inspection. Since EDM notches would seem to be more reflective than a natural crack, what ratio, if any do you use. For example, for a .025 by .500 critical crack, what size EMD notch would you use? Are there any studies comparing the UT response from EDM notches and natural cracks? Is there a standard ratio used in your industry, such as 2 to 1? Any help would be appreciated. ------------ End Original Message ------------

04:35 May-07-2004 Terry Oldberg Engineering, Mechanical Electrical Nuclear Software Consultant, USA, Joined Oct 1999 ^{42}

Re: EDM notches vs natural cracks in UT This is to address a statement of John Rudlin that may be misleading. When he states that that "The ratio varies widely according to the inspection situation," this implies the existence of such a ratio, given an "inspection situation." Does such a ratio exist?

To answer this question meaningfully, one must define what is meant by "ratio." If it means the ratio of EDM to natural crack size that produces the same signal-to-noise ratio, this ratio often exists in a given inspection situation. However, if it means the ratio of EDM to natural crack size that produces the same or similar inspection reliability statistics, this ratio does not, for the usual defect detection test, exist. That the EDM notch produces the same signal-to-noise ratio does NOT imply that it produces the same or similar inspection reliability statistics and this is true even in a "given inspection situation." On the other hand, what is needed from a laboratory mockup is the same or similar inspection reliability statistics.

Why, then, do people like Dave Katz concern themselves with creating artificial flaws that produce the same signal-to-noise ratio as natural flaws? One possibility is that they are fooled by the ambiguity of the associated language. Possible appearances to the contrary, knowing a signal-to-noise ratio does not imply knowledge of either the signal or the noise. The latter are what determine the inspection reliability statistics.

----------- Start Original Message ----------- : My colleague Charles Schneider has supplied me with the following information regarding this question. : Unfortunately, there is no ratio that can be universally applied. The ratio varies widely according to the inspection situation. It is particularly sensitive to: : a) angle of misorientation (i.e. angle of incidence for pulse-echo testing) : b) the ratio sigma/lambda (s/l) of the roughness of the crack face to the ultrasonic wavelength. : See J M Coffey 'Quantitative assessment of the reliability of ultrasonics for detecting and measuring defects in thick-section welds' in Proc Conf 'Tolerance of flaws in pressurised components' (London, 16-18 May 1978, IMechE) for typical examples at 0 degree and 20 degree incident angles. (These examples were based on formulae in M V Berry 'The statistical properties of echoes diffracted from rough surfaces', Proc Royal Society Vol 273A, 611-654, 1973.) : A more general discussion of the problem can be found in J A Ogilvy 'Theory of wave scattering from random rough surfaces' (Bristol, Adam Hilger, 1991). This book includes some useful approximations to the ratio you want for the following limiting cases: : - s/l ¤ 0.3 (where diffuse scatter dominates) : - very small s/l, under near-specular conditions (where perturbation theory gives a reasonable approximation). : However, for many inspection situations where the roughness falls between the limiting cases above, I think it is best to estimate the ratio using established modelling software. TWI operates a number of such codes under licence. : All the above presupposes that you know the roughness s of the crack face. Charles Schneider may the best source of such information for many of the typical cracking mechanisms in steel.