|   685 views |
 Technical Discussions | | Jeff
 | Jeff
|  03:46 Jun-14-2002
Flaw Detection based on 2:1 Noise to Flaw Ratio Does anyone have any information on this? We are using a 2:1 base metal-to-defect ratio for detection in Friction Stir Weldments. Example: lack-of-penetration is defined when the indication is 6dB higher (2:1) than the noise floor in base metal. Any information on techniques that use this would be helpful. Thanks - Jeff
| |
| | | |  Godfrey Hands
Consultant,
PRI Nadcap, United Kingdom, Joined Nov 1998,  304  | | 09:39 Jun-15-2002
Re: Flaw Detection based on 2:1 Noise to Flaw Ratio : Does anyone have any information on this? We are using a 2:1 base metal-to-defect ratio for detection in Friction Stir Weldments. Example: lack-of-penetration is defined when the indication is 6dB higher (2:1) than the noise floor in base metal.
.
: Any information on techniques that use this would be helpful.
.
: Thanks -Dear Jeff,
The problem with a 2:1 Flaw/Noise ratio is that the noise can add to the flaw amplitude or subtract from the flaw amplitude. This means that in the worst case scenario where the flaw amplitude (independent of noise) is twice the noise amplitude, and this is then reduced in amplitude by the noise, you have no resultant signal left above the noise.
Conversely, when the amplitude is just below the "threshold" without noise (i.e. acceptable imperfection), and is added to by the noise, this then exceeds the threshold and becomes an apparently rejectable condition.
A 3:1 ratio is more often used, as this gives closer calls to the real condition.
Is thereany way that you can improve the signal to noise ratio and work with 3:1 ? e.g. Higher frequency probes etc ? Good Luck, Godfrey Hands
.
: Jeff
.
| |
| | | | David Ponder
| David Ponder
| 07:04 Jun-15-2002
Re: Flaw Detection based on 2:1 Noise to Flaw Ratio : : Does anyone have any information on this? We are using a 2:1 base metal-to-defect ratio for detection in Friction Stir Weldments. Example: lack-of-penetration is defined when the indication is 6dB higher (2:1) than the noise floor in base metal.
: .
: : Any information on techniques that use this would be helpful.
: .
: : Thanks -
.
: Dear Jeff,
: The problem with a 2:1 Flaw/Noise ratio is that the noise can add to the flaw amplitude or subtract from the flaw amplitude. This means that in the worst case scenario where the flaw amplitude (independent of noise) is twice the noise amplitude, and this is then reduced in amplitude by the noise, you have no resultant signal left above the noise.
: Conversely, when the amplitude is just below the "threshold" without noise (i.e. acceptable imperfection), and is added to by the noise, this then exceeds the threshold and becomes an apparently rejectable condition.
: A 3:1 ratio is more often used, as this gives closer calls to the real condition.
: Is there any way that you can improve the signal to noise ratio and work with 3:1 ? e.g. Higher frequency probes etc ?
.
: Good Luck,
.
: Godfrey Hands
: .
: : Jeff
: .
.Dear Fellow TechniciansI think that increasing the frequency of the probes will only increase the metal noise. The heat affected zone will give you all kinds of false indications with higher frequency. You will not be able to tell the difference between an actual defect and grain boundaries in the material. If you even have the option go with lower frequency. If the data you are using is anything like the AWS specs all they are saying is, any indications that are 6dB higher than the metal noise shall be considered lack of penetration. AWS uses alot of the same type stuff, check AWS D1.5
| |
| | | |  Terry Oldberg
Engineering, Mechanical Electrical Nuclear Software
Consultant, USA, Joined Oct 1999, 42 |
Terry Oldberg
Engineering, Mechanical Electrical Nuclear Software
Consultant,
USA,
Joined Oct 1999
42
| 09:14 Jun-29-2002
Re: Flaw Detection based on 2:1 Noise to Flaw Ratio The subject of this posting contains an internal inconsistency that makes the subject itself nonsensical. In particular, "noise" and the related concept "signal" are defined statistically but flaw detection tests do not obey statistics ( see http://www.ndt.net/article/v04n05/oldberg/oldberg.htm ).
| |
| | | |
|
