Home study correspondence courses (UT, ECT, LPI and MPI) including NDT Problem Solver Software (solving standard equations in UT, RT and ECT)
|TABLE OF CONTENTS|
We make certain assumptions about the test conditions and presume that changes in time or amplitude are caused by variation in the parameter of interest. The assumptions made are based on all parameters being constant except the one we are interested in measuring changes in. For example, when performing a thickness measurement we assume the acoustic velocity of the test piece we are measuring is the same as the acoustic velocity in the calibration piece. We further assume that the temperature at which tests and calibrations are made are not important. Yet either or both of the parameters assumed fixed (materials velocity and temperature) can affect our test results. Variables affecting the test results will be divided into 4 groups:
Another factor relating to the results of an inspection is the Human Factor, this is a widely debated subject. The subject is not discussed in this chapter nor is the related subject of Probability of Detection. For more information on POD please refer to earlier publications on NDTnet
Scope - The primary variable in the scope is the linearity of the time base. Verification methods will usually require a tolerance in accuracy to a percentage of the total screen range (typically +/- 2%). This ensures no distance measured will be in error by more than 2%, e.g. for a 250mm range it may be possible to have an error of +/-5mm maximum in steel.
Pulser-Receiver - Amplitude uncertainties will result from variations in the linearity of the vertical deflection of the scope or due to inaccuracies in the amplitude control. Scope vertical linearity ensures that the relationship between two signals of different amplitudes is maintained over the entire range of the screen height. This is done by comparing the relative height of two echoes at different screen heights. e.g. setting two echoes 6 dB apart starting with one at 80% FSH, the other at 40% FSH adjustments are made to first increase the 80% FSH signal to 90% and 100%. The lower signal should be 45% and 50% respectively. Reducing the higher signal in 10% FSH increments, the lower should continue to be half its height. Tolerance for this parameter is +/-5% of the screen height. This ensures that the signal ratio of two different amplitudes truly indicates the size or distance effects. This would be most important for DGS type comparisons.
The other aspect of vertical linearity variability is the amplitude gain control. This applies to the calibrated gain control usually found in dB increments on a flaw detector. Since the dB is derived from (dB = 20 log A2/A1 changing the dB gain by a fixed amount should change the ratio of the signals. This allows us to expect a signal at 50% FSH to increase to 100% FSH when 6dB is added to the receiver gain. ASME code requires scanning of a weld to be done using 14dB over reference. This means a signal that was 20% of the reference amplitude at reference gain would then come up to the reference level denoted by the DAC. If the receiver gain is not linear the smallest recordable indication may be greater or less than the intended level. This will be another source of incorrectly sizing a defect with respect to a reference.
The effect of bandpass filters on displayed signal amplitude has been discussed in Chapter 6. The effect on amplitude is to reduce signal height if the centre frequency of the received signal is beyond (greater or less than) the bandpass region of maximum response. This can be a factor when it is known that the reflected or re-transmitted signal off a defect has a frequency spectrum determined by the characteristics of the defect (this is the basis of acoustic spectrum analysis or frequency analysis).
BS 4331 Part 3*, recommends the following probe/system performance checks;
on rough surfaces, such as castings, twice daily
||monthly and when large changes in probe angle or index are observed
||monthly and whenever repairs have been made to either probe or instrument and if one instrument is replaced with another
||daily and after repairs or replacement as above
||monthly and after repairs or replacement as above|
The above monitoring items apply to contact testing probes. The wear experienced by movement on metal surfaces tends to accelerate changes in performance. Some of the changes introduced by wear can alter test results significantly. As an example, consider beam angle change. If at the start of the day a nominal 60° probe was found to have an actual angle of 62°, an indication is found with a soundpath of 150mm in a 100mm thick plate butt weld. A quick calculation made by the operator allows plotting to position the defect 132.4 mm along the surface from the exit point and then 70.4 mm below the test surface. After several hours of heavy scanning the operator has unwittingly worn the probe down on the nose and the actual angle changes to 58° (both 62° and 58° are within acceptable tolerances for a nominal 60° probe). When the same indication is investigated again the operator finds an indication in that same area but it does not seem to be the same. The soundpath is only 133mm for the "new" indication. The operator, believing the refracted angle is 62° now plots a defect 62mm below the surface. This situation is illustrated in Figure 8-1.
Index point or beam exit point for an angle beam probe is easily established on the 11W block. This is used to establish the actual refracted angle so its accuracy within +/-1.5mm is essential. For longer soundpaths (>25mm the effect on positioning a flaw forward or backward would not be too critical. e.g. If the flaw plotted in Figure 8-1 was 1mm forward or backward due to the exit point being off 1mm from its scribed position it would have little bearing on the evaluation of the defect. If, however, the weld tested was on a 6mm thick pipe with a TIG root the root width might be about 2mm. An error in exit point placement could plot the defect on the wrong side of the weld.
In the chart of items checked as per BS 4331, the first three items are unique to contact probes, but the remaining items could be considered by any transducer evaluation, including immersion probes. Handling and aging can cause changes to the element's backing, degree of polarization, lensing material shape, lens material bond to the element or degree of loading (for the thin gold face on PVDF elements). These changes result in changes in both amplitude and frequency. The effect on performance is multi-stepped.
For example: if aging has resulted in a slight disbonding of the element from the high density backing of a standard ceramic element, its damping will be reduced. This will lead to an increased ringing. More ringing reduces resolution and increases the extent of dead zone due to the rattle. Decreased damping due to the disbondment, however, allows vibrations to be larger so sensitivity is increased. The reduction in backing load tends to change the centre-frequency to a higher value but the increased sensitivity, afforded by more and larger vibration displacements, reduces the bandwidth. The higher frequency increases the near zone as it is a function of wavelength. The angle of divergence is also changed (decreased ) as it too is a function of wavelength.
Operating probes in warm water (>50°C) or high radiation fields (several MegaRads) can cause blistering or disbonding of the epoxy material used for lensing. This could have similar effects to those noted for backing disbondment as well as distorting and redirecting the beam centre-line.
In addition to aging and environmental causes of alterations to the transducer performance, handling can also cause changes to occur. A sharp jolt from dropping a probe may result in similar disbond problems. With the availability of different pulse shapes it may be possible to deteriorate polarization in an element. A negative going pulse voltage is normally applied to probes but polymer elements tend to perform better with a positive going pulse. Polymer probes will show no deterioration if pulsed with negative going spikes but ceramic elements may experience depolarization over extended periods of time. Depolarization will reduce sensitivity and the increased gain required will manifest itself in a lower s/n ratio.
Sources of variation in transducer performance are many. Establishing a baseline with tolerances and then monitoring for changes in any of the parameters checked will help to ensure reliability of test results.
Entry surface variables include:
In addition to reduced coupling, which will reduce signal amplitudes, the rough surface increases the rate of wear on the probe. On an otherwise smooth surface isolated protrusions such as weld spatter can hinder or stop probe motion or in the case of mechanized systems there may be sufficient force to move the probe past the obstruction but this could result in damaging the probe by either tearing it from its mounting or severely scoring the plastic wedge. When the dirt on the test piece is very fine (similar to a flour texture) coupling can be prevented due to surface tension preventing the liquid couplant penetrating to the metal. Unless a transfer value has been established between test piece and calibration piece, this could go undetected.
In addition to affecting coupling, surface roughness tends to reduce signal amplitude by scattering and focusing the beam. This applies to both contact and immersion testing.
Whether uniform or irregular, a rough surface has the potential to present a scattering effect at an interface where a beam impinges. The degree of scattering is based on the ratio of roughness to wavelength. When roughness is less than about 1/10 a wavelength, scatter will be negligible. To reduce signal losses due to scattering an operator can select a lower frequency probe. With a wavelength of 0.37mm in water for a 4MHz probe, signal loss due to scatter can occur for irregularities as small as about 0.04mm. In addition to signal reduction another effect of surface irregularities is to redirect and mode convert some energy which when returned to the probe can be the source of spurious signals. In contact testing false indications from standing waves resulting from scatter on rough surfaces will normally have short soundpaths. They can be eliminated as true flaws by failing to locate any trace of indication from the full skip or from the opposite side.
Unless done properly, removal of surface roughness by mechanical means can result in further scattering problems. Small curved gouges left by a grinding wheel used to remove spatter or machining grooves can form small lenses. The affect of grinding can be unpredictable. Some of the lensing may concentrate the beam thereby increasing signal amplitude, or, the lens effect may be a de-focusing of the beam, again resulting in lower than expected signal amplitudes. Uniform surface preparation by sand or shot blasting usually provides a good surface for ultrasonic testing. Removal of excess metal by a hand held grinding wheel is commonly used on weld caps and roots. When a pipe weld has had its root ground flush and inspection can only be performed from the outside diameter, quality of grinding can result in unnecessary repair calls if grinding has been along the weld axis. The small grooves made by the grinding wheel run parallel to the root edge and are easily confused with lack of fusion, missed edge or undercut defects.
When thickness testing is done on a painted surface the paint thickness can add error to the reading. For example:
But attenuation and velocity of couplants are not independent properties. Each is a function of other parameters. Unless these parameters are controlled or in some way compensated for, gross variations from the reference value or calibration conditions can result.
Attenuation of couplants varies with material composition as would be expected. Published attenuation values are available for many materials as indicated in the table below. Attenuation coefficients are often quoted in nepers which allow for frequency dependence. 1 Np = 8.686 x f2 = dB/cm. Table 8-1 indicates attenuation of some common liquids.
|Material|| Attenuation (Np x 10-15)
In more practical terms, for water, this would mean an attenuation of about 5 dB per metre. Since such long water path lengths are not normally used the 0.005 dB/mm attenuation is considered negligible. But for the heavier oils attenuations 200 to 500 times greater can have significant effects on signal amplitude and frequency content. For the fixed delaylines or wedge materials used in contact testing attenuation variations can be far more pronounced and variation between manufacturers can cause considerable response differences. For example the plastics listed in table 8-2 are typical wedge materials selected by manufacturers and based on velocity for refraction purposes, but attenuation differences would result in noticeable amplitude response variation and frequency content of transmitted waveforms. Since the operator rarely knows what wedge material a manufacturer has used, little can be done to correct for potential variations in periodic inspections where results of tests taken with one or more years separation are compared.
|Material||Attenuation dB/cm @ 5 MHz||
||6.4 to 12.4
||2.75 to 2.61
lexan (poly carbonate)
||1.8 to 3.6
||2.32 to 2.48
||2.8 to 16
||2.6 to 2.77|
Attenuation is not a material constant. Under changes in conditions it can change. For example attenuation in water is inversely proportional to both temperature and pressure.
At standard pressure and temperature (1 atmosphere and 20°C) attenuation in water is 25.3 x 10-15 Np. When temperature is 0°C and water still liquid attenuation is 56.9 x 10-15 Np and at 40° it is 14.6 x 10-15 Np. At 1000 atmospheres attenuation drops to 12.7 x 10-15 Np and increases to 18.5 x 10-15 Np in a vacuum (zero atmospheres) when the temperature is held at 30°C.
Attenuation of couplants need rarely be considered when calibration and test conditions are the same couplant material, temperature and pressure. However, mechanical actions can add to variations in attenuation under some conditions e.g. liquid soap is often used in contact testing. Under static conditions it provides reasonable coupling, ease of probe movement and clean hands. When a part is inspected with more rapid probe motion than may be used for static calibration it is possible to lather the soap. As bubble density builds in the couplant attenuation will increase.
Far more pronounced on test results are the affects of velocity changes. As with attenuation of couplants, velocity is normally considered fixed for a given material. Providing all parameters affecting velocity are controlled the assumption is valid but subtle changes in parameters can have significant results.
Just as plastic compositions change in velocity so too does water. Pure water at 20°C and 1 atmosphere pressure has a velocity of 1480m/s. But water is not normally pure. As salinity increases as in sea water, acoustic velocity increases. At 20°C in sea water with a 3% salinity the acoustic velocity increases to about 1515m/s. For work done on off-shore structures, where immersion work would occur using the surrounding sea water, any calibrations done on-board ship must use water of the same salinity as will occur at the depth of test or probe placement and refracted angle will be mis-calculated. As well, acoustic velocity increases with pressure in water so with increasing water depth velocity also rises. This is relatively insignificant but may be corrected for by the equation.
Temperature is undoubtedly the most significant parameter affecting acoustic velocity. As such, its control or knowledge of its change is essential to ensure inspection accuracy. Change in acoustic velocity with change in material temperature is termed acoustic temperature dependence. Strangely, temperature dependence is not always the same sign, i.e. increasing material temperature will increase acoustic velocity in some materials and decrease it in others.
Table 8-3 illustrates some of the variations in temperature dependence (V/t) for some materials.
|Material||Acoustic Velocity @25°C||
(V / t (m/s °C)
polyvinyl chloride (hard)
By comparison most metals have a temperature dependence of between -0.5 to -5.0 m/s/°C depending on mode and axis of propagation.
|Table 8 - 3a: |
Velocity change with Temperature
| Velocity |
|5 ||1440 ||3,14
||15 ||472 ||2,65
||25 ||1498 ||2,15
||35 ||1520 ||1,67
||45 ||1536 ||1,18
||55 ||1548 ||0,68
||65 ||1555 ||0,20
Figure 8-5 illustrates the effect of temperature on refracted angles in steel for three common fixed wedge angles.
Incident angles indicated in the legend in Figure 8-5 are those to produce nominal 45°, 60° and 70° refracted angles under standard conditions. The plastic calculated for is UVA II with an acoustic velocity of 2760 m/s. The affect on angulated longitudinal waves is even more pronounced, this due to the greater ratio difference in velocities between the plastic and steel in compressional mode.
Some sources also consider the relative hardness of the probe face with a greater coupling or contouring available from softer material such as plastics and virtually no contouring available.
To avoid machining calibration blocks for every possible radius and surface condition compensation is made by adding gain to the receiver. The amount of compensating gain can be determined by a simple transfer value or it can be calculated using formulae and charts. Examples of the charts used for convex curvature compensation are found in Figures 8-6 and 8-7. These are from Australian Standard 2207 - Methods For The Ultrasonic Testing of Fusion Welded Joints in Steel. Two conditions are considered. In the first figure a nomogram is used to correct for losses when the probe contact is made on the curved surface. The test part radius is located on the left-hand scale and a line made through the appropriate probes diameter on the middle scale. The point on the right-hand scale this line intersects is the amount of gain to add as a correction factor.
In the second figure the probe makes contact on a flat surface but the beam reflects off a convex curve thereby redirecting portions of the beam away and reducing the maximum possible reflected energy. The graph used does not consider probe diameter, instead, ratio of surface curvature to metal path thickness is used. moving vertically up from the ratio axis (horizontal axis) at the appropriate ratio for your work piece, the point on the vertical axis where the curve is intersected gives the necessary correction factor.
Other codes and specifications may use different equations or graphs but the intent is the same.
Attempts to compensate by simply adding gain may not be adequate. Improvement is had by contouring the plastic wedge to the test piece. However, since the probe can no longer be calibrated on a flat reference piece this makes machining of a reference piece of the exact same geometry a necessity. For very small parts even this may prove unsatisfactory due to the production of surface waves and other spurious signals associated with large time differences of the beam in the wedge or delayline. If contouring probes proves too noisy then immersion methods or even another NDT method may have to be considered.
Geometry is not only a consideration as a potential source of signal variation but also of feasibility. Consideration must be given to beam shape when interaction with a boundary occurs. Formation of unwanted surface waves and mode converted waves will result due to finite extents of a beam. The single ray drawn from the probe exit point at the refracted angle is merely a convenient presentation of the principle of the test. In reality portions of the beam will impinge at greater and lesser angles due to divergence or side lobes (in the near zone).
This occurs more often for curved surfaces than for flat surfaces because not only does the beam divergence increase the incident angle but the point of incidence on a curved surface is always receding so further increase in incident angle results.
Curved surfaces make plotting more difficult than the simple trigonometry for flat surfaces. Compare similar conditions for a 45° beam in flat and curved plate. Using a thickness of 20mm the signal obtained at 35mm for an inspection of a 100mm radius pipe occurs 19.2 mm from the OD test surface and 31.7mm along the test surface from the exit point to the point over the indication. On a flat piece this would indicate an indication 15.2mm down and 24.8mm from the exit point. See Figure 8-9.
Geometry can also limit inspections. Again, concentric geometries are a common problem in this regard. A critical ID/OD ratio exists that will not allow an angle beam from the OD to intersect with the ID. This is shown in Figure 8-10.
This critical angle occurs when the ray representing the centre of beam is tangential to the inside surface of the pipe. Two situations may exist; either the angle is fixed (as in contact testing) and the critical ratio is to be determined or the ratio of wall thickness to the OD is known and the maximum angle that can be used is sought.
The equation relating wall thickness t, pipe outside diameter D and refracted angle can be written
The other parameter being sought, maximum angle, would use the following equation:
|Sin||= 1 -||----|
Although the equations and graph indicate that high angles of refraction may be used to ensure both ID and OD are seen, the actual size of the pipe will limit the practical maximum angle due to beam width and the resultant surface waves that occur. For tubular products where wall thickness approaches the wavelength the wall is flooded with sound and plate waves discussed in Chapter 5 result.
Some authors have indicated that contact pipe inspection in the circumferential direction can be accomplished by simply moving the probe over the pipe surface by a distance equal to one full skip. The principle involved is shown in Figure 8-12.
This draws well for the centre of beam ray but when applied to real conditions this can supply little more than a go/no-go inspection. Beam spread, mode conversions and attenuation will not permit accurate locating of any defects occurring several skips away. In fact attenuation will probably limit detectability of defects by this method to pipes under 10-20cm diameter. Such a technique may be useful when access is limited to only half of the circumference. The presence (but not necessarily the location) of a flaw may be detected by first scanning in one direction to the obstruction then the other, as in Figure 8-13.
In cooling molten metals, solidification begins at many sites throughout the melt. At each site the growth pattern is determined by the surrounding liquid material and the surrounding temperature gradients. Crystals form as growth progresses. Eventually the crystals' growth is halted when another crystal is encountered and all the liquid has been consumed. Although we generally consider the metal to be homogeneous, on a microscopic scale the boundaries formed by the edges of the grains make it in-homogenous. For an ultrasonic wave each crystal presents a different acoustic impedance depending on orientation and degree of inter grain bonding. As well, there may be pores, gaps and non-metallic inclusions. All these factors will cause scattering.
Just as with surface roughness, scatter will be a function of wavelength. Krautkramer points out that for grain sizes up to about 1/100th of a wavelength scatter can be considered negligible. However, as grain size increases beyond that, it can become a significant factor adding to decreasing signal amplitudes. As grain sizes increase to greater than 1/10th the wavelength, inspection may not be possible by ultrasonics. Austenitic stainless steels are typical of metals with large grain structures. In the production of austenitic steels manufacturers often attempt to control or limit grain size. This is done by :
a) introducing small amounts of grain refining elements
b) limiting the temperature the steel is heated to
c) by hot working the steel to break up the austenite grains.
In spite of these effects the stainless steel product is not always consistent in its grain structure. When testing stainless steel forgings, it is possible to have areas of higher attenuation than others. In cases such as this it will require observant operators to recognize the increase in grass level.
Velocity changes with material and condition as well. Contact angle wedges are normally made for steel so the refracted angle indicated on the wedge assumes it will be used on a steel with a longitudinal velocity of about 5900m/s and a shear wave velocity of about 3250m/s. When the same wedge is used on aluminium plate with longitudinal velocity 6320 m/s and shear at 3100 m/s the 'nominal' refracted angle indicated will not indicate the true angle.
But one need not move to a completely different metal to illustrate velocity changes. Rolled plate for pipeline construction shows variations of 8-10% in shear wave velocities. This is attributed to rolling and heat treat differences and the resulting differences in grain elongation and orientation. Even alloys of steel show marked variation in acoustic velocity; 4340 steel has a shear velocity of 3240 m/s while in 4150 steel it is 2770 m/s, a difference of nearly 17%.
These variations point out the importance of material specific calibration pieces.
Finally, as with couplants, acoustic velocity of a test material varies with temperature. Most published values will indicate velocities determined at 20°C. For work at much higher or lower temperatures corrections will need to be made. This will require the temperature dependence for the material to be established and this will have to be in addition to similar corrections made for couplant changes.
Mode converted wave off defect at 65° strikes radius providing large amplitude signal that when plotted gives a virtual location of a defect different from its actual position
If the adjacent surface in 8-14 had been flat the reflected shear and mode converted signals would both have been undetected. Such signals produced by mode conversion can be differentiated from flaws by using a couplant wetted finger to rub the test surface. The compressional mode is noticeably damped.
Mode converted longitudinal and re-directed shear waves can be seen when drawing distance amplitude correction curves on some calibration blocks. The typical ASME calibration block uses a side drilled hole at quarter thickness (1/4t). The convex radius of the hole allows both situations to occur (see Figure 8-15).
Maximum reflection occurs off a free boundary. This is effectively the situation for non-fusion and cracks where the void is air. However, when a dissimilar material fills the void, as would be the case in a slag inclusion or tungsten inclusions in a TIG weld or carbide inclusions in castings or forgings, part of the sound incident on the boundary is transmitted. This will reduce the reflected signal. Added to the loss due to transmission into the next medium is the associated losses due to the reflection at any angle other than 0°. Reflection and transmission coefficients discussed earlier show how quickly amplitudes can change due to boundary material differences and incident angles.
Finally, ambient conditions of pressure can have profound effects on a signal amplitude. Studies have shown a signal from a surface breaking crack to be reduced by 20dB when the cracked sample was placed in compression. This effectively "closed the gap".