Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
14 Findings and Application 3.1 Critical Flaw Size Analytical parametric studies were performed in order to establish the critical flaw sizes considered rejectable for typical bridge CJP welds. The results were used to identify the critical flaw size that must be reliably detected and rejected to establish revised acceptance and rejection criteria. The critical flaw sizes were evaluated based on fracture mechanics using a BS 7910  FFS approach, and the internal flaws are assumed to be cracks. The parametric studies that were per- formed included various plate geometries, welds, residual stress fields, flaw types, flaw sizes, and locations. 3.1.1 Volumetric Flaws Flaw severity is dependent on the flaw type (i.e., planar vs. volumetric). Therefore, volumetric (i.e., non-planar such as slag and porosity) flaws and planar flaws were considered sep- arately. Fracture mechanics relies on an underlying assump- tion that a flaw is a crack, whether considering fatigue crack growth or fracture. This is a very conservative assumption for smooth, rounded volumetric flaws, but assuming that all volumetric flaws are void of any cracks or crack-like geom- etry is an unconservative assumption to make. During the literature review, it was apparent that all existing enhanced ultrasonic testing specifications do not distinguish between flaws that are volumetric as opposed to crack-like, other than including additional criteria for rejecting scattered volumet- ric flaws, such as a limit on the number of point-like reflec- tors over a specific weld length. In the round robin testing performed during this research project, inspectors were requested to document whether a detected weld discontinuity was volumetric or planar. Responses showed that volumetric discontinuities and planar discontinuities could not be differentiated with reasonable accuracy. Not only were volumetric discontinuities often reported as being planar, but planar discontinuities were often reported as being volumetric. It was clear that the eval- uation of truly planar discontinuities using an acceptance criteria developed for volumetric discontinuities will not capture the criticality of planar discontinuities and would be unconservative. Therefore, it was determined that PAUT acceptance criteria should not require flaw characterization as planar and volumetric flaws, but rather all indications should be evaluated against the same acceptance criteria. In order to develop target critical flaw size for volumet- ric flaws, existing experimental test data on CJP welds with such flaws was explored. Harrison [22â24] investigated the fatigue resistance and strength of butt welds with embed- ded slag inclusions and porosity. That research found that the effect of slag inclusions and porosity below 10% volume on ductile strength was negligible due to the overstrength of the weld metal. The porosity limit was set to 10% since this was considered the limit that should be allowed without masking other indications during RT inspection. Slag inclusions were found to typically have limited through-thickness heightâ the critical dimension according to fracture mechanicsâdue to occurring between weld passes. During these studies, data from high-cycle fatigue tests performed by multiple research- ers were collected by Harrison to develop an acceptance criteria based on S-N curves. This was used to set limits on the percent- age of the volume of weld metal containing porosity and the maximum length of slag inclusions that were later incorporated into BS 7910. These limits assume that the flaw is verified to be volumetric and was developed to be used with RT. BS 7910 states that âthe tolerable porosity sizes based on ultrasonic test- ing may be considerably less, particularly for thinner sections.â This is because BS 7910 recognizes that volumetric flaws can be easily undersized by ultrasonic testing. BS 7910 is silent on how specifically the allowable sizes should be reduced for ultrasonic testing; however, this would likely be determined through some kind of performance testing and POD evaluation. The critical dimensions for slag inclusions in BS 7910:2013 are given in Table 4 when combining the height requirements C H A P T E R 3
15 for fracture (no length requirement for fracture) and the length limit for fatigue (no height requirement for fatigue) assuming infinite fatigue life. As shown in this table, the criti- cal dimensions for slag inclusions in BS 7910 were already quite small before considering the effect of undersizing with ultrasonic testing. AWS D1.5 RT acceptance criteria include limits on maxi- mum indication length and spacing. While these limits are workmanship-based and not based on experimental results, they do help by highlighting the typical flaw size that has been traditionally accepted. A minimum slag height for RT sensi- tivity was also calculated by comparing the size and density of required wire-type image quality indicators (IQI) to determine the required slag height for the same change in density. This analysis assumed the density of slag to be 3.5 g/cm3 based on literature [25, 26]. The maximum slag length and minimum slag height for each plate thickness are shown in Table 5. This confirmed that the BS 7910 maximum slag height would likely be detectable under RT inspection and that the BS 7910 maxi- mum slag length requirements compare reasonably to the AWS D1.5 RT acceptance criteria for stress ranges of â¼5 ksi. The critical planar flaw sizes must also be considered when evaluating critical volumetric flaws since calculation of the stress intensity factors (KI) shows that any sharp or planar flaw extending from a slag inclusion or porosity will result in an equivalent planar flaw extending over the combined area of the volumetric and planar flaws. Thus, once a crack begins to extend from a volumetric flaw, it is equivalent to a crack which extends over the entire projected area. An appropriate delta-K threshold (DKth) to control fatigue crack growth of volumetric flaws was evaluated using fatigue test data. Using the combined fatigue and fracture limits on slag given by BS 7910 in Table 4, it was found that the DKth varied from 1.5 to 2.0 ksi in depending on the stress range. In addition, NCHRP Report 335: Acceptance Criteria for Steel Bridge Welds  recommended using 2.5 ksi in for DKth in the development of a modified RT acceptance criteria for bridge welds. NCHRP Report 335 included fatigue testing on weld samples with porosity and slag weld imperfections and reported that the calculated initial DK ranged from 2.25 to 3.7 ksi in . As will be discussed in the following section, 2.5 ksi in was used for DKth in this project for determining the critical flaw size for planar flaws. Since this is also appro- priate for volumetric flaws, planar and volumetric flaws have the same critical flaw size based on fatigue. Therefore, the tar- get critical planar flaw sizes shown in the following sections were used to develop target critical volumetric flaw sizes for the CIVA parametric matrix. 3.1.2 Critical Planar Flaw Sizes Based on Fatigue Various planar flaws were modeled using FFS procedures to determine the maximum flaw size that will not grow in fatigue (i.e., infinite life). Failure was defined as the limiting flaw size at which crack growth is expected at a given stress range since the intention is to evaluate welds in a fabrication shop environ- ment where the anticipated annual daily truck traffic (ADTT) would not need to be taken into consideration for each indi- vidual weld. The inputs for the analysis accounted for varia- tions in the plate thickness, flaw aspect ratio, flaw position, and magnitude of the applied live load stress range. Thickness transition weld geometries were also considered by accounting for stress concentrations through finite element analysis. It is noted that one could utilize a finite life approach and assume a number of cycles per day over the design life which could result in larger flaw tolerance; however, such an approach was not deemed to be implementable for several reasons. For example, the designer would need to provide detailed static and cyclic stress data for each weld. Then, the inspector would need to use this data to select the appropriate inspection criteria for the specific weld. Clearly, there is much room for error in this approach in addition to the fact it places considerable respon- sibility on the technician to interpret the data. Stress Range (ksi) Slag Height (2a) (0.5" thick weld) Slag Height (2a) (> 1" thick weld) Slag Length (2c) 6.09 0.063" 0.118" 0.098" 5.37 0.063" 0.118" 0.157" 4.64 0.063" 0.118" 0.394" 4.06 0.063" 0.118" 1.378" 3.34 0.063" 0.118" Infinite Table 4. Combined BS 7910 slag fatigue and fracture requirements. Plate Thickness (in ") Min. Slag Height (2a) Max. Slag Length (2c) 0.5" 0.028" 0.125" 1" 0.044" 0.313" 2" 0.057" 0.500" 3â4" 0.089" 0.500" Table 5. Slag RT sensitivity and AWS D1.5 length requirements.
16 220.127.116.11 Cyclic Stress Range To determine the stress range for evaluation, a reasonable approach is to use the stress range associated with infinite life; for example, using a stress range of 16 ksi for a Category B butt weld detail. However, the stress range associated with the CAFL (constant amplitude fatigue limit) of Category B (16 ksi) is very high in terms of actual in-service stress ranges for a fatigue- resistant detail such as a CJP butt weld. The FFS calculations indicated that a 16 ksi stress range results in a very small maxi- mum permissible flaw size to ensure infinite life. It is noted that NCHRP Report 335 recommends using about half of the fatigue limit for the detail under consideration, for example, 8 ksi to 9 ksi for Category B, as it better represents the upper bound in-situ effective stress range. NCHRP 335 even sug- gests using lower values if data exists for justification or if deemed acceptable by the engineer. Therefore, critical crack sizes were calculated for stress ranges from 4 ksi to 20 ksi in 4 ksi increments during the parametric study. 18.104.22.168 Delta-K Threshold (DKth) The delta-K threshold (DKth), which correlates to initiation of fatigue crack growth, was determined through review of previous research that had performed fatigue testing of welds in structural steels, including NCHRP Report 267: Steel Bridge Members Under Variable Amplitude Long Life Fatigue Load- ing  and NCHRP Report 181: Subcritical Crack Growth and Fracture of Bridge Steel . The appropriate value of DKth depends on the R ratio which is the ratio of minimum applied stress to maximum applied stress. The typical R ratio at a butt weld is quite high due to the presence of residual and dead load stresses. According to Barsom and Novak , DKth is equal to 2.05 ksi in for an R ratio of 0.8. According to Fisher et al. , the constant amplitude crack growth thresh- old approaches 3 ksi in at high R ratios, but he found that DKth approaches 2 ksi in for an R ratio of 0.8 for random vari- able block loading when only a small percentage of stress intensity factor ranges exceeds the constant amplitude crack growth threshold. Therefore, based on the previous testing performed on DKth for typical structural steels, 2.5 ksi in was used for DKth in this project for both planar and volumetric flaws (i.e., planar and volumetric flaws have the same critical flaw size for fatigue). While the specific residual stresses were not considered in the fatigue evaluation, the effect of the residual stresses on the fatigue crack growth was accounted for in the determination of the DKth through the R ratio. Since infinite life was utilized in the analysis, the effect of residual stresses on the da/DN curves did not affect the analysis. Therefore, the yield strength (i.e., grade) is not a variable in the test matrix. 22.214.171.124 Summary of Parametric Study Using the inputs of the fatigue-loading parametric study, the maximum crack height and length that would initiate crack growth for CJP butt welds was determined for various a/c ratios, crack position, plate thickness, stress range, and type of weld (equal thickness or thickness transition). Fig- ure 3 shows typical crack position and weld type. Signal FFS software  was used to perform this analysis by converting the fatigue crack growth problem to an equivalent fracture problem by setting the fracture toughness (Kc) equal to the DKth value of 2.5 ksi in and the primary stress equal to the applied stress range. The specific joint types that were evalu- ated for fatigue are summarized in Table 6. The fatigue analysis and findings are presented in the fol- lowing sections, along with the fracture analysis findings for each geometry type including equal thickness butt welds and thickness transition butt welds. The results of T and corner joints are presented in Appendix A. 3.1.3 Critical Planar Flaw Sizes Based on Fracture Similar to the fatigue loading models, planar flaws have been modeled using FFS procedures to determine the maxi- mum crack size that will not result in fracture. The Option 1 FFS procedure found in BS 7910:2013 was utilized to evaluate Figure 3. Typical flaw positions within weld.
17 the effects of various flaw types on the performance of CJP welds. While finite element models were developed to evalu- ate the effects of plasticity for transition welds, the Option 1 failure assessment diagram (FAD) curves were used to account for the inherent variability in the fracture mechanics inputs. A parametric study was performed to account for variations in the plate thickness, flaw aspect ratio, flaw position, magni- tude of the combined dead and live stresses, residual stresses, and material properties (i.e., strength and fracture toughness). 126.96.36.199 Yield and Tensile Strength Both yield and tensile strength must be included in the fracture analysis. Since steel products are generally delivered above the minimum specified yield strength, the analyses assumed that the expected yield strength is 10% greater than the minimum specified yield strength (i.e., FyExpected = 1.1FyNominal). Similarly, the expected tensile strength was also assumed to be 10% greater than the minimum specified ten- sile strength. The expected yield and tensile strength of the as-placed weld were assumed to match the expected yield and tensile strength of the base material. Specific material grades considered in the parametric study were 36, 50, 70, and 100 ksi. 188.8.131.52 Fracture Toughness A fracture assessment requires toughness to be known or at least assumed. At present, the U.S. bridge industry does not require any specific level of toughness or testing for as-placed production CJP welds or for the heat-affected zone (HAZ). In the absence of any specifications, one could con- servatively use the lower-bound estimate of KIc per the master curve of around 20 ksi in , where KIc is the critical fracture toughness at initiation of fracture. However, this will result in extremely conservative estimates of tolerable flaw sizes. Rather than assume a lower-bound toughness, the material toughness was included in the fracture analyses as a variable in the parametric study. Toughness was varied from a lower bound of 25 ksi in to 100 ksi in in 25 ksi in increments to establish the sensitivity and relationship between flaw size and toughness. It was assumed that the toughness used in the model corresponds to that of the as-placed weld and HAZ at the lowest anticipated service temperature. Charpy v-notch (CVN) requirements are included in ASTM A709  for base metal of production welds and AASHTO/AWS D1.5  for weld metal of groove weld qualification test plates. These welds do not necessarily match the thickness or joint geometry of the production welds. While these requirements were used to estimate weld metal fracture toughness, previous research has found that production welds may have lower fracture toughness than the procedure qualification record (PQR) test welds due to differences in the joint geometry, and so use of minimum Charpy requirements of PQR test welds may be unconser- vative in some cases . The CVN values were converted into KIc values using the Charpy/fracture toughness correlation in BS 7910. This approach utilizes the master curve to estimate the material toughness. Other factors in this equation are the probability that the material fracture toughness (Kmat) is less than esti- mated, the thickness of the material, and a factor (TK), which describes the scatter in the Charpy versus fracture toughness correlation. The probability that Kmat is less than estimated was set to 50% since the Charpy value used in the correlation is only the minimum specified value and the scatter of the data is unknown. The thickness was taken as 4â³ unless the Type of Crack/Position Type of Butt Weld A/C Ratio Thickness Applied Stress Range (Category B) Near Surface Equal Thickness 0.01-1.5 0.5", 1", 2", 3", 4" 4 ksi, 8 ksi, 12 ksi, 16 ksi, 20 ksi Transition 0.01-1.5 0.5" to 1", 1" to 1.5", 1" to 2", 2" to 3", 2" to 4", 3" to 4" 4 ksi, 8 ksi, 12 ksi, 16 ksi, 20 ksi Â¼ depth Equal Thickness 0.01-1.5 0.5", 1", 2", 3", 4" 4 ksi, 8 ksi, 12 ksi, 16 ksi, 20 ksi Transition 0.01-1.5 0.5" to 1", 1" to 1.5", 1" to 2", 2" to 3", 2" to 4", 3" to 4" 4 ksi, 8 ksi, 12 ksi, 16 ksi, 20 ksi Mid-depth Equal Thickness 0.01-1.5 0.5", 1", 2", 3", 4" 4 ksi, 8 ksi, 12 ksi, 16 ksi, 20 ksi Transition 0.01-1.5 0.5" to 1", 1" to 1.5", 1" to 2", 2" to 3", 2" to 4", 3" to 4" 4 ksi, 8 ksi, 12 ksi, 16 ksi, 20 ksi Table 6. FFS evaluation matrix for CJP butt welds.
18 permissible thickness for a certain grade of steel was smaller. The recommended TK term of +25Â°C was used to account for the scatter in the Charpy correlation. It was found that, except for non-fracture critical Grade HPS 100W, the base metal Charpy requirements resulted in lower KIc values than the weld metal requirements. In order to summarize the data, the minimum base or weld metal fracture toughness values in AASHTO Temperature Zone II at the maximum permissible thickness are shown in Table 7. Grades 36 through 50W have KIc of approximately 46 ksi in for fracture critical welds. Grade 70W has a KIc of approximately 60 ksi in , and Grade 100W has KIc of approxi- mately 75 ksi in for fracture critical welds. Therefore, use of 50 ksi in for Grade 36 through 50W and 75 ksi in for Grade 100W is probably reasonable. It should be noted that this is an approximate correlation and the fracture toughness in the HAZ is unknown. 184.108.40.206 Primary Stresses The primary stresses (i.e., combined dead and live load stresses) were assumed to be 75% of the minimum specified yield strength of the base material. This correlates to the allow- able stress level for the operating rating in the 2nd Edition Manual for Bridge Evaluation . The HL-93 rating factor for the operating rating is 1.3 times greater than the inventory rating factor. A reasonable stress level for inventory loading is 0.55*Fy, since this correlates to the historic allowable stress design limits. Multiplying 1.3*0.55 approximately results in the 0.75 factor. This is believed to be a reasonable level for maximum anticipated loads in a bridge member and that using 100% of the minimum specified yield strength would be overly conservative. 220.127.116.11 Residual Stresses The effects of residual stresses were included in analytical models for fracture by assuming a uniform tension stress. In reality, the actual residual stress profile highly depends on the type of weld, the welding procedure, etc. Furthermore, the effects of repairs and starts/stops can drastically affect residual stress fields. For double-bevel CJP welds, assumed residual stress profiles typically result in compressive residual stresses in the root of the weld. This compressive residual stress helps with resisting fracture for embedded discontinuities, but it results in different critical crack sizes depending on the location of the embedded discontinuity with respect to location within the weld. For new welds, the acceptance criteria would also need to be different depending on the weld type (e.g., single bevel, double bevel, electroslag) and the welding procedure (e.g., heat input). In addition, different criteria would need to be devel- oped for repair welds where the residual stress fields would also be considerably different. Therefore, a simplified approach for acceptance criteria for new welds was utilized due to the additional variables which would need to be included in the acceptance criteria (i.e., one would require knowledge of heat input, travel speed, top or bottom of the plate) in order to account for the different residual stress profiles and the loca- tion of the discontinuity in relation to the residual stress profile. Use of a uniform residual stress profile equal to 100% of the expected yield strength is recommended by BS 7910 for an initial assumption. BS 7910 includes additional non- uniform residual stress profiles in Annex Q, and API 579-1/ ASME FFS-1  released updated non-uniform residual stress profiles with the 2016 edition of the code. For both codes, the residual stress is equal to the assumed material yield strength at the surface of the weld. For embedded flaws, the maximum residual stress in the middle two-thirds of the thickness was evaluated. For BS 7910, this results in a residual stress of approximately 70% of the assumed material yield strength. For API 579-1/ASME FFS-1, the residual stress pro- file depends on the heat input of the weld divided by the weld thickness, with lower heat inputs or thicker welds resulting in higher residual stresses. Previous research has found that typical bridge heat inputs are between 50 kJ/in to 90 kJ/in . If the heat input was 50 kJ/in, the low heat input category would be invoked for welds greater than 1.5â³ thick. If the heat input was 90 kJ/in, the low heat input category would be invoked for welds 3â³ thick or greater. For the low input case, the maximum residual stress in the mid- dle two-thirds of the thickness according to API 579-1/ ASME FFS-1 would be 100% of the assumed material yield strength. The medium heat input case results in residual stresses of approximately 60% of the assumed material yield strength in the middle two-thirds of the thickness. Two levels of residual stress were evaluated during the para- metric study. The first level was residual stresses equal to the expected yield strength of the base metal (1.1*FyNominal), as it seems that residual stresses of 100% of the expected material yield strength are an appropriate assumptions for surface Steel Minimum Base or Weld Metal in Zone II at Maximum Thickness (ksi in) Non-FC Gr. 36 - HPS 50W NA Non-FC Gr. HPS 70W >56.51 Non-FC Gr. HPS 100W 68.4 FC Gr. 36 - HPS 50W 45.8 FC Gr. HPS 70W >60.41 FC Gr. HPS 100W >75.11 Note: 1T0 was slightly conservative since minimum Charpy energy was rounded down to either T27J or T40J. Table 7. Typical fracture toughness values for bridge welds.
19 flaws. The second level was equal to two-thirds (66%) of the expected yield strength of the base metal, as it seems that using 66% of the assumed material yield strength is appropriate for embedded flaws in most cases. BS 7910 includes an additional equation which accounts for global relief of residual stresses due to primary loading of the structure. This equation was utilized in the evaluation which resulted in a reduction of the actual residual stresses at the surface from 100% of the expected yield strength to approximately 75% of the expected material yield strength for equal thickness butt welds and 60% of the expected yield strength for thickness transition butt welds. 18.104.22.168 Summary of Parametric Study Using the inputs of the fracture parametric study, the max- imum crack height and length that would resist fracture was determined for various a/c ratios, yield and tensile strength, fracture toughness, primary stress, and residual stress using SignalFFS software . The specific joint types that were evaluated for fracture are the same as those evaluated for fatigue shown in Table 6. The fracture analysis and findings are presented in the fol- lowing sections, along with the fatigue analysis findings for each geometry type, including equal thickness butt welds and thickness transition butt welds. The results of T and corner joints are presented in Appendix A. 3.1.4 Critical Flaw Sizes for Equal Thickness CJP Butt Welds For the analysis of CJP butt welds, the embedded and near surface flaws were assumed to be centered in a 12â³ wide plate. As expected, increasing the stress range resulted in a large decrease in the maximum flaw size that would not result in fatigue crack growth. Likewise, decreasing the fracture toughness of the material resulted in a large decrease in the maximum flaw size that can resist fracture. The maximum flaw size was similar for embedded cracks at the Â¼ point and the midpoint through the thickness of the plate except for thin plates (i.e., 0.5â³ thick) at a 4 ksi stress range or for frac- ture toughness of 100 ksi in . In these cases, a slight increase in maximum flaw size is seen for cracks at the midpoint com- pared with the Â¼ point through the thickness of the plate. Due to the similarities of the Â¼ point and midpoint embedded crack results, the results from the Â¼ point were used for the critical crack size of embedded flaws. It was found that increasing the thickness of the plate resulted in an increase in the critical flaw size for both surface and embedded flaws for both the fatigue and fracture analyses, but the sen- sitivity in critical flaw size to changes in plate thickness was much lower for limiting fatigue crack growth than limiting fracture. In other words, the critical flaw size was very similar for all thicknesses when accounting for fatigue crack growth. The results from the fatigue and fracture parametric studies were combined in order to determine the target criti- cal flaw size. Obviously, there was a wide spectrum of target critical flaw sizes depending on the inputs. Plots were created for both 100% and 66% residual stress assumptions for each plate thickness and for both surface and embedded cracks. Due to the wide variation in flaw size due to the underlying input parameter assumptions, the critical crack size data were grouped based on selected parameters that were reasonable for typical bridge welds. Based on the discussion of inputs noted above, the assumptions given in Table 8 were deter- mined to be most reasonable for target critical flaw sizes of highway bridge welds. After completing the review and grouping data, simpli- fied tables for the combined fatigue and fracture analysis of equal thickness CJP butt welds were developed. The cor- responding target critical flaw height and length at various aspect ratios for the cases in Table 8 are shown in the follow- ing tables: Table 9 for surface flaws in equal thickness welds, and Table 10 for embedded flaws in equal thickness welds. The results in Table 9 and Table 10 were compared to the closed fracture mechanics equations for a surface crack and embedded crack in an infinite plate. Using the equations provided by Anderson , the stress intensity factor for a circular surface crack in an infinite plate is KI = 1.29s a and KI = 1.13s a where âaâ is the entire crack height for the surface crack and half the crack height for the embedded crack. These equations can be easily modified for fatigue crack growth by exchanging DKth for KI and the cyclic stress range (Sr) for stress (s), resulting in DKth = 1.29 Sr a and DKth = 1.13 Sr a for surface and embedded cracks, respec- tively. Table 11 gives the results of the critical size of circular flaws in an infinite plate for fatigue loading only using inputs of 2.5 ksi in for DKth and 4 ksi or 8 ksi for Sr. The 8 ksi stress Residual Stresses (% of Expected Yield Strength) 100% for surface flaws, 66% for embedded flaws Yield Stress Grade 36â50 Grade 70â100 Grade 36â100 Fracture Toughness Kc=50 ksi in Kc=75 ksi in Kc=75 ksi in Fatigue Stress 4 ksi4 ksi 8 ksi Table 8. FFS inputs for target critical flaw sizes.
20 range results given in Table 11 matched the target surface and embedded flaws for 4â³ thick plates given in Table 9 and Table 10. The 4 ksi stress range results given in Table 11 are larger than the target surface and embedded flaws for 4â³ thick plates given in Table 9 and Table 10 since these cases were controlled by fracture rather than fatigue crack growth. It quickly became apparent that the critical crack sizes were quite small in some cases. To verify if the approach was yield- ing reasonable estimates, the simplified results were compared to the flaw size acceptance criteria in ISO 19285:2017 . This PAUT inspection document includes acceptance criteria to be determined by either measuring the height and length of the flaw or by measuring the length and the maximum amplitude. There are three levels of quality specified in this document, but Level 2 would typically be used for bridge welds. Table 12 includes the maximum flaw height and length for surface and embedded flaws over various plate thicknesses. For thin plates (i.e., 0.5â³), the critical flaw height tends to be similar to those calculated during this project, but the critical Yield Stress Grade 36â50 Grade 70â100 Grade 36â100 Fracture Toughness Kc=50 ksi in Kc=75 ksi in Kc=75 ksi in Fatigue Stress Plate Thickness Flaw Height x Flaw Length Flaw Height x Flaw Length Flaw Height x Flaw Length 0.5" 0.059" x 1.076" 0.063" x 0.736" 0.149" x 0.199" 0.037" x 1.061" 0.065" x 0.218" 0.095" x 0.127" 0.024" x 1.082" 0.032" x 0.187" 0.058" x 0.115" 0.063" x 0.084" 1" 0.063" x 1.044" 0.136" x 0.339" 0.155" x 0.207" 2"â4" 0.064" x 1.059" 0.137" x 0.343" 0.157" x 0.209" 8 ksi4 ksi4 ksi Table 9. Target surface flaw in equal thickness weld. Yield Stress Grade 36â50 Grade 70â100 Grade 36â100 Fracture Toughness Kc=50 ksi in Kc=75 ksi in Kc=75 ksi in Fatigue Stress Plate Thickness Flaw Height x Flaw Length Flaw Height x Flaw Length Flaw Height x Flaw Length 0.5" 0.141" x 1.010" 0.221" x 0.276" 0.097" x 1.023" 0.125" x 0.314" 0.200" x 0.200" 0.061" x 1.020" 0.063" x 0.571" 0.093" x 0.170" 0.143" x 0.143" 1" 0.181" x 1.066" 0.254" x 0.462" 0.107" x 1.066" 0.125" x 0.430" 0.251" x 0.251" 0.062" x 1.041" 0.063" x 0.963" 0.095" x 0.174" 0.152" x 0.152" 2"â4" 0.196" x 1.031" 0.253" x 0.550" 0.110" x 1.004" 0.125" x 0.462" 0.262" x 0.262" 0.063" x 1.046" 0.096" x 0.174" 0.153" x 0.153" 8 ksi4 ksi4 ksi Table 10. Target embedded flaw in equal thickness weld. Defect Type 4 ksi Stress Range 8 ksi Stress Range Surface 0.234" x 0.469" 0.058" x 0.117" Embedded 0.612" x 0.612" 0.153" x 0.153" Table 11. Circular flaw in an infinite plate with fatigue loading only. Plate Thickness Surface Flaws Embedded Flaws Flaw Height x Flaw Length Flaw Height x Flaw Length 0.5" 0.039" 0.079" x 0.5" 0.039" 0.079" x 0.5" 1" 0.039" 0.079" x x x x x 1" 0.039" 0.157" x 1" 2"â3" 0.079" 0.118" x 1.969" 0.079" 0.197" x 1.969" 4" 0.118" 0.157" x 2.362" 0.118" 0.236" x 2.362" x x x x Table 12. ISO 19285:2017 Level 2 acceptance criteria.
21 flaw length tends to be a bit longer for the ISO acceptance cri- teria. As the plates become thicker, the ISO acceptance criteria allows larger flaws than were calculated during this project. Overall, the ISO acceptance criteria compares reasonably well with the critical flaw sizes computed in this project. ASME Code Case 2235-13  includes an acceptance criteria based on measurements of flaw height and length. The acceptance criteria vary by aspect ratio (a/c) and flaw height (a/t) for thicknesses 1â³ and greater. Therefore, at least three different aspect ratios were evaluated to compare the range of acceptable flaw sizes. It was found that the results developed in this study are typically conservative compared to the sizes in the ASME Code Case for the high-strength steels or high stress ranges. For thin plates (i.e., 0.5â³), the critical flaw height tends to be within 1â16â³ for approximately similar length flaws, but for thick plates (i.e., 3â³) the differ- ence between the critical crack sizes developed in this project and the ASME limits increases and reaches values of approxi- mately 5â8â³ in some cases. For the 4 ksi stress range cases, the critical crack sizes compare quite well, with the ASME Code Case allowing slightly smaller crack sizes for thin plates and slightly larger crack sizes for thick plates. For the 8 ksi stress range case, the critical crack size developed in this project is typically much smaller than the ASME limits. This is due to the limit placed on fatigue crack growth in this project, which was not a consideration during the development of the ASME Code Case 2235 limits . 3.1.5 Critical Flaw Sizes for Thickness Transition CJP Butt Welds Butt welds with a transition in thickness have a stress con- centration at the start of the transition. This stress concen- tration may decrease the fatigue and fracture resistance of cracks, especially if they are located near the surface of the plate. Typical girder flange welds utilize single-sided (offset) transitions since the web plate remains at a consistent height, as shown in Figure 4. The butt weld is typically located in the region with the greatest stress concentration, at the point where the thin plate begins the thickness transition. The gen- eral location of the butt weld and HAZ is visually apparent in Figure 4. In order to account for this effect, the magnitude of the stress concentration needs to be determined for vari- ous thickness transitions, as shown in the specimen matrix in Table 6. Finite element analysis was used to calculate the mag- nitude of the stress concentration using the commercial solver ABAQUS. All of the thickness transitions used a 1 to 2.5 slope, which is the maximum allowed by AASHTO for Category B butt welds. The thickness transitions with a minimum plate thickness of 0.5â³ were modeled as web plates, while the thickness transitions with a minimum plate thickness equal to or greater than 1â³ were modeled as flange plates. For the flange plate models, the flanges were assumed to be 18â³ wide while the web was 0.5â³ thick by 36â³ deep. For the web plate models, various flange sizes were used to capture different levels of stiffness. In order to capture the restraint provided by the web, the finite element analysis was performed using a 3D model. Typ- ical transition butt weld geometries have a smooth radius as shown in Figure 4. Therefore, to better represent the actual conditions, a 2â³ radius was used at the change in slope. The model was loaded with a unit traction in the axial direction on the thin plate side while the thick plate side was restrained from movement in the axial direction. Figure 5 displays the typical profile for stresses in the lon- gitudinal direction of the girder. This figure shows a view cut through the flange at the point of maximum axial stress. As expected, there is a concentration of axial stresses at the radius of the transition to the thin plate. The lowest axial stresses occur where the thickness transition ends at the thick plate. The peak stress occurs within the radius, but the stresses in front of the radius within the thin plate are still increased in the region very close to the radius. Typically, the butt weld would be located in the thin plate near the thickness transition as shown in Figure 4. In this configura- tion, the far bevel face is located at the start of the transition radius. The upper portions of the weld near the top surface will therefore experience increased stresses compared to the rest of the thin plate. The results from various thickness transitions were com- pared by plotting the stress at each integration point along a path extending from the point of maximum stress con- centration vertically through the thickness of the flange. The location along the thickness of the flange was then normalized in order to compare the various transition geometries. As expected, the greatest stress concentration factor (SCF) occurred at the transitions with the greatest relative change in thickness, such as 1â³ to 2â³ or 2â³ to 4â³. It should be noted that the stress concentration occurred within the top half of the plate, and decreased quickly away from the plate surface. Thin Plate Thick Butt Weld Figure 4. Typical thickness transition butt weld on bridge girder flange plate.
22 In order to estimate the effect of the stress concentration on the fatigue resistance of transition welds, a polynomial trendline was fit to the results from the three transitions with the greatest SCFâ1â³ to 1.5â³, 1â³ to 2â³, and 2â³ to 4â³ transi- tionsâto obtain the following result, where t is the normal- ized depth in the plate, with t = 0 on the transition-side face and t = 1 on the opposite face: 1.6587 4.8158 5.0949 3.0826 1.79214 3 2( ) = â + â +SCF t t t t tFlange The SCF estimation was then multiplied by the various nom- inal stresses shown in Table 6 to obtain a through-thickness fatigue stress profile for each nominal stress level. The polyno- mial stress profiles were then entered into Signal FFS to obtain the maximum crack size that would not initiate fatigue crack growth for both embedded and surface cracks. A separate web thickness transition model was developed for thickness transitions where the thin plate was less than 1â³ thick. Rather than the thickness transition occurring only on one side of the web, the web plate transition occurred on both sides of the web so that the web plate remained centered on the flange. The web plate thickness transition used in this analysis was 0.5â³ to 1â³ thick, since this was assumed as the largest anticipated thickness transition for a web plate. Unit axial traction was placed on the web and flange on the thin side of the weld. It was found that the SCF of web thickness transitions had a dependency on the relative stiffness of the flange plate. There- fore, the flange size was varied as follows: no flange, 1â³ Ã 6â³, 1.5â³ Ã 9â³, and 4â³ Ã 18â³. The location of the greatest SCF in the web thickness transition was located near the flange, as shown in Figure 6. This figure shows axial stresses in the model with the lower limit on the color palette set to the nominal stress. When there is no flange, the web SCF is small since the average stress level is equal to the nominal stress. As the flange stiff- ness increases, the average stress level through the thickness of the web increases at the point of maximum SCF. Since the maximum SCF in web thickness transitions is dependent on the flange stiffness, the SCF was increased slightly compared to the highest FE results for the estimation to be used in the Signal FFS fatigue evaluations in order to be conservative for possible flange/web combinations. The following equation displays the resulting estimation of the SCF for web transi- tions, where t is the normalized depth in the plate with t = 0 on the transition-side face and t = 1 on the opposite face: 0.8 0.8 1.42SCF t t tWeb ( ) = â + The equation for the SCF estimation was then multiplied by the nominal stress values given in the specimen matrix in Table 6 to obtain a through-thickness fatigue stress profile for each nominal stress level. The polynomial stress profiles were then entered into Signal FFS to obtain the maximum crack size that would not initiate fatigue crack growth for both embedded and surface cracks. Similar finite element models were used for fracture resis- tance analysis as the fatigue resistance. Due to greater applied stresses in the fracture analysis, plasticity effects needed to be accounted for through the use of nonlinear FE analysis using stress-strain curves for ASTM A709 bridge steels. The plasticity effects flatten the SCF curves near the surface of the thickness transition due to localized yielding and increase the depth of Max: 1.8 Figure 5. Stresses for 1â to 2â transition weld.
23 the SCF below the surface. When all of the thickness transitions were plotted, it was discovered that the thickness transitions with the highest SCF were nearly linear for points between the maximum SCF and the midpoint of the plate. Since the critical flaws will be those on the top surface or Â¼-point through the thickness, the model of the stresses in the FFS approach only needs to accurately represent the stress profile over the top half of the plate. Therefore, it was determined that a linear stress approximation could be used rather than a polynomial. The linear SCF estimation used for the fracture analysis of transitions was represented by using the following equations where Sm is the membrane (nominal) stress and Sb is the bending stress: 0.75 0.6 S F S S m y b m = = The values of Sm and Sb were then used in the Signal FFS software for the primary stresses along with the previously noted inputs for secondary stresses, thicknesses, fracture toughness, etc., to calculate the limiting crack size to resist fracture. Using this approach, the thickness transition CJP butt weld geometries in Table 6 were modeled as flat plates with a 12â³ width and a linearized primary stress profile for the primary stresses in the Signal FFS software to calculate the limiting crack size to resist fracture. The results for the target surface and embedded weld flaws in thickness transition welds are given in Table 13 and Table 14, respectively. In these tables, the plate thick- ness correlates to the thickness of the thinner plate. The FFS results were determined based on the SCF listed above, which were appropriate for the largest thickness transitions Figure 6. Web SCF for 4â Ã« 18â flange. Yield Stress Grade 36â50 Grade 70â100 Grade 36â100 Fracture Toughness Kc=50 ksi in Kc=75 ksi in Kc=75 ksi in Fatigue Stress 4 ksi4 ksi 8 ksi Plate Thickness Flaw Height x Flaw Length Flaw Height x Flaw Length Flaw Height x Flaw Length 0.5" 0.039" x 1.051" 0.063" x 0.254" 0.102" x 0.136" 0.028" x 1.029" 0.032" x 0.318" 0.073" x 0.097" 0.013" x 1.010" 0.016" x 0.111" 0.032" x 0.046" 1"â2" 0.032" x 1.057" 0.065" x 0.187" 0.081" x 0.107" 0.024" x 1.057" 0.032" x 0.176" 0.048" x 0.137" 0.061" x 0.082" 0.008" x 1.005" 0.016" x 0.046" 0.019" x 0.034" 0.020" x 0.027" 3"â4" 0.031" x 1.041" 0.063" x 0.181" 0.080" x 0.106" Table 13. Target surface flaw in thickness transition weld.
24 evaluated. Therefore, these results should be valid for typi- cal thickness transitions, regardless of the thickness of the thicker plate. Comparing the results for thickness transition CJP butt welds to the equal thickness CJP butt welds shown in Table 9 and Table 10, a larger reduction in critical crack sizes was seen for surface flaws in transition welds than for embedded flaws in transition welds. The reduction is also much greater for plates with a thickness 1â³ or greater since these plates were assumed to be flange plates with offset thickness transitions rather than web plates with centered thickness transitions for 0.5â³ plate thickness. The thickness transition CJP weld results were compared to ISO 19285 and ASME Code Case 2235. Neither of these documents has separate acceptance criteria for thickness transition welds, so accounting for the SCF decreased the critical crack sizes calculated for thickness transition CJP welds in this project compared with those allowed in ISO 19285 and ASME Code Case 2235. In general, the target flaw sizes developed during this project are similar to those allowed in the other acceptance criteria for 0.5â³ and 1â³ plate thicknesses with a 4 ksi stress range. For the other cases, the target critical crack sizes developed during this proj- ect are smaller than those allowed for the other acceptance criteria, especially for an 8 ksi stress range. Along with not accounting for the stress concentration of thickness transi- tion welds, the fatigue crack growth failure mechanism was not considered in the development of ASME Code Case 2235 . It is not known whether either of these factors was considered during the development of ISO 19285, as no lit- erature was found that documents the development of that acceptance criteria. The results of T and corner joints are presented in Appendix A. 3.2 Round Robin Results 3.2.1 Introduction A round robin testing program was performed in order to gain insight into the capabilities of the current tech- nicians in the steel bridge industry and to identify best practices for improved flaw detection and flaw charac- terization. The round robin testing program was used to determine the minimum flaw size that could be reliably detected with enhanced ultrasonic methods and how the advanced methods compare with the historical conventional UT method. Scanning of a set of specimens containing weld flaws was performed by five PAUT technicians, two TOFD technicians, and five conventional UT technicians. Data was only received by four of the five PAUT technicians who par- ticipated. Conventional UT and PAUT technicians were qual- ified according to the requirements in AWS D1.5 as ASNT Level II for UT and PAUT, respectively. There was tremen- dous difficulty identifying TOFD technicians to participate in the round robin testing program due to lack of availability or lack of equipment. While eleven TOFD technicians were contacted and five of these technicians initially agreed to take part, only two technicians actually committed to testing of the plates. It is believed that this is directly related to the lack of TOFD technicians in the bridge and building fabrication industry. The directions and weld bevel drawings supplied to conventional UT, PAUT, and TOFD technicians who par- ticipated in the round robin testing program are provided in Appendix H. PAUT technicians were requested to scan the plates in accordance to the requirements in Annex K and submit the scan plan details to the research team. Each technician pro- vided details on the number of line scans; location of line Yield Stress Grade 36â50 Grade 70â100 Grade 36â100 Fracture Toughness Kc=50 ksi in Kc=75 ksi in Kc=75 ksi in Fatigue Stress Plate Thickness Flaw Height x Flaw Length Flaw Height x Flaw Length Flaw Height x Flaw Length 0.5" 0.099" x 1.041" 0.126" x 0.349" 0.205" x 0.205" 0.077" x 1.031" 0.095" x 0.278" 0.127" x 0.196" 0.172" x 0.172" 0.038" x 1.097" 0.066" x 0.101" 0.094" x 0.094" 1" 0.110" x 1.003" 0.125" x 0.448" 0.245" x 0.245" 0.075" x 1.003" 0.095" x 0.250" 0.125" x 0.192" 0.175" x 0.175" 0.035" x 1.004" 0.063" x 0.089" 0.084" x 0.084" 2"â4" 0.114" x 1.037" 0.125" x 0.521" 0.264" x 0.264" 4 ksi4 ksi 8 ksi Table 14. Target embedded flaw in thickness transition weld.
25 scans (i.e., face and side of weld scanned and corresponding index offset); incidence angle range; angular sweep increment; calibration/TCG block details; equipment and transducer make, model, and settings; along with any other informa- tion included in AWS D1.5 Table K.2. Rather than providing the same scan plan to the technicians, having the technicians develop their own scan plan allowed for documentation of the variations in possible scan results for the same plate scanned within the requirements of Annex K. For instance, one tech- nician scanned the thickness transition plates with six line scans, one technician used five line scans, one technician used four line scans, and one technician used only two line scans. If the research team had provided a scan plan to the techni- cians, the scatter of results would be artificially limited since they would be provided with additional requirements beyond what is included in Annex K. Also, by providing a scan plan, the Research Team would have artificially influenced the flaw rejection results for PAUT by having the technicians either use an index offset which would maximize or possibly minimize the amplitude response of the known flaws. Table 15 provides the details of the flaws included in the round robin testing program. There was a total of 19 flaws implanted within the 11 plates that were circulated in the round robin. Some plates had multiple flaws while others did not have any flaws. In addition to the testing by the PAUT, TOFD, and UT technicians, the plates were tested with digital radiography and the full matrix capture/total focus- ing method (FMC/TFM) PAUT to aid in determining the as-built flaw sizes. The digital RT and FMC/TFM results are also shown in the following sections. It was determined that digital RT gave the best estimate of the true flaw location and length for volumetric flaws while FMC/TFM gave the best estimate of the true flaw height. The digital RT images of each plate are shown in Appendix D. It should be noted that not all flaws could be distinguished on the digital RT scans, especially lack of fusion (LOF) flaws, which do not produce a density change in the plan view. Testing was also performed using tra- ditional film RT, but it was found that the contrast of the flaws was poor after digitizing the film. There were also many film artifacts and scratches. For these reasons, the digital RT images were exclusively used for flaw location and length sizing. The FMC/TFM testing was performed blind so as not to skew the results by providing the intended flaw height. Two flaws were not detected by the FMC/TFM technician, and so flaw heights were not reported for these flaws. The flaw sizes shown in Table 15 are the sizes based on the flaw height from FMC/TFM and flaw length from digital RT, where this information was available. If the flaw height or length was not available based on the FMC/TFM or digital RT results, the flaw sizes provided by FlawTech for planar flaws or the intended flaw sizes for implanted volumetric flaws were used. A few of the implanted volumetric flaws also had unintended peripheral flaws which were noted on the digital RT scans. The peripheral flaws were included in the updated flaw loca- tion and length measurements. For these flaws, the accuracy in reported flaw location by the round robin technicians was considered both with and without the peripheral flaws in order to capture all possible hits, since some technicians may have included the peripheral flaws while others may not. The shear wave velocity of the round robin plates was measured and compared to a standard AISI 1018 calibration block to verify whether the plates would be considered acous- tically equivalent to typical calibration standards. The shear wave velocity of all of the round robin plates was found to be within 0.5% of the AISI 1018 calibration block, which had a shear wave velocity of 0.1275 in/Âµs. As will be explained further, this is within the Â±1% tolerance that was determined to result in negligible error in the amplitude and incidence angle. The anisotropic ratio (i.e., ratio of shear wave velocity in the rolled and transverse to rolled direction) was 0.5% and lower. This is also within the Â±1% tolerance that was noted to result in acoustic anisotropic behavior and beam splitting. The highlights of the results of the round robin testing program are briefly summarized in the following sections, along with a summary of the scatter in height and length measurements given in Appendix B, and the scatter in reported amplitude given in Appendix C. Flaw ID Flaw Type Height (in) Length (in) 1 LOF 0.03 0.06 2 LOF 0.22 0.40 3 LOF 0.20 0.40 4 Toe Crack 0.02 0.04 5 Crack 0.43 0.80 6 Crack 0.40 0.74 7 Crack 0.17 0.40 8 LOF 0.45 1.00 9 LOF 0.20 0.60 10 LOF 0.43 0.80 11 LOF 0.23 1.00 12 Porosity 0.09 3.31 13 Slag 0.32 0.37 14 Slag 0.16 0.18 15 Slag 0.10 0.90 16 Porosity 0.13 3.27 17 Slag 0.24 0.49 18 Slag 0.06 0.03 19 Slag 0.17 3.61 Table 15. Round robin flaw details.
26 3.2.2 Flaw Detection and Location The accuracy in reported flaw location was very poor for many of the PAUT and conventional UT technicians. Thus, a âhitâ (simply defined as the technician noting that they detected an indication which matched a known flaw) was originally defined as any reported indication where any part of the reported indication was within 1 inch along the longi- tudinal weld axis of any part of a known flaw in the plate. In other words, the reported indication did not have to line up with a known flaw or even overlap with a known flaw, as long as the gap between the extents of the reported indication and known flaw was less than 1 inch. Clearly this is a very liberal criterion for counting a reported flaw as a hit, but all techni- cians, including PAUT, TOFD, and conventional UT, seemed to struggle with accurately locating flaws. For PAUT and TOFD technicians, this may be due to inaccuracy in encoder calibration or incorrectly using the encoder, while for conven- tional UT technicians it may be due to physical measurement error. No consideration was made for correctly measuring the through-thickness location since the flaw depth was not con- sistently reported (i.e., top of flaw or maximum amplitude). It was determined not to be too stringent with developing a criterion for determining whether a reported flaw correlated to the intended flaw in order to have adequate data for flaw height and length sizing accuracy, along with variations in reported amplitude. API RP 2X: Recommended Practice for Ultrasonic and Magnetic Examination of Offshore Structural Fabrication and Guidelines for Qualification of Technicians  includes formulas for scoring the performance of ultra- sonic technicians when performing a qualification exami- nation. To be correctly located, this document recommends that the centerline of the reported indication be within the boundary of the actual indication or within Â½ inch of the actual centerline of the indication (whichever is greater). This requirement is obviously stricter than the requirement developed during the initial review of the round robin data where flaws did not need to overlap at all as long as the gap between any portion of the reported flaw and actual flaw was less than or equal to 1 inch. The hit/miss data are reported using both criteria in Tables 16â18, with hits labeled as â1â and misses labeled as â0.â The criteria when the gap between the actual and reported flaw is less than 1 inch is shown in Table 17, and the API RP 2X flaw location criteria is shown in Table 18. Since these tables highlight accuracy of flaw detection, the data is only shown for the PAUT results where all relevant indica- tions were asked to be reported regardless of the amplitude. The average hit rate for all of the flawsâplanar flaws (Flaws 1â11) and volumetric flaws (Flaws 12â19)âis shown in Table 16. The overall hit rate for PAUT, con- ventional UT, and TOFD was quite similar when using the 1-inch gap criteria for flaw location. For planar flaws, PAUT had the highest hit rate and conventional UT had the low- est hit rate. For volumetric flaws, TOFD had the highest hit rate, and PAUT had the lowest hit rate. Under the more strin- gent criteria included in API RP 2X, the methods that utilize encoded line scanning had a larger drop in overall hit rate than the manual, raster scanned conventional UT method. For planar flaws, TOFD had the highest hit rate, and PAUT had the lowest hit rate. For volumetric flaws, conventional UT had the highest hit rate, and TOFD had the lowest hit rate. The detection rate results for individual flaws were found to be highly variable. A small slag inclusion (Flaw 18: 0.06â³ high Ã 0.03â³ long) was not detected by any PAUT or conventional UT technician, but was detected by both TOFD technicians using the 1-inch gap criteria. It was noted in the inspection report for one of the TOFD technicians that this flaw had a low signal response. Another slag inclusion (Flaw 15: 0.10â³ high Ã 0.90â³ long) was detected by every PAUT and conventional UT inspection, but was missed by one of the TOFD technicians. 3.2.3 Flaw Rejection and Flaw Type Characterization Both flaw location criteria (i.e., the 1-inch gap criteria devel- oped by the authors and the API RP 2X criteria) were evaluated to compare the rejection rate for PAUT Annex K and conven- tional UT. Flaw classification and rejection rates are shown in Tables 19â22. The criteria when the gap between the actual and reported flaw was 1 inch or less is shown in Table 21, while the API RP 2X flaw location criteria is shown in Table 22, with rejection labeled as â1â and acceptance labeled as â0â in both tables. Location Criteria Method Planar Flaws Only Volumetric Flaws Only All Flaws 1-Inch Gap PAUT 0.97 0.68 0.86 Conventional UT 0.91 0.83 0.87 TOFD 0.95 0.94 0.95 API RP 2X PAUT 0.65 0.64 0.64 Conventional UT 0.75 0.83 0.78 TOFD 0.91 0.63 0.79 Table 16. Average hit/miss rate.
27 Flaw rejection from TOFD inspections cannot be compared to PAUT or conventional UT inspections since there is no acceptance criteria for this NDT technique in AWS D1.5. For the flaw to be considered rejected, it had to be located correctly as well as meeting any other criteria for rejection included in AWS D1.5. All flaws except Flaw 18, which is the very small slag inclusion, were considered rejectable by at least one con- ventional UT technician when using the liberal flaw detection criteria of the gap being 1 inch or less between the actual and reported flaw. The average rejection rate for PAUT and conventional UT is shown in Table 19 for planar flaws, volumetric flaws, and all flaws. The rejection rates were lower for PAUT per Annex K than for conventional UT, especially when considering the API RP 2X location criteria. The rejection rate for PAUT and conventional UT are much closer for planar flaws when using the 1-inch gap criteria while PAUT has a much lower rejection rate than conventional UT for volumetric flaws. It was found that while PAUT has a lower rejection rate for the small planar flaws such as Flaw 1 and 4, all of the planar flaws 0.2â³ high by 0.4â³ long or larger were rejectable per Annex K for the 1-inch gap criteria while some of these flaws were not detected by conventional UT. Due to inaccuracy of locating these large planar flaws with PAUT, many of these large planar flaws would not have been rejected by PAUT according to the flaw location criteria in API RP 2X. It is important to recognize that the PAUT and conven- tional UT rejection rates are for the same set of samples and is providing a direct comparison of the two methods. This is unlike some previous studies where the rejection rates of conventional UT and PAUT Annex K were computed using different sample sets, which could skew the data. It should be noted that the rejection rates include Flaw 18, even though this flaw was not rejectable according to any conventional UT or PAUT technicians. Therefore, this flaw has equally low- ered the rejection rate for PAUT and conventional UT, but the rejection rates of critical flaws would be slightly increased. Two indications (Flaw 1 and Flaw 14) were rejected by PAUT technicians, per Annex K, due to being characterized as cracks rather than due to amplitude and length. The current AWS D1.5 Annex K10.2 states that âindications characterized as cracks shall be considered unacceptable regardless of length Specimen Details PAUT1 Conventional UT TOFD Flaw ID Flaw Type PAUT1 PAUT2 PAUT3 PAUT4 PAUT Avg UT1 UT2 UT3 UT4 UT5 UT Avg TOFD1 TOFD2 TOFD Avg 1 LOF 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 LOF 1 1 1 1 1 1 1 1 1 1 1 1 3 LOF 1 1 1 1 1 1 1 1 1 1 1 1 4 Toe Crack 0 1 1 1 0.75 0 0 1 1 1 0.6 1 0 0.5 5 Crack 1 1 1 1 1 1 1 1 1 1 1 1 6 Crack 1 1 1 1 1 1 1 1 1 1 1 1 7 Crack 1 1 1 1 1 1 1 1 1 1 1 1 8 LOF 1 1 1 1 1 1 0 1 1 1 0.8 1 1 1 9 LOF 1 1 1 1 1 1 1 1 0 1 0.8 1 1 1 10 LOF 1 1 1 1 1 1 0 1 1 1 0.8 1 1 1 11 LOF 1 1 1 1 1 1 1 1 1 1 1 1 1 1 12 Porosity 0 0 1 1 0.5 0 1 1 1 1 0.8 1 1 1 13 Slag 0 1 0.5 1 1 1 1 1 1 1 1 1 14 Slag 1 1 1 1 1 1 1 1 1 1 1 1 15 Slag 1 1 1 1 1 1 1 1 1 0 1 0.5 16 Porosity 1 1 1 1 1 0 1 1 0.8 1 1 1 17 Slag 1 1 1 1 1 1 1 1 1 1 1 1 18 Slag 0 0 0 0 0 0 0 0 0 0 0 1 1 1 19 Slag 1 1 1 1 1 1 1 1 1 1 1 1 1 1 Total Hits 15 7 8 18 48 16 15 17 17 18 83 18 18 36 Hit Rate Avg. 0.79 0.78 0.89 0.95 0.86 0.84 0.79 0.89 0.89 0.95 0.87 0.95 0.95 0.95 1Cells blacked out were not tested by technician Table 17. Hit (â1â)/miss (â0â) comparison for 1-inch gap criteria.
28 Specimen Details PAUT1 Conventional UT TOFD Flaw ID Flaw Type PAUT1 PAUT2 PAUT3 PAUT4 PAUT Avg UT1 UT2 UT3 UT4 UT5 UT Avg TOFD1 TOFD2 TOFD Avg 1 LOF 1 0 1 1 0.75 0 1 0 1 1 0.6 0 1 0.5 2 LOF 1 1 1 1 1 1 1 1 1 1 1 1 3 LOF 1 1 1 1 1 1 1 1 1 1 1 1 4 Toe Crack 0 0 1 1 0.5 0 0 0 0 0 0 1 0 0.5 5 Crack 1 0 0.5 1 1 1 1 1 1 1 1 1 6 Crack 1 0 0.5 1 1 1 1 1 1 1 1 1 7 Crack 1 0 0.5 1 1 1 0 1 0.8 1 1 1 8 LOF 1 1 0 1 0.75 1 0 1 1 1 0.8 1 1 1 9 LOF 1 0 0 1 0.5 0 1 1 0 0 0.4 1 1 1 10 LOF 1 0 1 1 0.75 1 0 1 1 1 0.8 1 1 1 11 LOF 1 0 1 0 0.5 1 0 1 1 1 0.8 1 1 1 12 Porosity 0 0 1 1 0.5 0 1 1 1 1 0.8 0 1 0.5 13 Slag 0 1 0.5 1 1 1 1 1 1 0 1 0.5 14 Slag 1 0 0.5 1 1 1 1 1 1 0 1 0.5 15 Slag 1 1 1 1 1 1 1 1 1 0 1 0.5 16 Porosity 1 1 1 1 1 0 1 1 0.8 1 1 1 17 Slag 1 1 1 1 1 1 1 1 1 1 0 0.5 18 Slag 0 0 0 0 0 0 0 0 0 0 0 0 1 0.5 19 Slag 1 1 1 1 1 1 1 1 1 1 1 1 1 1 Total Hits 15 2 6 13 36 14 14 15 15 16 74 13 17 30 Hit Rate Avg. 0.79 0.22 0.67 0.68 0.64 0.74 0.74 0.79 0.79 0.84 0.78 0.68 0.89 0.79 1Cells blacked out were not tested by technician Table 18. Hit (â1â)/miss (â0â) comparison for API RP 2X criteria. Location Criteria Method Planar Flaws Only Volumetric Flaws Only All Flaws 1-Inch Gap PAUT 0.79 0.50 0.68 PAUT w/o K10.2 0.76 0.45 0.64 Conventional UT 0.87 0.83 0.85 API RP 2X PAUT 0.53 0.50 0.52 PAUT w/o K10.2 0.50 0.45 0.48 Conventional UT 0.75 0.83 0.78 Table 19. Average rejection rate. or amplitude.â The rejection rate of PAUT Annex K without invoking Clause K10.2 (i.e., these flaws would instead be acceptable due to low amplitude) is included in Table 19, since PAUT flaw characterization is not always accurate and many PAUT technicians may not be comfortable with or capable of characterizing discontinuities as cracks. In fact, both of these indications were mischaracterized as cracks. Flaw 1 is a small lack of fusion indication which was acceptable per all other PAUT inspections but was rejectable to most conventional UT inspections. Flaw 14 is a slag indication and was accept- able per the other PAUT inspection but was rejectable to all conventional UT inspections. While the rejection rate for PAUT and conventional UT was similar for many individual flaws, some individual flaws had much lower rejection rates for PAUT than for conven- tional UT. There were two reasons for these differences: the flaw classification (Class AâD) based on amplitude for PAUT was often lower than conventional UT, and the reported location of the flaw was often more inaccurate for PAUT as compared with conventional UT.
29 For instance, flaws with lower classifications according to PAUT as compared to conventional UT include Flaw 1 and Flaws 12â15. These include a small LOF flaw, a group of porosity flaws, and three slag inclusions. In order to illus- trate this difference, the average reported flaw classification for these five flaws are shown in Table 20. PAUT technicians had a much higher scatter in how these five flaws were clas- sified. As shown in the table, the classification ranged from Class B, C, or D to not reporting the flaw at all. However, these flaws were typically reported as Class A for conventional UT. It is thought that this difference is mainly due to the fact that PAUT scanning in accordance with Annex K is performed by line scanning where the reported amplitude is not maximized compared with the manual, raster scanned conventional UT approach. Another very important reason for these variations is due to differences in the amplitude limits between the flaw clas- sifications for the PAUT code and the conventional UT code. In other words, the Class AâD limits for the PAUT code may not be equivalent to the Class AâD limits of the conventional UT code, even if the amplitude was maximized for each method. Another set of flaws, Flaws 5â7 and Flaw 11, shows large decreases in the rejection rate for PAUT when applying the more stringent API RP 2X flaw location criteria. This can be seen when comparing the results for these individual flaws Method No Reported Flaw Class D Class C Class B Class A PAUT 21% 21% 29% 21% 7% Conventional UT 4% 4% 4% 0% 88% Table 20. Average flaw classification for Flaw 1 and Flaw 12â15. Specimen Details PAUT Annex K1 Conventional UT Flaw ID Flaw Type PAUT1 PAUT2 PAUT3 PAUT4 PAUT Avg UT1 UT2 UT3 UT4 UT5 UT Avg 1 LOF 0 0 0 12 0.25 0 1 1 1 1 0.8 2 LOF 1 1 1 1 1 1 1 1 1 3 LOF 1 1 1 1 1 1 1 1 1 4 Toe Crack 0 0 0 0 0 0 0 1 0 1 0.4 5 Crack 1 1 1 1 1 1 1 1 1 6 Crack 1 1 1 1 1 1 1 1 1 7 Crack 1 1 1 1 1 1 1 1 1 8 LOF 1 1 1 1 1 1 0 1 1 1 0.8 9 LOF 1 1 1 1 1 1 1 1 0 1 0.8 10 LOF 1 1 1 1 1 1 0 1 1 1 0.8 11 LOF 1 1 1 1 1 1 1 1 1 1 1 12 Porosity 0 0 0 1 0.25 0 1 1 1 1 0.8 13 Slag 0 1 0.5 1 1 1 1 1 1 14 Slag 12 0 0.5 1 1 1 1 1 1 15 Slag 0 0 0 1 1 1 1 1 1 16 Porosity 1 1 1 1 1 0 1 1 0.8 17 Slag 1 1 1 1 1 1 1 1 1 18 Slag 0 0 0 0 0 0 0 0 0 0 0 19 Slag 1 1 1 1 1 1 1 1 1 1 1 Total Rejected 13 5 5 15 38 15 15 17 16 18 81 Rejection Rate 0.68 0.56 0.56 0.79 0.68 0.79 0.79 0.89 0.84 0.95 0.85 1Cells blacked out were not tested by technician 2Rejected due to crack classification rather than amplitude and length Table 21. Rejection rate (reject â1â/accept â0â) for 1-inch gap criteria.
30 Specimen Details PAUT Annex K1 Conventional UT Flaw Type PAUT1 PAUT2 PAUT3 PAUT4 PAUT Avg UT1 UT2 UT3 UT4 UT5 UT Avg 1 LOF 0 0 0 12 0.25 0 1 0 1 1 0.6 2 LOF 1 1 1 1 1 1 1 1 1 3 LOF 1 1 1 1 1 1 1 1 1 4 Toe Crack 0 0 0 0 0 0 0 0 0 0 0 5 Crack 1 0 0.5 1 1 1 1 1 1 6 Crack 1 0 0.5 1 1 1 1 1 1 7 Crack 1 0 0.5 1 1 1 0 1 0.8 8 LOF 1 1 0 1 0.75 1 0 1 1 1 0.8 9 LOF 1 0 0 1 0.5 0 1 1 0 0 0.4 10 LOF 1 0 1 1 0.75 1 0 1 1 1 0.8 11 LOF 1 0 1 0 0.5 1 0 1 1 1 0.8 12 Porosity 0 0 0 1 0.25 0 1 1 1 1 0.8 13 Slag 0 1 0.5 1 1 1 1 1 1 14 Slag 12 0 0.5 1 1 1 1 1 1 15 Slag 0 0 0 1 1 1 1 1 1 16 Porosity 1 1 1 1 1 0 1 1 0.8 17 Slag 1 1 1 1 1 1 1 1 1 18 Slag 0 0 0 0 0 0 0 0 0 0 0 19 Slag 1 1 1 1 1 1 1 1 1 1 1 Total Rejected 13 2 3 11 29 14 14 15 15 16 74 Rejection Rate 0.68 0.22 0.33 0.58 0.52 0.74 0.74 0.79 0.79 0.84 0.78 1Cells blacked out were not tested by technician 2Rejected due to crack classification rather than amplitude and length Flaw ID Table 22. Rejection rate (reject â1â/accept â0â) for API RP 2X criteria. Actual Flaw Type Reported Flaw Type Crack Planar(Non-Crack) Volumetric No Type Reported Crack 22% 44% 11% 22% LOF 21% 71% 4% 4% Porosity 25% 25% 50% 0% Slag 9% 64% 27% 0% False Calls 71% 0% 29% 0% Table 23. Flaw characterization using PAUT. in Table 21 and Table 22 for PAUT and conventional UT. These flaws include three cracks and one LOF flaw. If one were to group the results of these flaws, the average rejection rate for PAUT decreased from 100% when using the more liberal 1-inch gap criteria to 50% when using the API RP 2X criteria, while the conventional UT results only decreased from 100% to 90% using the same criteria, respectively. Therefore, poor flaw location of these planar flaws resulted in a large decrease in the rejection rate for PAUT compared with conventional UT. Table 23 shows the findings for all the PAUT flaw character- izations during the round robin testing program. Cracks were only correctly characterized 22% of the time, while many false calls were incorrectly characterized as cracks. It should also be noted that planar flaws such as cracks and lack of fusion were sometimes incorrectly characterized as volumetric dis- continuities (11% and 4%, respectively). This is a major con- cern for developing separate acceptance criteria for planar and volumetric discontinuities. If a planar flaw is mischaracterized as volumetric, it would result in an unconservative evaluation. Table 24 compares the flaw characterization results for PAUT and TOFD. It was found that the average TOFD technicians did not classify flaws as well as the average PAUT technician. While the classification of planar flaws was similar, TOFD more often incorrectly classified volumetric flaws as being planar. 3.2.4 False Calls The reported indications that were not within Â±1 inch of the total extents of the actual flaw were initially determined to be false calls. The digital RT results revealed that some of the plates included unintended weld flaws, especially near the edge. This was especially true for the FlawTech plates, which
31 seemed to have groups of sparse porosity intermittently within some of the plates, as shown in the digital RT images in Appendix D. Therefore, indications that overlap with these unintended weld flaws should not be indicated as false calls. It was found that the number of false calls which the techni- cians classified as rejectable was relatively small after removal of these unintended weld flaws which were apparent on the digital RT results as shown in Table 25. 3.2.5 API RP 2X Scoring Results API RP 2X includes formulas for scoring the performance of ultrasonic technicians during a qualification examination. These formulas evaluate the detection of indications and include a penalty for false calls. It includes suggested mini- mum performance scores for qualification examinations. The performance scores are defined as follows: 100 Formula 1 R 1 100 Formula 2 1 1 P L L L L L L c a c f = Ã = ï£«ï£ ï£¶ ï£¸ â ï£« ï£ ï£¶ ï£¸ Ã Where: P = percentage of actual reflectors correctly detected and sized R = overall rating including penalty for false calls, 0 to 100 La = length of actual reflector contained in the test plate Lc = credited length for indications that have been correctly sized and located. (Credit is given for the lesser of the reported length or actual length of the flaw.) L1 = accumulative length of all indications by the techni- cian, right or wrong Lf = accumulative length of indications above the stated disregard level where no reflector exists To be correctly sized, API RP 2X recommends that the reported dimensions be within a factor of two of true dimen- sions (i.e., one-half to twice the actual dimension). To be cor- rectly located, it is suggested that the centerline of the reported indication be within the boundary of the actual indication or within Â½ inch of the actual centerline of the indication (whichever is greater). API RP 2X suggests that minimum performance for ultra- sonic technicians be a score of 70 or above for Formula 1 and a score of 50 or above for Formula 2. These criteria were applied to the round robin data to compare the performance of PAUT and conventional UT technicians. The small slag inclusion (Flaw 18), which was not detected by any PAUT or conventional UT technicians, was not included in the analysis since it is not likely to be critical for any bridge structures. The data was analyzed two ways: (1) only accounting for correctly measuring flaw length along with flaw location and (2) accounting for correctly measuring flaw length and flaw height along with the flaw location. Due to the presence of unintended peripheral indications outside of the extent of the intended indications, technicians were given credit if the reported location and dimensions were within the required tolerances for either the main grouping (intended) of the indication or the total flaw including the unintended peripheral indications. The length of the total flaw, including unintended peripheral indications, was used for the actual length (La). Table 26 displays the results considering detected discontinuitiesâeven if they were not deemed rejectableâ and disregarding height sizing errors for PAUT and TOFD technicians. No PAUT or TOFD technicians met the mini- mum requirements for P and R, but two conventional UT technicians passed both requirements. The average score was much lower for the PAUT and TOFD technicians than the conventional UT technicians. Table 27 displays the results for Reported Flaw Type Actual Flaw Type PAUT TOFD Planar Volumetric No Type Reported Planar Volumetric No Type Reported Planar 85% 6% 9% 86% 10% 4% Volumetric 40% 60% 0% 86% 0% 14% False Calls 71% 29% 0% 100% 0% 0% Table 24. Comparison of flaw characterization of PAUT and TOFD. Technician Number of Rejectable False Calls Total Length of Rejectable False Calls PAUT1 1 3.11" PAUT2 0 0" PAUT3 0 0" PAUT4 3 2.76" UT1 0 0" UT2 1 0.45" UT3 0 0" UT4 0 0" UT5 0 0" TOFD1 2 0.95" TOFD2 0 0" Table 25. Rejectable false calls.
32 PAUT and TOFD when length and height sizing were both required to be within one-half to twice the actual dimension, along with being properly located. Due to the large inaccura- cies with height sizing and the lack of reported flaw height for TOFD results, no PAUT or TOFD technicians were close to passing the minimum performance requirements. This simple performance test highlights that there is con- siderable room for improvement to the current PAUT pro- cedures. For instance, if one were to compare the numerical values to what a student might obtain in a college engineering course, the grades for PAUT technicians are generally in the âDâ to âFâ range at best for Table 26 and at worst when con- sidering Table 27. In the opinion of the authors, this strongly highlights the need for performance testing of ultrasonic technicians in the bridge industry in order to test their abili- ties to detect (and reject) critical weld flaws. 3.2.6 Modifications to Future Scanning Procedures Based on Round Robin Results The initial round robin testing phase proved to be extremely useful and revealed much about the state of the practice as related to UT and PAUT. However, based on the results of the round robin, moving forward with development of accep- tance criteria based solely on the physical measurement of the flaw size was deemed not to be feasible with the present accuracy and reliability of flaw size measurements using the current PAUT workforce. Acceptance criteria based on flaw height and length sizing was still provided as an alternative method if the PAUT technician develops a written procedure according to specified requirements and successfully com- pletes a performance test on samples of similar material and with flaws similar to the critical flaw size. This would allow for other advanced methods such as FMC/TFM (Full Matrix Capture/Total Focusing Method) PAUT to be used in inspect- ing bridge welds provided that they can detect and reject crit- ical weld flaws if the shop, technician, engineer of record, etc., would request such methods. The conclusion was also that implementation of TOFD will be very difficult in the bridge industry. TOFD testing is spe- cialized and not readily available; in addition, it is clear that there are very few TOFD technicians that perform testing on structural welds, and it appears that there is not much interest in testing bridge welds with TOFD. TOFD technicians noted that testing butt welds where there is a plate thickness transi- tion is also difficult with TOFD due to the special attention and jigs needed to scan these welds. The focus of the research, according to the research objec- tives, was to develop acceptance criteria based on maximum amplitude and flaw length for PAUT in the AWS D1.5. From the round robin results, it appeared that PAUT, when per- formed in accordance with Annex K, shows improved detec- tion of planar flaws as compared with conventional UT since data collected showed that PAUT did not miss large flaws when using the encoded line scans and the reported flaw location only had to be within 1 inch of the actual flaw to be counted as a hit. However, it seems that additional training or performance testing is required to improve the accuracy of encoded scans. The testing also confirmed that the lack of raster scanning, to maximize the amplitude associated with a given indication, reduces the rejection rate that results from using the current version of Annex K compared with using conventional UT. In other words, flaws that are rejected today using conventional UT would not be rejected using Annex K, although there is no fitness-for-service or engineering basis for allowing this to be the case. Again, while ideally one would prefer to reject the flaw based on the size of the flaw and the type of flaw (planar vs. volumetric), the round robin testing shows that neither of these can be reliably performed with the current workforce. To ensure that flaws that would be rejected today using con- ventional UT are not accepted with PAUT using line scan- ning alone, additional raster scanning of selected indications P R P R P R TOFD1 47 39 PAUT1 68 56 UT1 67 78 TOFD2 65 82 PAUT2 44 50 UT2 75 86 Avg. TOFD 56 61 PAUT3 48 60 UT3 61 57 PAUT4 64 50 UT4 79 59 Avg. PAUT 56 54 UT5 78 48 Avg. UT 72 66 Table 26. API RP 2X results, not including height sizing error. P R P R TOFD1 17 14 PAUT1 29 24 TOFD2 29 37 PAUT2 10 11 Avg. TOFD 23 26 PAUT3 20 25 PAUT4 33 26 Avg. PAUT 23 22 Table 27. API RP 2X results for PAUT and TOFD, including height sizing.
33 coupled with line scanning using an encoder is prudent. This will help ensure that the small flaws that were accepted in the round robin when using line scanning alone would be rejected in practice. PAUT has the advantage of producing ultrasound over a range of angles, so additional raster scan- ning would not only maximize the amplitude at that angle but would do so over a wide range of angles. This should also help improve the rejection rate of volumetric flaws as well as small planar flaws. Further, use of an encoder would allow for a permanent record of the initial line scan. In its most basic form, the proposed approach for inspec- tion is to scan a weld using PAUT using an encoder along a specific index offset to be set during the development of the scan plan. If an indication is identified that meets a certain threshold, the technician would return to that location and raster scan using the PAUT probe in order to maximize the length and decibel reading. Then, the decision to accept or reject the indication would be based on an amplitude and length table. The round robin testing showed conventional UT is already being used to reject very small planar flaws (0.03â³ Ã 0.06â³). Thus, it does not seem that the proposed approach will result in increased rejection rates for small flaws (when using PAUT), which would have been acceptable under conven- tional UT and thereby resulting in an unreasonable increase in weld repairs. However, as discussed previously, while the line scanning approach is generally repeatable, the variability of the manual scanning approach after an indication is found using PAUT is problematic. 3.3 CIVA Modeling To better minimize the variability found during the round robin testing program, simulations were performed using CIVA-UT  to aid in the initial procedure development through modeling. The modeling incorporated weld flaws similar to the critical flaw sizes developed during the ana- lytical program. These flaws serve as a âlower-boundâ flaw set from which improved acceptance criteria were developed to consistently reject these flaws. As long as flaws of this size or larger are consistently rejectable, the procedure can be deemed effective at removing critical flaws from service. Therefore, the acceptance criteria are grounded in fracture mechanics but will not use flaw height measurement for evaluation. Two parametric modeling studies were performed using CIVA-UT. The first parametric study addressed the differ- ences in reference amplitude possible due to differences in acoustic properties of various steel bridge base metal grades. This study involved (1) development of benchmarked CIVA material models based on experimental test results and (2) CIVA modeling of various probes and incidence angles to develop recommendations for AWS D1.5 in order to limit the amplitude differences between the calibration block and test object. The second parametric study using CIVA evaluated the effects of variations in the amplitude response of weld flaws deemed critical per the FFS parametric study. This analysis provides a rational comparison of the amplitude from the target critical weld flaws to the acceptance cri- teria amplitude limits. These data were used to develop a rational fracture-mechanics-based PAUT inspection pro- cedure and acceptance criteria which will detect and reject critical weld flaws. The parametric models varied the plate thickness along with the flaw type, size, position, tilt, and skew of target critical weld flaws in order to compare the maximum amplitude of the indication response with the reference amplitude and the associated acceptance criteria amplitude limits. The detailed results of the two CIVA parametric studies are included in Appendix E. The results of the first paramet- ric study were used to develop recommendations for AWS D1.5 Annex K to account for differences in acoustic attenua- tion and shear wave velocity, as discussed in Section 3.4 and 3.5. The results of the second parametric study were used to develop recommendations for AWS D1.5 Annex K for scan- ning procedure requirements and acceptance criteria which will result in improved detection and rejection of critical weld flaws as discussed in Section 3.7 and 3.8. 3.4 Acoustic Attenuation 3.4.1 Experimental Testing 22.214.171.124 Experimental Testing Program Unexpected but significant observations were made during pilot experimental testing to evaluate the effects of coat- ings. After removing the paint coatings and grounding the steel surface smooth, a more than 4 dB difference in material attenuation was found between side-drilled holes (SDH) located at a depth of 0.5â³ in six different steel specimens when using a 5 MHz shear wave probe. Currently, the AWS D1.5 Bridge Welding Code procedures (for conven- tional UT and Annex K) explicitly assume that all carbon bridge steels possess the same attenuation characteristics and no correction or consideration needs to be taken dur- ing the inspection of bridge welds. As a result, the difference in attenuation found during this sequence of testing directly led to a controlled experimental evaluation of the ultra- sonic attenuation in different base metals typically used in bridge construction. The objective was to investigate the effect of different variablesâsuch as ultrasonic frequency, wave mode, material microstructure, and material acoustic velocityâon the magnitude of material attenuation.
34 Nine steel specimens were fabricated and tested using con- ventional UT and PAUT. Table 28 outlines the samples tested and their properties. Two specimensâID 36 and ID 50 from the initial testing that looked at the effects of bridge paints on PAUTâwere reused during this portion of the study. These specimens represented the least and most attenuating specimens previously tested within the field. To fully evaluate these differences in a more controlled setting, samples were cut from the girders and brought into the laboratory. ID 36 was an âhistoricalâ A36 steel; ID 50 was a modern A709 Gr. 50 steel. An additional seven modern high-performance steels (HPS) were added to further extend the evaluation. The addition of these new specimens set out to further evaluate if there were differences in the ultrasonic attenuation charac- teristics in different plates. Three of the new specimens were of the quenched and tempered (QT) variety at the mill, while four of the new specimens were produced using the thermo- mechanical control process (TMCP). A review of previous literature [39â41] has confirmed that chemical composition, grain size, and microstructure have all been found to affect the acoustic properties and propa- gation of sound through material. The chemical composi- tion of each specimen was obtained and found to meet the requirements of its respective ASTM steel standard. For each specimen, a metallurgical analysis of the grain size and microstructure was also performed by an outside consul- tant. Analysis showed the following: Figure 7 shows the grain structure perpendicular to the rolled direction for each speci- men at the central region of the plate magnified at 100X with Nital etchant. â¢ Specimen 36 consisted of a WidmanstÃ¤tten pattern of ferrite and pearlite. â¢ Specimen 50 consisted of ferrite and pearlite. â¢ Specimen 70 had a general structure of fine acicular ferrite with small spherical carbides, but also visible were bands of ferrite and low-carbon martensite and bainite. â¢ Specimens 101 and 102 consisted of QT martensite. The TMCP specimens all had a variation in grain structure near the surface in comparison with the central regions. â¢ Specimen TMCP 1 had acicular ferrite with elongated pearlite and long bands of pearlite in the central region. On the near surface region, a fine acicular ferrite and short bands of pearlite were seen. â¢ Specimen TMCP 2 had elongated ferrite with bands of pearlite and bainite in the central region. On the near sur- face region, elongated ferrite and short bands of pearlite and bainite existed. â¢ Specimen TMCP 3 had a fine acicular ferrite with patches of pearlite in the central region. On the near surface region, a more refined structure of fine acicular ferrite and patches of pearlite were seen. â¢ Specimen TMCP 2 was further analyzed parallel to the rolled direction. Parallel to rolled, the central region and near surface regions both consisted of elongated ferrite with bands of pearlite and bainite. Figure 7 shows the grain structure perpendicular to the rolled direction for each specimen at the central region of the plate magnified at 100X with Nital etchant. Grain size measurements were made in accordance with ASTM E112-13 Standard Test Methods for Determining Aver- age Grain Size. Per ASTM E112, grain size measurements can be conducted numerous ways, but all methods include counting the number of grains or number of grain bound- aries along a specified line within a known area. A table is provided in ASTM E112 to rate the grain size from 00 up to 14.0, with 00 representing the largest average grain size and 14.0 the smallest average grain size . Table 29 presents the grain sizes measured for the group of specimens tested for this project. It should be noted that in this table, for Speci- mens 101 and 102, the prior austenite grain size is measured and presented. In this case, the prior austenite grain size was that of the steel before quenching and tempering occurred. ID Steel Properties Fabrication Year Thickness (in) Width (in) 36 A36 1973(1) 1.25 1.87 50 A709 Gr50 2013 1.25 1.87 70 HPS 70W QT 2015 1.50 1.87 101 HPS 100W QT circa 2000s 2.00 1.87 102 HPS 100W QT circa 2000s 1.50 1.87 TMCP 1 HPS 70W TMCP 2009 1.57 1.87 TMCP 2(2) HPS 70W TMCP 2014 1.25 1.87 TMCP 3 HPS 70W TMCP 2011 2.00 1.87 (1) Date the bridge was put into service (2) Two specimens, one in the rolled direction and one in the transverse to rolled direction, were fabricated with this steel plate Table 28. Steel specimens.
35 36 50 70 (QT) 101 (QT) 102 (QT) TMCP 1 TMCP 2 TMCP 3 Figure 7. Microstructure for specimens at 100X (perpendicular to rolled direction).
36 The variability in ultrasonic inspection of NGI-ESW welds was then assessed following the evaluation of base metal. Unlike the consistent microstructure of base metal, welding produces different zones of varying grain structures. From the electroslag welding process, the HAZ consists of two grain structure zones. The portion of the HAZ bordering the base metal is comprised of fine grains and the inner portion is comprised of coarse grains. The weld may have an additional two or three zones itself of coarse columnar and/or equiaxed grains . Speci- mens were fabricated to facilitate the comparison of attenuation among base metal, HAZs, and weld metal. The electroslag weld samples were donated by the Federal Highway Administration (FHWA) and supplied by two different fabricators. As shown in Figure 8, the individual grain structure of the NGI-ESW welds varied considerably for these specimens. Table 30 outlines the details of the specimens and their material properties. The specific details of the experimental testing program which measured the acoustic attenuation of base metal and NGI-ESW are given in Appendix F. The results of these tests are summarized following. 126.96.36.199 Summary of Experimental Findings During the portion of the research that was intended to investigate the effects of coatings on scanning, major dif- ferences were observed in the attenuation characteristics in different steel plates. In has been generally assumed by the industry that the attenuation characteristics of the test block can be used to represent those of the test compo- nent. However, upon closer study and laboratory testing, it was concluded that this assumption is not appropriate in all cases. Some steels will have much different attenuation characteristics than that of the typical test block. Clearly, if the test piece (and weld) are more attenuating than the test block, the amount of sound returned from a reflector will be less than the same reflector in the test block. The result is obvious: unacceptable flaws will be accepted. The converse is also true if the test block is more attenuating than the test component. One of the issues that is specific to PAUT (and Annex K) is the fact that the effect of attenuation error becomes more pronounced at higher transducer frequencies. Cur- rently, Annex K permits the transducer frequencies rang- ing from 1 MHz to 6 MHz, while D1.5 Clause 6 limits the frequency to 2.25 MHz for conventional UT. Interestingly, other specifications openly acknowledge this issue and require technicians to calibrate using a test block made of a material with the same acoustic properties as the piece being inspected unless it is confirmed that the attenuation characteristics are comparable as previously stated. This would, of course, include attenuation associated with the weld metal. In lieu of requiring such calibrations, the experimental results suggest that for most cases, these effects in base metal will generally become small and possibly negligible when the transducer frequency is less than or equal to 2.25 MHz. The trade-off is that lower-frequency transducers have less resolution, but the CIVA and experimental studies suggest this will not adversely affect the overall reliability. ID(1) Grain Size ASTM Grains per Unit Area (in2) at 100X(2) 36 ASTM 2-1/2 2.83 50 ASTM 7 64.0 70 (QT) ASTM 10 512.0 101 (QT) ASTM 8 128.0 102 (QT) ASTM 8 128.0 TMCP 1 ASTM 11 (central)/ ASTM 11 (surface) 1024.0 TMCP 2 ASTM 11 (central)/ ASTM 11 (surface) 1024.0 TMCP 2 (parallel to roll) ASTM 10 (central)/ ASTM 10 (surface) 512.0 TMCP 3 ASTM 8 (central)/ ASTM 12 (surface) 128.0/ 2896.3 (1)Perpendicular to rolled direction unless noted otherwise (2)ASTM E112, 2013 Table 29. ASTM grain size classifications. Figure 8. Variation in NGI-ESW grain structure.
37 While limiting the probe frequency may address the attenuation issues associated with base metal, the attenua- tion characteristics of the weld itself, in particular attenua- tion associated with NGI-ESW, was found to be quite high. In this case, use of a lower frequency, such as 2.25 MHz, still resulted in a large loss of amplitude from sound pass- ing through the weld metal. In other industries, when the attenuation of base metal or weld metal is suspected as being different than the calibration block, a transfer cor- rection is formulated through a pitch-catch scanning pro- cedure to measure the actual attenuation. It is clear that the material attenuation issues observed must be accounted for in some way. 3.4.2 AWS Recommendations for Acoustic Attenuation Differences in acoustic attenuation for common grades of base metals were noted during the experimental testing. The differences in attenuation were significantly greater for 5 MHz probes than 2.25 MHz probes, which resulted in large amplitude variations depending on the differences between the calibration block and the test object. CIVA models repre- senting various grades of bridge base metals were developed and benchmarked to the experimental test results. The CIVA models were used to evaluate the amplitude differences that would result from testing different grades of base metals after performing calibration on a typical calibration block with AISI 1018 steel base metal. Based on this analysis, proposed modifications to AWS D1.5 Annex K were developed in order to improve the calibra- tion requirements to account for the differences in acoustic attenuation. A marked-up version of Annex K and associ- ated commentary with the recommended changes is given in Appendix G. These modifications include that a 2.25 MHz probe should be used for plate thicknesses exceeding 0.5â³ unless the attenuation of the calibration block and test object is similar or the differences are accounted for through the use of a transfer correction. A transfer correction accounts for the differences in attenuation and coupling losses due to surface roughness between the calibration block and the test object. Transfer correction is referred to in many UT references [12, 13, 19, 37, and 44] specifically as a method to account for acoustic differences between the calibration block and the test object. Two methods may be used to perform a transfer correction: 1. Fabricate a block with the same reference reflector includ- ing type, size, and depth as the calibration block, and note the differences in the signal response between the calibra- tion block and the test object. The sound path for this measurement should correspond to the longest sound path used for the inspection. 2. Perform pitchâcatch measurement on the calibration block and the test object over two sound paths to develop the relationship between attenuation and sound path that can be used to correct the amplitude from the calibration block. Since the first option involves fabrication of a new calibra- tion block, the second option is often much more suitable for checking whether any amplitude correction is necessary between an existing calibration block and the test object. In this method, two probes are used, with one acting as a trans- mitter and the other as a receiver. The sound is skipped off of the backwall of the calibration block and test object in a single-V and double-V path (i.e., one skip and two skips off of the backwall), as shown in Figure 9, and the amplitude is measured at each location with the same transducer settings. These amplitude measurements are then plotted against the sound path and lines are drawn through the measurements corresponding to the calibration block and the test object, as shown in Figure 10. The difference in amplitude at the maximum inspection sound path can be directly obtained from these lines. If the calibration block line is above the test object line, then the calibration block has less attenuation than the test object and additional gain must be added to the inspection of the test object. If the calibration block line is below the test object line, then the calibration block has more attenuation than the test object. While gain may be removed from the inspec- tion of the test object to correct for the difference, in this case, this could result in removal of too much gain at shorter sound paths. The thickness of the calibration block and the test object does not need to match, but the amplitude measurements must Specimen ID Fabricator Base Metal Fabrication Year Thickness (in) Side A Side B P1 Fabricator A HPS 70W (QT) HPS 70W (QT) 2015 2.0 P2 Fabricator A 50W 50W 2015 2.0 P3 Fabricator A 50W HPS 70W (QT) 2015 2.0 P4 Fabricator B HPS 70W HPS 70W 2013 2.0 Table 30. NGI-ESW specimen.
38 be made in the far field in order to ensure that the change in amplitude is due to attenuation. API RP 2X requires that the single-V-path-measurement sound path exceeds 4 inches. If the single-V-path measurement is within this sound path, it is recommended that additional skip distances be used to obtain longer sound paths. While the inspection will be performed with PAUT probes, use of conventional UT probes with similar frequency and aperture that will be used for the PAUT inspection will be adequate to account for the attenuation. It is recommended that the conventional UT wedge be chosen to produce the largest incidence angle used in the PAUT inspection, since this will increase the sound path and is more sensitive to amplitude differences due to velocity variations. During experimental testing, 2.25 MHz and 5 MHz conventional UT and PAUT probes were used for attenuation testing, and the results of PAUT and conventional UT were very similar for the same incidence angle, frequency, and aperture. According to ISO 17640 , no correction is required if the greatest difference in amplitude between the calibration block and the test object at the longest sound path is 2 dB or less. This code requires correction for amplitude differences greater than 2 dB but smaller than or equal to 12 dB. If the amplitude difference exceeds 12 dB, this code requires that the reason for this difference shall be considered or the scan- ning surface reconditioned. In order to make the transfer correction when TCG is used, either (1) the entire TCG curve is corrected by modifying the reference gain, which would offset the entire TCG curve up/down by the same amount, or (2) separate corrections could be performed manually to each TCG point for each focal law. The first method is very simple, but may result in overcorrection at short sound paths. For instance, shifting the entire amplitude by the transfer correction at the longest sound path in Figure 10 would result in the test object amplitude being overestimated at shorter sound paths. This would be conservative, but the oppo- site is true if the transfer correction resulted in removal ampli- tude (i.e., too much amplitude would be removed at short sound paths resulting in lower sensitivity than reference). Therefore, in order to avoid unconservative corrections, Annex K should not allow for the removal of amplitude unless consideration is given for the amplitude difference at shorter sound paths. The second method of TCG correction for attenuation differences described above involves manual adjustment to each TCG point for each focal law. While this would result in accurate correction over all sound paths, it would be a time-consuming process that would require input of at least 93 corrections considering 31 focal laws for a 40Â°â70Â° inci- dence angle range with 1Â° increment and a 3-point TCG. Therefore, unless automated in some way, this method is more sensitive to human error and unlikely to be practical for shop application. The research team recommends that AWS Annex K be modified to require verification of the differences in acoustic properties between the test object and calibration block. If the amplitude difference exceeds 2 dB at the longest sound path and the calibration block is less attenuating than the test object, a correction for this difference is required. Guidance for four methods of correction has been produced for the AWS Annex K commentary. They involve the following: 1. Using a lower frequency probe, such as 2.25 MHz. 2. Using a calibration block that better matches the test object and results in less than a 2 dB amplitude difference at the longest sound path. 3. Reconditioning the scanning surface of the test object in order to better match the calibration block. 4. Adding gain equal to the difference between the calibra- tion block and the test object at the longest sound path. It is noted that this may be overly conservative at shorter sound paths and it is not recommended to be performed for differences greater than 12 dB. It is very likely that the base material on each side of a butt splice will be from different mills, grades, or heats. When this is the case, a separate transfer correction is required on each plate at the weld to correct for differences in acoustic properties. It is possible that the equipment settings may differ for inspection of the same weld depending on which side of the weld is being inspected. Figure 9. Transfer correction probe locations. Figure 10. Transfer correction amplitude.
39 The research team recommends that additional requirements be provided to account for the significant loss of amplitude found when sound propagates through coarse grained NGI- ESW welds. These requirements include verifying the ampli- tude and location of a 1.5 mm (0.06â³) diameter SDH in a full- scale mockup of the weld. (Note, full-scale means simply full thickness. The specimen need only be a few inches wide.) The reflector would be required to be placed in a location that will maximize the sound path traveling through the weld metal. This recommendation will allow for an amplitude correction to be determined on a case-by-case basis in order to account for the high attenuation found when the sound beam passes through the weld metal of some NGI-ESW welds. Recommendations are also provided that NGI-ESW welds shall be inspected from the outside of each fusion face. This recommendation, pro- vided in the section that covers scanning coverage of welds, will ensure that detection of fusion flaws will not depend on sound passing through the entire weld volume. 3.5 Shear Wave Velocity 3.5.1 Experimental Testing When inspecting a component using shear waves, sound enters the material at some defined angle, usually ranging between 45Â° and 70Â°. Conventional UT utilizes a single angleâ e.g., 70Â°âwhereas PAUT utilizes a range of angles, typically from 45Â° through 70Â°, or any specific group thereof (e.g., 55Â° through 65Â°). An accurate estimate of the velocity through the thickness is required since it is needed (using Snellâs Law) to predict the refraction angle of the sound entering the steel. For example, in conventional UT, when one uses a 70Â° wedge, the geometry of the wedge is based on the assumed velocity of 0.127 in/Âµs (3,230 m/s) for shear waves to ensure the sound is traveling through the steel at 70Â°. Unfortunately, even a very small difference has a large impact on this calculation, espe- cially for higher beam angles. According to research performed in Japan and the CIVA simulation performed by the RT, dif- ferences as little as 2% or even less need to be accounted for, as the sound will enter at a different angle than assumed. The result is twofold. First, a given flaw will be mislocated since the assumed geometry is not correct (e.g., the beam is not at 70Â°, but rather, something greater). Second, since the actual sound path that is needed to maximize the reflection is differ- ent (and longer) and much of the sound energy is not actu- ally hitting the reflector due to refraction into surface waves, the resulting returning sound associated with a given reflec- tor will have lower amplitude. From the perspective of the technician, a given reflector will appear to be at the incorrect location and to reflect less sound. The most pronounced effects are observed at steeper inci- dence or sound beam angles, such as 60Â° and greater, where a significant portion of the sound beam begins to reach the critical angle at which the shear wave converts to a surface wave (to be discussed in detail later). Note, when the term âincidence angleâ or âsound beamâ is used, it is simply refer- ring to primary angleâfor example, 70Â°âthough some of the beam naturally spreads above and below, similar to how a flashlight beam spreads out. The result is that someâor in some cases, mostâof the sound never actually enters the plate as a shear wave but rather mode coverts and travels along the surface of the plate at the steeper incidence angles. This has been confirmed by both experimental testing and CIVA simulations performed by the RT. Again, this is most pronounced at beam angles greater than 60Â°. If a constant index is used, there may be very little sound coverage in the region of interest due to the beam entering at a different angle than assumed, and if there is some sound reflected from a flaw, the amplitude is reduced, sometimes significantly. To investigate this further, the acoustic shear wave velocity in a number of TMCP HPS 70W plates was measured using an Electro Magnetic Acoustic Transducer (EMAT). Using a shear wave, acoustic velocity was measured in the rolled and transverse to the rolled direction. The acoustic velocities and anisotropic ratios are listed in Table 31. The anisotropic ratio is the measured difference between the acoustic velocities in the rolled direction (RD) and transverse to rolled (â¥RD) direc- tion. The amount of anisotropy is defined by the anisotropic ratio, which is the ratio of the velocity in the rolled direction to the transverse to rolled direction, as shown below. % 1 100( ) = â¥ â ï£« ï£ ï£¬ ï£¶ ï£¸ ï£· ÃAnisotropic Ratio Velocity in RD Velocity in RD ID Acoustic Velocity (in/Âµsec) Anisotropic Ratio (%) Rolled Direction Transverse to Rolled Direction 70 (QT) 0.1271 0.1272 0.05 101 (QT) 0.1272 0.1274 0.15 TMCP 1 0.1328 0.1266 4.95 TMCP 2 0.1304 0.1241 5.07 TMCP 3 0.1293 0.1255 3.09 Table 31. Shear wave acoustic velocities of steel specimens.
40 The QT specimensâspecimens 70 and 101âhave a very low anisotropic ratio. In comparison, it is clear that all of the TMCP specimens demonstrate high anisotropic ratio. Plates with a high anisotropic ratio(1â2% or greater) are considered an acoustically anisotropic material . All TMCP plates tested demonstrated significant acoustic anisotropy. In other words, there is directional variation in velocity in the as-rolled and cross-rolled directions. The average velocity through the thickness differed by as much as 5%, which, as noted, at this level has a significant effect on the refraction angle. After the observation, an additional literature review was performed to establish if other researchers had observed this same behavior in TMCP plates. Previous research in Japan has documented this effect for TMCP plates [45, 46]. The Japanese JIS Z 3060 UT code  specifies that the shear wave velocity be measured in the direction in which the inspection will occur in the test object and compared to the calibration block. Depending on the ratio of the shear wave velocity in the calibration block and test specimen, either a new calibration block is required with a velocity that matches the test object more closely or restrictions are placed on the incidence angle that may be used in the inspection. All three TMCP specimens used during this study would have exceeded the limits that would have corrective action as acoustically anisotropic using the Japanese criteria. It is noteworthy that this is also an issue for conventional UT, as the 70Â° probe is the prescribed transducer for most applications. As will be shown for some TMCP plates, due to the refraction issues, the 70Â° probe can be quite ineffec- tive in detecting flaws. This is concerning and, though out of the scope of this project, this will also need to be addressed in the current AWS D1.5. Other specifications require the acoustic velocity of the plate be measured and used to deter- mine the true angle of refraction with respect to the assumed scanning angle. When a significant difference between mea- sured and assumed velocity is observed (1% or greater is considered significant), the scanning angle is often limited to no greater than 60Â°. For brevity, the specific details of the experimental testing program which measured the shear wave velocity of base metal are given in Appendix F. Additional velocity measurements were made on other grades of base metals commonly used in bridges using a normal incidence shear wave probe, shown in Table 32. Machined specimens were not fabricated from these plates, so a correlation of the shear wave velocity to the amplitude response of a SDH is not provided in these cases. These data are simply provided to aid in determining whether other grades are susceptible to velocity variations. The A709-HPS 50W and A709-HPS 70W specimens both had a shear wave velocity in the rolled direction that would differ from a stan- dard velocity of 0.127 in/Âµs by more than 1%, along with aniso- tropic ratios greater than 1%. Therefore, both of these steels would be susceptible to calibration issues due to mismatch of the calibration block and test object shear wave velocity as well as beam splitting when testing along an oblique orientation to the rolled direction due to acoustic anisotropy. 3.5.2 AWS Recommendation for Shear Wave Velocity 188.8.131.52 Rolled Direction or Transverse to Rolled Direction Differences in measured shear wave velocity for common grades of base metals were noted during the experimen- tal testing. For TMCP processed steels tested, the velocity increased in the rolled direction and decreased in the trans- verse to rolled direction. These materials are considered acoustically anisotropic. Since the incidence angle is related to the ratio of the wedge velocity to the steel velocity by Snellâs Law, differences in the incidence angle result in errors in properly locating flaws as well as affecting the amplitude reflected from the indications (i.e., a reduction in amplitude). Experimental testing found that the amplitude was sig- nificantly decreased at high incidence angles (60Â° or greater) when scanning was performed in a direction of increased velocity. CIVA models were developed that represented one of the TMCP processed steel samples from the experimental test- ing with a shear wave velocity of 0.133 in/Âµs (4.5% increase compared to the calibration block). The CIVA models showed that the difference in amplitude of the TMCP steel to the 1018 calibration block would be â¼2 dB at a 7â³ depth with 45Â° inci- dence angle, would be â¼2 dB at a 1â³ depth and â¼4 dB at a 7â³ depth with 60Â° incidence angle, and would exceeded 2 dB Grade Acoustic Velocity (in/Âµsec) Anisotropic Ratio (%) Rolled Direction Transverse to Rolled Direction A709-50 0.1281 0.1280 0. 05 A709-50W 0.1279 0.1275 0. 36 A709-HPS 50W 0.1287 0.1270 1.29 A709-HPS 70W (TMCP) 0.1294 0.1262 2.54 Table 32. Shear wave acoustic velocities of other bridge base materials.
41 for all sound paths with the 70Â° incidence angle. Therefore, while limiting the incidence angle range to 40Â°â60Â° will lessen the impact from the changes in the shear wave velocity, it will not limit the amplitude differences to 2 dB or less over all possible sound paths. Additional CIVA analysis was performed in order to identify the limiting velocity difference compared to the standard steel velocity of 0.127 in/Âµs (3,230 m/s) which would result in an amplitude difference of 2 dB or less over a sound path cover- ing 7â³ depth for the 60Â° and 70Â° incidence angles. A velocity increase of 2.5% compared to the standard velocity resulted in an amplitude difference of 2 dB or less across the 40Â°â60Â° incidence angles. Therefore, assuming a shear wave velocity of 0.127â0.128 in/Âµs for the calibration block, which is typi- cal for most steels, the 40Â°â60Â° incidence angles should be appropriate for plates with velocity of 0.130â0.131 in/Âµs. Two of the three TMCP samples from the experimental test- ing had a shear wave velocity of 0.130 in/Âµs or less, and the amplitude difference for these two samples compared to the Grade 50 sample were 2 dB or less at the 60Â° incidence angle. This correlates well with the CIVA results regarding a recom- mended limit of 2.5% velocity difference for inspection up to a 60Â° incidence angle. Limiting the incidence angle range to 40Â°â60Â° for inspection of TMCP processed steels will result in much less amplitude error than using 40Â°â70Â° and may limit the amplitude error to 2 dB or less for many TMCP pro- cessed steels, however there may be certain heats of TMCP processed steels where the amplitude loss may exceed 2 dB at the 60Â° incidence angle. A velocity increase of 1.0% compared to the calibration block resulted in an amplitude difference of 2 dB or less across the 40Â°â70Â° incidence angles. Assuming a shear wave velocity of 0.127â0.128 in/Âµs, the 40Â°â70Â° incidence angles should be appropriate for plates with velocity of 0.128â0.129 in/Âµs. One of the three TMCP samples from the experimental testing had a shear wave velocity of 0.1293 in/Âµs, and the amplitude difference for this sample compared to the Grade 50 sample (measured shear wave velocity of 0.1274 in/Âµs) was 2â4 dB at the 70Â° incidence angle. The velocity of this TMCP sample was 1.5% greater than the Grade 50 sample, so a slight increase in sound loss exceeding 2 dB is expected at the 70Â° incidence angle. While a 1.0% limit on the velocity difference between the calibration block and the test object seems reasonable to limit the amplitude difference to 2 dB or less over the 40Â°â70Â° incidence angle range, the actual amplitude difference may exceed 2 dB in some cases, especially when also including any differences in material attenuation, as discussed previously in Section 3.4. A marked-up version of Annex K and associated commen- tary with the recommended changes is given in Appendix G. The research team is recommending that the acoustic properties of the test object and calibration block be verified to be within certain tolerances. It is recommended that the shear wave velocity in the direction of sound propagation in the test object and the calibration block be required to be within Â±2.5% of each other. When the difference exceeds this amount, it is recommended that a new calibration block with a veloc- ity within 2.5% of the test object be fabricated or otherwise acquired. When the difference in shear wave velocity exceeds Â±1%, the research team recommends that the incidence angle be limited from 40Â°â60Â°. If the velocity of the test object and calibration block is measured and found to be within Â±1% of each other, a 40Â°â70Â° incidence angle range may be used. Since most steels have a velocity of 0.127â0.128 in/Âµs, these steels would all be able to use the same calibration block, assuming that attenuation is properly accounted for. JIS Z 3060:2015  includes three different methods to directly or indirectly measure the shear wave velocity of the test object and calibration block: 1. Fabricate a block with a reference reflector from the test object material. Calculate the incidence angle of the test object and the calibration block by using measurements of the physical distance of the reflector from the index point and the depth of the reflector. 2. Perform pitchâcatch measurement on the test object and calibration block in a single-V path, and calculate the inci- dence angle by using measurements of the physical distance between the transducer index points and the thickness of the plate. 3. Directly measure the shear wave velocity of the plate by using a normal incidence shear wave probe to measure backwall signals. The shear wave velocity of the calibra- tion block and test object can either be directly computed using a calibration feature with successive shear wave backwall skip signals (similar to velocity measurements using a normal incidence longitudinal wave probe), or the velocity ratio between the calibration block and test object can be computed using the known thickness and measured shear wave backwall signals of each plate. If the incidence angle is measured, the velocity ratio can be calculated using Snellâs Law as follows: sin sin V V test object cal block test object cal block Î¸ Î¸ = Assuming that the calibration block incidence angles are truly 60Â° and 70Â° (i.e., velocity of calibration block matches the assumed velocity), the incidence angles measured in the test object are 62.6Â° and 74.4Â° for a 2.5% increase in velocity and 61Â° and 71.6Â° for a 1.0% increase in velocity. All of these methods rely on the orientation of the probe used to measure the velocity or incidence angle matching the
42 orientation of the probe that will be used during scanning. For instance, if the PAUT line scanning will be performed with the probe oriented along the rolled direction, then the measure- ment of the shear wave velocity or incidence angle shall be performed in the same direction. As an example, checking the depth and index distance from the corner trap signal, which is produced from scanning the edge of the plate transverse to the rolled direction, cannot be used to verify the velocity along the rolled direction. Fabrication of a calibration block is also sensitive to the orientation of the plate in relation to the orientation of the scanning direction. In fact, the direction of sound propagation in regards to the through-thickness direction is also critical. For instance, typically calibration blocks are cut from material and flipped so that the through-thickness direction forms the width of the calibration block, as shown for the block on the left side of Figure 11. This is performed in order to provide for greater depths of the SDH reflectors and avoid the need for very thick plates. Due to the complex and layered grain structure found in TMCP steels, the velocity at the surface of the plate will differ from the velocity in the middle of the plate. Therefore, calibration blocks for TMCP steels must have the correct scanning surface and scanning direction (e.g., through- thickness and rolled direction) compared with the test object, as shown on the right side of Figure 11. 184.108.40.206 Oblique to Rolled Direction Since the TMCP steels have very different velocity in the rolled direction and the transverse to the rolled direction, they are considered acoustic anisotropic materials. When scanning is performed oblique to the rolled direction (i.e., not parallel or perpendicular), the beam splits into two waves traveling at different velocities, as shown in Figure 12 for the normal incidence shear wave probe. Since the beam is split, the amplitude is effectively halved, and it is possible for two indications to appear on the screen, as shown in Figure 13. Experimental testing was performed on Specimen TMCP 2 plate, cut at a 45Â° orientation to the rolled direction with 1â16â³ dia. SDHs at 0.6â³ and 1.0â³ depth. The machined speci- men dimensions are similar to the other TMCP specimens. The amplitude difference for the 45Â°, 60Â°, and 70Â° incidence angles on the 45Â° orientation to rolled direction TMCP block compared to the Grade 50 block is shown in Figure 14 for the 2.25 MHz Zetec PAUT probe (AXL-2.25 MHz with AXL-55SW wedge with 16 active elements) and an active aperture of 16 mm Ã 20 mm. The Grade 50 block tradition- ally served as reference for the other TMCP testing. The results from testing using the same heat of steel but in the rolled and transverse to rolled orientation are shown in Figure 15. This testing was performed using a 2.25 MHz conventional UT probe, which had a similar aperture to the PAUT probe. Comparing these figures, the 45Â° orientation to the rolled direction block had more loss of amplitude at the 45Â° and 60Â° incidence angles than the rolled and transverse to rolled blocks. This is likely due to the beam-splitting effect for the 45Â° orientation block. The loss of amplitude in the 45Â° orientation block at the 70Â° incidence angle was comparable Scanning Surface Rotated Figure 11. Calibration block rotation. 1st skip 2nd skip 3rd skip Figure 12. Normal incidence shear wave probe at 45Ã§ angle.
43 Figure 13. 0.6â deep SDH in 45Ã§ TMCP plate. 0.6" 1.0" 0.6" 1.0" 0.6" 1.0" -4.0 -2.0 0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 dB Search Angle 45áµ 60áµ 70áµ Reference Figure 14. 45Ã§ orientation TMCP plate amplitude difference for 2.25 MHz PAUT probe. 0.6" 1.0" 0.6" 1.0" 0.6" 1.0" -4.0 -2.0 0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 dB Search Angle 45áµ 60áµ 70áµ 45Â° 60Â° 70Â° TMCP 2 Roll -rollx x + Figure 15. Rolled and transverse to rolled orientation TMCP plate amplitude difference for 2.25 MHz conventional UT.
44 to the results of the rolled direction for the 1â³ deep hole but was lower than the rolled direction for the 0.6â³ hole. Based on these results, the research team recommends addi- tional requirements for acoustically anisotropic materials. It recommends that these requirements be placed on materials that have an anisotropic ratio (ratio of velocity in the rolled and transverse to rolled direction) of 1% or greater. These rec- ommendations include a requirement that the incidence angle range be limited to 40Â°â60Â° for scanning of acoustic anisotro- pic plates (e.g., TMCP) at an oblique orientation to the rolled direction. It is also recommended that 4 dB required to be added to the reference sensitivity to account for the sound loss due to the beam splitting. While normal incidence shear wave probes can be used to quickly ascertain whether a plate has acoustic anisotropic behavior (i.e., the velocity is different in the rolled and trans- verse to rolled directions) by rotating the probe from polariza- tion in the rolled direction to the transverse to rolled direction, caution may be necessary to use it for measuring the velocity in the oblique orientation. For scanning along oblique orien- tations in acoustic anisotropic materials, JIS Z 3060 requires that the velocity be accounted for by measuring the incidence angle using an angle beam probe(s) rather than directly mea- suring the velocity with the normal incidence shear wave probe. While the technical reason for this requirement is not explained in JIS Z 3060, it is likely due to the fact that two dif- ferent measurements are possible with the normal incidence shear wave probe, which could result in miscalculations. Mea- surement of the incidence angle through either fabrication of a reference standard or using the pitchâcatch method results in measurement of the true refraction angle at the maximum amplitude indication. There may still be two indications apparent on the screen, but the maximum amplitude indica- tion would be used for measurement of the incidence angle. The JIS code does not require additional amplitude be added for scanning in the oblique orientation, but it does require that the calibration standard be acoustically equivalent to the test object with velocity within Â±2% and sensitivity correction within Â±2 dB compared to the test block along with additional limits on the incidence angle depending on the ratio of the velocity of the test object to the calibration block in the orien- tation of the scanning direction. Based on the experimental results, additional requirements are recommended for acoustically anisotropic materials. It is recommended that all materials with an anisotropic ratio (a ratio of velocity in the rolled and transverse to rolled direc- tion) of 1% or greater be defined as acoustically anisotropic in Annex K. Based on the experimental test results, additional requirements on the incidence angle and addition of ampli- tude are recommended for scanning of acoustically aniso- tropic materials in the oblique orientation with respect to the rolled direction. It is recommended to limit the incidence angle to 40Â°â60Â° for scanning of acoustic anisotropic plates at an oblique orientation to the rolled direction. It is also rec- ommended that 4 dB be added to the reference sensitivity to account for the sound loss due to the beam splitting. These requirements do not require the velocity or beam incidence angle be measured in the oblique orientation. There- fore, while guidance is provided in the commentary that cau- tion must be taken when measuring the velocity of the plate in the oblique orientation using a normal incidence shear wave probe, this measurement is not necessary. Rather, acous- tic anisotropic materials are identified using the shear wave velocity measurements in the rolled and transverse to rolled directions and checking against the 1% limit on difference. 3.6 Calibration Blocks While correction for differences in attenuation and velocity between the calibration block and the test object may be pro- vided through the use of certain probes and incidence angles or through the use of a transfer correction, it can also be provided by using a calibration block that is acoustically equivalent to the test object. Thus, fabrication of additional calibration blocks that are acoustically equivalent to the various steels commonly used in a shop may be prudent. As discussed pre- viously, the orientation and scanning face of the calibration block is also critical for acoustic anisotropic material such as TMCP processed steel. For this reason, guidance is provided on recommendations for proper calibration block design. Due to the necessary sensitivity for flaw detection and rejection dis- cussed in the following sections, it is recommended that the 1.5 mm (0.06â³) diameter SDH reference standard reflector be used for calibration and setting reference amplitude. This is the same reference standard reflector that is currently being used in AWS Annex K; no changes to the code in regards to the reference standard reflector are necessary. Calibration blocks can be machined from a strip of steel removed from a plate with 0.06â³ (1.5 mm) diameter SDHs drilled through the width. In order to provide enough sound paths for TCG calibration, it is recommended that one SDH be placed near the surface of the plate and the other placed either in the center or third-point, depending on the plate thickness. This provides for many possible TCG points, as shown in Figure 16. It is recommended that the current mini- mum of 3-point TCG be carried forward in the new version of Annex K, but nine or more TCG points are possible from this block. The hole near the surface should be placed at least 0.2â³ away from the surface in order to distinguish the first leg indication from the second leg indication. The hole 0.2â³ away (minimum) from the surface should not be scanned with a skip off of the near surface since the small ligament between the edge of the hole and surface of the plate can result in increased amplitude. This is similar to a corner trap from a
45 surface breaking flaw. CIVA analysis was performed to deter- mine the minimum ligament for the SDH from the surface of the plate in order to use it for skipping off of the backwall. A 0.5â³ minimum depth from the surface of the plate should be provided in order to skip from the near surface. It is recommended that this limit be included in the calibration block geometry requirements in Annex K. The width of the block should be large enough to accommodate the beam spread without funneling the sound beam along the calibration block at long sound paths. The JIS Z 3060 code includes the following equation for estimat- ing a suggested minimum width of the calibration block: 2W S D> Ã l Ã Where: W: width of the calibration block l: wavelength S: maximum sound path to be used D: width of the transducer Using typical values in this equationâfor example, a 11.7â³ sound path, which would represent a 4â³ depth at 70Â° incidence angleâthe minimum width of the block would be 2.1â³ for a 2.25 MHz probe with a 16 mm width and 1.7â³ for either a 2.25 MHz probe with a 20 mm width or a 5 MHz probe with a 10 mm width. Therefore, a 2â³ width will likely be adequate for most blocks, although a narrower block may be appropriate if it is used over shorter sound paths. It is recommended that Annex K include the requirement that the calibration block be of sufficient width to allow for adequate beam spread at the longest sound path used for calibration. A marked-up version of Annex K and associ- ated commentary with the recommended changes is given in Appendix G. The JIS code equation is also provided in the commentary as a recommended minimum width of the calibration block. Finally, the length of the block should be adequate to accommodate multiple skips for TCG calibration and pitchâ catch comparison to the test object. It is recommended that the spacing of the two holes be wide enough to provide for the double-V-path in the pitchâcatch setup for the 70Â° inci- dence angle. This will also provide adequate clearance for multiple skips for the TCG calibration. The holes should be spaced far enough from the end of the plate to limit the corner trap signal from affecting the TCG calibration. For instance, as a PAUT probe is swept over the SDH for TCG calibration, the high incidence angle can hit the corner trap before the low incidence angle has been cali- brated. If the corner trap signal has greater amplitude than the SDH, the corner trap amplitude will incorrectly be used for the TCG calibration rather than the SDH amplitude. There are two ways to avoid this issue: (1) space the SDH far enough from the end of the plate so that the corner trap is not reached at the high incidence angles, or (2) separate the incidence angle range into multiple groups and perform TCG calibration on each group separately. For instance, the 40Â°â70Â° incidence angle range could be split into a 40Â°â60Â° range and a 60Â°â70Â° range and each range swept over the SDH independently. The TCG for each angle range is then com- bined within the instrument software to provide for a single TCG covering the entire angle range. This is particularly an issue for long sound paths since the coverage between the 40Â° and the 70Â° beams is quite large. Based on preliminary CIVA analysis, it is recommended that the SDHs should be spaced at least 5â³ from the end of the plate with a 1â³ plate thickness, although the TCG may still require splitting into two groups for long sound paths. The recommended dimensions for a 1â³ thick calibration block are shown in Figure 17, but the SDH spacing and plate width may be modified based on the specific probe used in the inspec- tion. The dimensions for the plate length, width, SDH depth, spacing, and placement may all be different for different plate thicknesses. For instance, the recommended dimensions for a 2â³ thick calibration block are shown in Figure 18. This plate would also require the TCG be split into two groups for long sound paths, such as the third leg 4.5â³ depth TCG point. The 2.5â³ width was determined by using the 70Â° incidence angle for a 2.25 MHz 16 mm wide probe (equivalent to 5 MHz 8 mm wide probe) for the third leg 4.5â³ depth TCG point since this point would cover first and second leg scanning for 2â³ thick material. It should be noted that machining a 0.06â³ dia. SDH through 2.5â³ thick material may be difficult due to the short length of available drill bits. Recommended Figure 16. Possible TCG scanning positions.
46 placement of ultrasonic transducers for transfer correction and TCG calibration are shown in Figure 9 and Figure 16, respectively. Finally, the research team recommends that Annex K require the calibration block to be similar in temperature to the test object when calibration is performed. During the attenuation experimental testing, differences in attenuation measurements were noted when measurements were taken at different temperatures. It is recommended that the temperature limits from the AWS D1.1 PAUT proposal of Â±25Â°F be included in AWS D1.5 Annex K. 3.7 Amplitude Limit for Rejection of Flaws 3.7.1 CIVA Modeling Modifications to the AWS D1.5 Annex K acceptance criteria were developed in order to improve detection and rejection of critical weld flaws. The approach to the inspection proce- dure is to use line scanning for detection of weld flaws and follow-up raster scanning to evaluate rejection of weld flaws. The amplitude to be compared to the acceptance criteria limits will be the maximum amplitude measured during ras- ter scanning. Raster scanning will involve scanning the indi- cation over the entire incidence angle range from both sides of the weld while also rotating the probe. Compared with line scanning, raster scanning will maximize the amplitude as it will involve movement of the probe to account for unknown parameters such as flaw tilt, flaw skew, and flaw location. In order to develop the amplitude limit for flaw rejection, CIVA analysis was performed using a 2.25 MHz PAUT probe with an active aperture of 16 mm by 16 mm and an angu- lar range of 45Â°â70Â°. Using CIVA, the maximum amplitude that would be found during typical raster scanning with TCG was determined. The CIVA analysis was performed on flaws similar in size to the critical planar and volumetric flaws. The amplitude from the indications was referenced to the ampli- tude of a 1.5 mm (0.06â³) diameter SDH at a similar sound path, with positive amplitude having a higher response than the SDH and negative amplitude having a smaller response than the SDH. Therefore, positive results are more severe than negative results. Flaw tilt of planar flaws were evaluated at 0Â°, Â±5Â°, Â±30Â°, and Â±45Â°. The least positive maximum amplitude measured over all of the evaluated flaw tilts is the control used to reject critical flaws regardless of their tilt. In other words, the smallest possible amplitude that would result from raster scanning, regardless of the actual flaw tilt, forms the basis of the rejection limit. Since the critical flaw size depends on the stress range, stress concentration from thickness transitions, and the through- thickness location of the flaw, different critical flaw sizes are possible depending upon these inputs. The critical flaw sizes for surface and embedded flaws in equal thickness and thick- ness transition welds were evaluated for 4 ksi and 8 ksi stress ranges. Plate thicknesses of 0.5â³ and 2â³ were used to account for variations due to the probe near field and natural beam shape. The results of the parametric CIVA modeling for the controlling maximum amplitude compared to the 1.5 mm (0.06â³) diameter SDH are shown in Table 33 and Table 34 for critical planar and volumetric flaws, respectively. The results of this analysis are further summarized in Table 35 by combining the results of embedded and surface flaws for the same stress range and weld type (i.e., equal thick- Figure 17. Recommended calibration block for 1â thick plate. Figure 18. Recommended calibration block for 2â thick plate.
47 ness or thickness transition). This forms the basis of possible acceptance criteria amplitude limits to be applied to the raster scanning results. The acceptance criteria amplitude limit varies from â6 dB (indication amplitude 6 dB below refer- ence amplitude) for critical planar flaws in equal thickness welds under 4 ksi stress range to â18 dB for critical embedded volumetric flaws in thickness transition welds under 8 ksi stress range. It should be noted that critical surface planar and volu- metric flaws in thickness transition welds under 8 ksi stress range are not included in this table since the critical flaw size according to fracture mechanics was so small that the stan- dard analytical CIVA models are not valid (approximately half the wavelength of the 2.25 MHz probe). 3.7.2 Comparison to Current Amplitude- based Acceptance Criteria Figure 19 compares the acceptance criteria amplitude limits from the CIVA analysis in Table 35 to the current accep- tance criteria in AWS D1.5 Annex K for flaws in the middle half of tension welds. In this figure, the maximum amplitude from the indication compared to the 1.5 mm (0.06â³) diam- eter SDH is on the vertical axis and the indication length is along the horizontal axis. Combinations of amplitude and length that fall below the line would be accepted while those above the line are rejected. For instance, AWS Annex K would Stress Range Flaw Location Weld Type Critical Planar Flaw Comparable CIVA Analysis Height (in) Length (in) Height (in) Length (in) Plate Thickness (in) Controlling Maximum Amplitude (dB) 4 ksi Embedded Equal 0.20 0.20 0.20 0.20 0.5/2 -6 Transition 0.17 0.17 0.15 0.15 0.5/2 -8 Surface Equal 0.10 0.13 0.10 0.15 0.5/2 -2 Transition 0.06 0.08 0.05 0.10 0.5/2 -10 8 ksi Embedded Equal 0.14 0.14 0.15 0.15 0.5/2 -8 Transition 0.08 0.08 0.10 0.10 0.5/2 -10 Surface Equal 0.06 0.08 0.05 0.10 0.5/2 -10 Transition 0.02 0.03 0.025 0.025 0.5 -38 Table 33. CIVA results of controlling maximum amplitude for critical planar flaws. Stress Range Flaw Location Weld Type Critical Volumetric Flaw Comparable CIVA Analysis Height (in) Length (in) Diameter (in) Plate Thickness (in) Controlling Maximum Amplitude (dB) 4 ksi Embedded Equal 0.22 0.41 0.25 0.5/2 -7 Transition 0.24 0.24 0.25 0.5/2 -7 Surface Equal 0.23 0.31 0.25 0.5/2 -7 Transition 0.08 0.11 0.08 0.5/2 -13 8 ksi Embedded Equal 0.14 0.14 0.125 0.5/2 -13 Transition 0.08 0.08 0.08 0.5/2 -18 Surface Equal 0.06 0.08 0.08 0.5/2 -13 Transition 0.02 0.03 0.03 0.5 -18* *Result may not be valid since standard analytical CIVA model is not valid due to small flaw size relative to the wavelength Table 34. CIVA results of controlling maximum amplitude for critical volumetric flaws. Stress Range Weld Type Planar Flaws Volumetric Flaws 4 ksi Equal Thickness -6 dB -7 dB 4 ksi Thickness Transition -10 dB -13 dB 8 ksi Equal Thickness -10 dB -13 dB 8 ksi Thickness Transition (Embedded Flaws Only) -10 dB -18 dB Table 35. Raster scanning acceptance criteria limits from CIVA.
48 accept an indication that was up to 5 dB above reference [i.e., 5 dB greater amplitude than the 1.5 mm (0.06â³) diameter SDH] as long as the indication length is 0.5â³ or less. Also, AWS Annex K would accept any indication which is more than 6 dB below reference [i.e., amplitude of â6 dB or less compared with the 1.5 mm (0.06â³) diameter SDH] regardless of the indication length. It is apparent from Figure 19 that the CIVA amplitude limit results are more conservative than the AWS Annex K accep- tance criteria limits since the AWS Annex K line is above the CIVA results. For instance, Annex K acceptance criteria has higher amplitude limits for Class A and B flaws than any of the CIVA results. Therefore, there is more area for indications to fall under the Annex K line in the acceptable zone rather than the CIVA results. Once again, since Annex K also allows for evaluation of amplitude based on the initial line scan, it is unlikely that the amplitude used to compare to the acceptance criteria will be the maximum flaw amplitude due to possible flaw tilt, skew, and location in relation to the probe. The lack of maximizing the flaw amplitude, which is compared to the acceptance criteria for Annex K, would compound the differ- ences between an evaluation using Annex K and using raster scanning with the CIVA-determined limits. In order to compare the proposed amplitude limits from CIVA to the conventional UT amplitude limits, the conven- tional UT amplitude limits must be inverted from positive to negative integers. This is due to the fact that the conven- tional UT and PAUT have different approaches to obtain the indication amplitude. Conventional UT, according to AWS D1.5, requires modifying the instrument gain in order to force the indication amplitude to match the reference amplitude of the 1.5 mm (0.06â³) diameter SDH. Indications with high amplitude result in a low (or negative) indication rating since instrument gain is removed in order to bring the indication amplitude down to reference amplitude. For instance, in Figure 20 (left), the indication exceeds the refer- ence amplitude at reference gain; this indication has greater amplitude than the reference standard. In Figure 20 (right), 4 dB gain was removed from the instrument to bring the indication amplitude to the level of the reference ampli- tude. Assuming that the attenuation factor would be zero in this case, this indication would have an indication rating of â4 dB. This is the opposite of PAUT testing where the indications with high amplitude result in a more positive reported amplitude compared to the reference amplitude. For instance, PAUT inspection of the indication in Figure 20 would have a reported amplitude of +4 dB compared to the reference amplitude. Figure 21 compares the acceptance criteria amplitude limits from the CIVA analysis in Table 35 to the current acceptance criteria in AWS D1.5 Clause 6 conventional UT for flaws in the middle half of the tension welds after inverting the positive to negative values for the conventional UT tables. Conventional UT in Clause 6 includes separate criteria depending on the plate thickness and incidence angle used in the inspection. Rather than having one line like Annex K, there are eight dif- ferent lines for conventional UT acceptance limits. It is appar- ent that the conventional UT acceptance criteria are more -20 -15 -10 -5 0 5 10 0 1 2 3 4 5 M ax im um A m pl itu de (0 .0 6 in ch d ia . S DH fo r R ef er en ce ) (d B) Indication Length (in) CIVA -6 dB CIVA -7 dB CIVA-10 dB CIVA-13 dB Rejectable Acceptable More Conservative Class A Limit Class B Limit Class C Limit Figure 19. Comparison of CIVA results to AWS D1.5 Annex K acceptance criteria.
49 conservative than AWS D1.5 Annex K since the CIVA results from â6 dB to â13 dB correlate to the conventional UT limits while they are in the acceptable range for Annex K. Although it may seem that the CIVA results are overly conservative when compared to Annex K, they correlate well to the conventional UT acceptance criteria, which appears to have provided good historical performance when used for UT inspection. It is also important to remember that the amplitude limits from CIVA were derived completely inde- pendent of the Clause 6 conventional UT acceptance criteria by modeling critical flaws derived from fracture mechanics. While the similarities of the CIVA results to traditional acceptance criteria help to bolster confidence in the newly derived amplitude limits, these limits are not based solely on the historical performance of workmanship-based accep- tance criteria but rather on FFS. The 2018 edition of the CSA W59 code  includes an alternative acceptance criteria based on TCG rather than the fixed attenuation approach. This acceptance criteria was based on the CSA fixed attenuation conventional UT tables, which are similar to the AWS D1.5 Clause 6 conventional UT tables. The CSA W59 alternative TCG acceptance criteria may be applied to conventional UT or manual raster scanned PAUT. Encoded PAUT is specifically noted as an alternative ultra- sonic system which is subject to a written agreement between the engineer and contractor along with development of an appropriate scanning procedure and acceptance criteria. The CSA TCG acceptance criteria were derived by combining the Figure 20. Conventional UT amplitude measurement. -20 -15 -10 -5 0 5 10 0 1 2 3 4 5 M ax im um A m pl itu de (0 .0 6 in ch d ia . S DH fo r R ef er en ce ) ( dB ) Indication Length (in) CIVA -6 dB CIVA -7 dB CIVA-10 dB CIVA-13 dB CIVA -18 dB Conventional UT 5/16"-3/4" PL 70 Deg Conventional UT >3/4"-1.5" PL 70 Deg Conventional UT >1.5"-2.5" PL 70 Deg Conventional UT >1.5"-2.5" PL 60 Deg Conventional UT >1.5"-2.5" PL 45 Deg Conventional UT >2.5"-4" PL 70 Deg Conventional UT >2.5"-4" PL 60 Deg Conventional UT >2.5"-4" PL 45 Deg Rejectable Acceptable More Conservative Class A Limit Class B Limit Class C Limit Figure 21. Comparison of CIVA results to AWS D1.5 conventional UT acceptance criteria.
50 amplitude limits for the various incidence angles in the fixed attenuation tables along with correction for the difference in true attenuation compared to the fixed attenuation model [14â16]. By taking this approach, the CSA code attempts to recreate the same level of quality using the TCG acceptance criteria as the traditional fixed attenuation acceptance criteria. Figure 22 compares the TCG acceptance criteria in CSA W59-18 to the CIVA results for flaws in the middle half of the tension welds. Since the CSA TCG acceptance criteria depends on the plate thickness, there are three different criteria plotted in this figure. It can be seen that the CIVA results of â7 dB and â10 dB cross the CSA acceptance criteria and, therefore, are quite similar. The CIVA result of â6 dB would have been rejectable regardless of the indication length, according to the CSA TCG acceptance criteria for all of the plate thicknesses; therefore, the â6 dB limit would be slightly less conservative to the CSA TCG limits. On the other hand, the CIVA results of â13 dB and â18 dB would be acceptable regardless of the indication length and therefore would be more conservative. 3.7.3 Verification Testing of Weld Flaw Samples Initial verification of the CIVA analysis was performed by rescanning the round robin specimens with two different PAUT probes and using a standard AISI 1018 steel calibra- tion block with 1.5 mm (0.06â³) diameter SDHs for TCG. Testing was performed using a Zetec Topaz 16 with a Zetec AXL-2.25 MHz PAUT probe with 16 active elements (active aperture of 16 mm by 20 mm) and with a Zetec AM-5 MHz PAUT probe with 16 active elements (active aperture of 9.6 mm by 10 mm). In general, it was found that the ampli- tude from the flaws was greater using the 5 MHz probe as compared with the 2.25 MHz probe. This is not surprising since the wavelength is smaller for the 5 MHz probe and it is more sensitive to small flaws. This testing found that the lowest maximum amplitude from a planar flaw that would be rejectable according to fracture mechanics (excluding the extremely small critical size calculated for surface flaws in thickness transition welds under 8 ksi) was +3.6 dB for a vertical crack which was 0.17â³ high by 0.40â³ long. Therefore, this flaw would have been rejectable according to the amplitude limits found during the CIVA analysis. The round robin specimens also included some very small planar flaws which would only be rejectable according to fracture mechanics as a surface flaw in thickness transition weld under 8 ksi. The lowest maximum amplitude from these flaws was â5.3 dB for a surface breaking crack which was 0.02â³ high by 0.04â³ long. This flaw would also have been rejectable according to the amplitude limits found during the CIVA analysis. The lowest maximum amplitude from a volumetric flaw was â13 dB for a near surface group of porosity, which was 0.09â³ high by 3.31â³ long with a maximum pore diameter measured with RT of 0.05â³. This pore diameter is similar to -20 -15 -10 -5 0 5 10 0 1 2 3 4 5 M ax im um A m pl itu de (0 .0 6 in ch d ia . S DH fo r R ef er en ce ) ( dB ) Indication length (in) CIVA -6 dB CIVA -7 dB CIVA-10 dB CIVA-13 dB CIVA -18 dB CSA W59-18 5/16"1.5" PL CSA W59-18 >1.5"-2.5" PL CSA W59-18 >2.5"-4" PL Rejectable Acceptable More Conservative Class A Limit Class B Limit Class C Limit Figure 22. Comparison of CIVA results to CSA W59 TCG acceptance criteria.
51 the critical volumetric flaw used in the CIVA analysis for the 4 ksi thickness transition and 8 ksi equal thickness cases. This flaw would have been rejectable according to the amplitude limits found during the CIVA analysis for these cases. Current RT acceptance criteria would have also rejected this flaw. Final verification of the CIVA analysis was performed by an ASNT Level III UT/Level II PAUT technician. Round robin specimens with low amplitude indications were utilized for this testing along with additional flawed weld specimens that were acquired specifically for this testing. These specimens were acquired because they included known weld flaws that were small in size relative to many of the flaws included in the round robin specimens. The final verification testing involved line scanning the samples for flaw detection followed by raster scanning for evaluation of acceptance. As the flaw detection amplitude limit, which would require follow-up raster scanning, was unknown at the time (explained further in the follow- ing section), all indications with an amplitude greater than â20 dB during the initial line scanning were further investi- gated with raster scanning to determine the maximum ampli- tude. Testing was performed with an Olympus Omniscan MX2 with a 2.25L16-AWS1 PAUT probe, which is a 2.25 MHz probe with 16 active elements (active aperture of 16 mm by 16 mm). Table 36 shows the results from this testing for all intended flaws located in the test plates. This table includes the flaw type, intended flaw height and length, the rejection rate from the round robin results for conventional UT, the maximum measured flaw length from the line scanning, and the maxi- mum amplitude from the follow-up raster scanning. The flaw length was measured using the 6dB drop method, and the flaw length of scattered indications not separated by more than 2L was combined to determine the overall flaw length. Cau- tion should be taken if comparing the results of this testing to the current Annex K acceptance criteria as the reported amplitude for each flaw was peaked during raster scanning. Drawing Details UT Rejection Rate Line Scan Results Raster Scan Results Maximum Amplitude Rejection Limit Flaw Type Flaw Height Flaw Length Maximum Measured Length -13 dB -13 dB & 1" long -10 dB -8 dB -6 dB HAZ Crack 0.18 0.52 NA 0.99 8.0 Y N Y Y Y Porosity 0.10 0.73 NA 0.79 -6.2 Y N Y Y N LOF 0.11 0.63 NA 1.14 10.8 Y Y Y Y Y HAZ Crack 0.14 0.57 NA 0.83 -0.6 Y N Y Y Y Slag 0.10 0.74 NA 0.79 2.3 Y N Y Y Y Crack 0.14 0.37 NA 0.51 -12.1 Y N N N N LOF 0.18 0.50 NA 1.10 11.8 Y Y Y Y Y LOF 0.12 0.64 NA 1.14 15.5 Y Y Y Y Y Toe Crack 0.19 0.49 NA 0.55 2.9 Y N Y Y Y Toe Crack 0.11 0.61 NA 0.87 -1.0 Y N Y Y Y Slag 0.09 0.92 NA 0.63 -4.2 Y N Y Y Y HAZ Crack 0.14 0.45 NA 0.87 3.7 Y N Y Y Y IP 0.10 0.88 NA 0.83 5.5 Y N Y Y Y Slag 0.32 0.37 100% 0.55 1.7 Y N Y Y Y Slag 0.16 0.18 100% 0.47 2.9 Y N Y Y Y Slag 0.10 0.90 100% 1.73 -1.8 Y Y Y Y Y Porosity 0.09 3.31 80% 3.39 -8.0 Y Y Y Y N Toe Crack 0.02 0.04 40% 0.28 -8.4 Y N Y N N LOF 0.03 0.06 80% 0.36 -3.6 Y N Y Y Y Slag 0.06 0.03 0% 0.39 -8.2 Y N Y N N Slag 0.17 3.61 100% 3.47 0.9 Y Y Y Y Y Table 36. Experimental verification testing of flaw rejection amplitude limit.
52 The maximum amplitude and measured length were used to evaluate each flaw using five different criteria based on the CIVA results: 1. Rejection of flaws with maximum amplitude â¥ â13 dB 2. Rejection of flaws with maximum amplitude â¥ â13 dB and â¥1â³ long (i.e., reject scattered low-amplitude flaws such as porosity) 3. Rejection of flaws with maximum amplitude â¥ â10 dB 4. Rejection of flaws with maximum amplitude â¥ â8 dB 5. Rejection of flaws with maximum amplitude â¥ â6 dB All of the intended flaws had a maximum amplitude equal to or greater than â13 dB. Therefore, all of the intended flaws are rejectable under the first criteria (â¥ â13 dB). The second criteria was meant to reject low-amplitude flaws, such as slag or porosity, that were over 1â³ long and would likely be applied in conjunction with another criteria based solely on the amplitude. Six flaws are rejectable under this criteria including three lack-of-fusion flaws, two slag flaws, and one porosity grouping. All of the rejectable flaws under this criteria were also rejectable to all of the other amplitude-only criteria with the exception that the poros- ity would have been acceptable if the rejection limit was set to â6 dB. Therefore, it may not be necessary to apply the second criteria in conjunction with the other amplitude- only criteria. The third criteria (â¥ â10 dB) rejected all intended flaws except for a 0.14â³ Ã 0.37â³ crack. This crack is vertical and embedded so the amplitude response relied on tip diffraction. This specimen included weld reinforcement on both faces of the weld, so the entire incidence angle range could not be swept over the flaw before the front of the probe contacted the weld reinforcement. Because of this, the incidence angles were confined to high angles, and the maximum amplitude was measured at the 67Â° incidence angle. CIVA analysis for similar flaws had a maximum amplitude at â¼60Â° incidence angle (since the reinforcement was not modeled in CIVA). According to the critical flaw size for embedded planar flaws, the critical flaw size for a similar aspect ratio for the 4 ksi stress range in an equal thickness weld was approximately 0.13â³ Ã 0.31â³. Therefore, this crack would have been critical, but since this plate was not included in the round robin testing program, the rejection rate according to conventional UT is unknown. Three of the intended flaws are acceptable according to the fourth criteria (â¥ â8 dB). This includes the previously men- tioned vertical, embedded crack along with a very small surface breaking crack (0.02â³ Ã 0.04â³) and a very small slag inclusion (0.06â³ Ã 0.03â³). The small surface breaking crack had a rejection rate of 40% according to the conventional UT round robin results while the small slag inclusion had a 0% rejection rate for conventional UT. Five of the intended flaws are acceptable according to the fifth criteria (â¥ â6 dB). This includes the three flaws from the fourth criteria along with two groupings of porosity. Once again, it should be mentioned that this testing was performed on the round robin plates with the lowest ampli- tude indications. Therefore, it is likely that the indications in the other round robin plates would have been rejected with this criteria. Finally, any amplitude-based acceptance criteria will have variability from differences in probe parameters, calibration procedures and standards, probe pressure, and final probe location at maximum amplitude. Therefore, while verifi- cation testing is important to provide physical test results to verify the CIVA modeling, specific maximum amplitude values measured by a technician could be expected to vary by â¼Â±4 dB. 3.7.4 Recommendation for AWS A marked-up version of Annex K and associated com- mentary with the recommended changes is included in Appendix G. Based on the CIVA analysis of critical planar flaws and the experimental testing, it is recommended that the acceptance criteria amplitude limit for flaw rejection be set at 10 dB under the amplitude from the 1.5 mm (0.06â³) diameter SDH (i.e., â10 dB). As shown in Table 36, this limit would result in rejection of all intended flaws from the veri- fication testing except for the 0.14â³ Ã 0.37â³ embedded crack. As stated previously, it is believed that the weld reinforce- ment that limited accessâand thus limited the use of the incidence angles that could be swept over this flawâresulted in the low amplitude response. Therefore, it is recommended to set the Automatic Rejection Level (ARL) as 10 dB under Standard Sensitivity Level (SSL) for tension welds. These indications would be considered Class A defects and be automatically rejected regardless of length. The CIVA analysis of critical volumetric flaws found that the amplitude associated with critical pores may be as low as â13 dB. Volumetric flaws such as slag and porosity are typically made up of scattered groupings of individual discontinuities. In order to reject large groups of volumetric discontinuities that include a critical sized pore, it is recommended that indications with a maximum amplitude between â13 dB and â10 dB during the follow-up raster scanning be rejected if the length of the entire grouping of discontinuities exceeds 1â³. It is recommended that a new amplitude limit referred to as the Evaluation Level (EVL) be set at 13 dB under SSL, and indications which exceed the EVL but are less than the ARL be defined as Class B indications. Measurement of the flaw length will be needed for rejec- tion evaluation of Class B indications or limits for repair of Class A defects. It is recommended that the length
53 measurement for flaws use the 6 dB drop method during the manual raster scan. Some PAUT acceptance criteria use a standard amplitude limit for length measurement rather than the 6 dB drop. In these cases, the length is determined to encompass the full extent of the flaw, which has ampli- tude greater than this limit. While use of this method has some merit, oversizing of indications with saturated signals may occur. It is worth commenting that the research team initially intended on recommending that an amplitude limit of 18 dB below SSL be used for the length measurement of indi- cations which are above the EVL but less than the ARL (i.e., Class B). (The â18 dB limit corresponds to the amplitude limit for detection of flaws during the encoded line scan, as will be described in the following section.) In this case, length mea- surement of Class B flaws would have involved measuring the extents where the signal exceeds the flaw detection limit. In essence, this would have ensured that any indications that would be considered as detectable are included in the evalu- ation against the 1â³ length limit. Since 6 dB drop on flaws with maximum amplitude between â13 dB and â10 dB will essentially be equivalent to the â18 dB limit (i.e., â19 dB and â16 dB, respectively), it was finally decided to recommend the use of the familiar 6 dB drop method for all length measure- ments. This will provide consistency for Class A and Class B flaws, along with flaws in compression welds. It should also be noted that the existing requirements on spacing between Class B flaws from each other and the edge of the plate have been retained. Finally, it is recommended that for all indications inves- tigated in follow-up manual raster scanning, a screenshot should be required and data documented at the location of maximum indication amplitude. 3.8 Amplitude Limit for Detection of Flaws 3.8.1 CIVA Modeling As stated previously, the approach to the inspection pro- cedure is to use line scanning for detection of weld flaws and raster scanning to evaluate rejection of weld flaws. The amplitude limit for detection of weld flaws will need to be set such that it will detect critical weld flaws during the line scan regardless of the flaw tilt, skew, and position in rela- tion to the probe. The detection amplitude limit, by defi- nition, must be lower than the acceptance criteria rejection limit since the rejection limit is compared to the maximum possible amplitude following raster scanning. In order to develop the amplitude limit for flaw detection, CIVA analysis was performed using the same probe that was used in the flaw rejection limit (2.25 MHz PAUT probe with an active aperture of 16 mm by 16 mm and an angular range of 45Â°â70Â°). The CIVA analysis was performed on the same size flaws used in the flaw rejection, which are similar in size to the critical planar and volumetric flaws. Since the ampli- tude was always referenced to the 1.5 mm (0.06â³) diameter SDH at a similar sound path, the reported amplitude is simi- lar to that which would be found during typical line scanning with TCG. Positive amplitude represents a higher amplitude response than the SDH, and negative amplitude represents a smaller amplitude response than the SDH. Encoded line scanning involves using the incidence angle range of PAUT focal laws to provide coverage of the entire weld volume and HAZ. As shown in Figure 23, the PAUT probe is moved in a direction parallel to the weld axis at a constant index offset and with the probe orientation remain- ing perpendicular to the weld axis. Due to the amplitude being strongly affected by the interaction of the flaw tilt and sound beam incidence angle, full coverage of the weld volume and HAZ should be pro- vided in two crossing directions (i.e., nearly perpendicular sound beam directions), as shown in Figure 24. This can be provided by either scanning from both sides of the weld or combining 1st and 2nd leg index offset scans from the same side of the weld. Due to the effects of attenuation and beam spread on long sound paths, scanning from both sides of the weld is preferred rather than relying on second leg scans with long sound paths. This has already been incor- porated in Annex K by requiring butt welds be tested from the same face but opposite sides of the weld axis where access is possible. JIS Z 3060 , which uses a DAC curve approach, requires that the plate be flipped and scanning performed in the 1st leg from the other face of the weld when the sound path exceeds 250 mm (9.8â³). For a 70Â° incidence angle, this would cor- respond to a depth of 3.4â³ which would be exceeded for 2nd leg scanning of plates thicker than 1.7â³. Therefore, it may be reasonable to set limits on the maximum sound path which can be used for sound coverage in order to limit the effects of attenuation and beam spread. For instance, coverage could still be provided for shorter sound paths through the use of additional line scans at a different index offset or flipping the Figure 23. Encoded line scanning.
54 plate and scanning from both weld faces. A reasonable limit may be limiting the sound path used for full coverage to 12â³, since this would still allow for full coverage to be provided at the 70Â° incidence angles for the 2nd leg in 2â³ thick plates (i.e., 4â³ deep TCG point). The recommendations for changes to Annex K included a statement that the probe dimensions shall be chosen in order to optimize the beam formation within the area of coverage. No exact sound path limit was provided. In order to develop the amplitude limit for the detection of flaws, the probe was moved across the weld flaws perpendicu- lar to the weld axis (i.e., the index offset was varied) in the CIVA simulations. The largest amplitude for all of the focal laws (incidence angles) was documented at 6 mm (0.24â³) increments of the index offset, as shown in Figure 25. The maximum amplitude across the incidence angle range for each index offset represents the largest amplitude that would occur if a line scan was performed using the same index offset. The indication amplitude was documented at a small increment of possible index offsets. The amplitude limit for flaw detec- tion was subsequently determined in order to detect critical weld flaws for any possible index offset used in line scanning (i.e., combination of possible probe and flaw locations). The only stipulation is that the flaw must be within the cover- age of the incidence angle range, which is already provided through minimum scanning coverage requirements. In other words, as long as full coverage was provided of the weld and HAZ, the amplitude from the critical weld flaw would surpass this limit and therefore be detected during line scanning. Since the flaw tilt is also an unknown parameter, the tilt of planar flaws was evaluated at 0Â°, Â±5Â°, Â±30Â°, and Â±45Â°. Flaw tilt away from the probe (the maximum amplitude in 1st leg) was defined as positive tilt while tilt towards the probe (the maximum amplitude in 2nd leg) was defined as negative tilt, as shown in Figure 26. Since the minimum scanning cover- age requirements provide for full coverage in two crossing directions, flaws tilted away from the probe when scanned from one side of the weld would be tilted towards the probe when scanned from the other side of the weld. Therefore, all necessary analysis could be performed by sweeping the probe over the flaw in 1st and 2nd leg. The smaller of the amplitude from the 1st leg results for positive flaw tilt or 2nd leg results for negative flaw tilt was used as the controlling amplitude for flaw detection in order to ensure that the flaw would be detected by only requiring sound coverage in two crossing directions. Therefore, it did not matter whether the sound beam that would impact the flaw was provided in 1st or 2nd leg as long as the full weld volume and HAZ was covered by sound in two crossing directions. Figure 27 displays the amplitude of a 0.2â³ Ã 0.2â³ embedded planar flaw at mid-thickness depth in a 2â³ thick plate, as the probe is swept over the flaw in 1st and 2nd leg. When there is no tilt, the amplitude varies from â5 dB to â15 dB compared to the 1.5 mm (0.06â³) diameter SDH until the weld flaw is no Figure 24. Line scanning sound coverage. Figure 25. Index offset increment.
55 longer within the sound beam coverage, at which point the amplitude drops quickly (0â³ index offset). When the flaw is tilted from vertical by 30Â° or 45Â°, the maximum amplitude increases to +10 dB when the sound beam is nearly perpen- dicular to the weld flaw. For the negative tilt cases, this occurs at index offsets of â4â³ to â2â³ where the flaw is impacted by sound in the 2nd leg. For positive tilt cases, it occurs at index offsets of â1â³ to 0â³ where the flaw is impacted by sound in the 1st leg. When the sound beam is nearly parallel to the tilted flaws (1st leg of negative tilted flaws and 2nd leg of posi- tive tilted flaws), the amplitude drops off considerably with amplitudes of â¼â25 dB. Therefore, sound coverage in two crossing directions is required in order to provide for detec- tion of tilted flaws. The critical flaws from Table 33 and Table 34 for the CIVA analysis used to develop the acceptance criteria rejection limits were also used in the study for the amplitude limits for flaw detection. The controlling amplitude over all possible index offset positions was tabulated for various flaw tilt and possible incidence angle ranges. The controlling amplitude was used to determine the overall minimum amplitude possi- ble during line scanning for critical weld flaws. This amplitude could then be used to form the basis of the amplitude detec- tion limit for encoded line scanning, which would require follow-up raster scanning for evaluation of acceptance. Table 37 includes the results from the CIVA flaw detection analysis for planar flaws without any skew (i.e., the flaw length is parallel to the weld axis). This table presents the minimum possible peak amplitude during a line scan as long as the flaw was within sound coverage provided in two crossing direc- tions. It presents parametric results for various-sized planar flaws, various flaw tilt, and various incidence angle ranges. The controlling line scan amplitudes (i.e., minimum peak ampli- tude depending on the chosen index offset) are highlighted for each flaw and incidence angle range. The maximum ampli- tude (from raster scanning) is also provided in the table. For some of the flaws, especially those with large tilt, the amplitude is very sensitive to the probe location. For instance, Figure 26. Flaw tilt during line scanning. -40 -35 -30 -25 -20 -15 -10 -5 0 5 10 15 -7 -6 -5 -4 -3 -2 -1 0 1 Ch an ge fr om R ef er en ce A m pl itu de (d B) Index Offset (in) 0.20"x0.20" 0 Tilt 0.20"x0.20" 5 Tilt 0.20"x0.20" -5 Tilt 0.20"x0.20" 30 Tilt 0.20"x0.20" -30 Tilt 0.20"x0.20" 45 Tilt 0.20"x0.20"x -45 Tilt 0.20" x 0.20" Tilted Embedded Flaw in 2" Plate Figure 27. Amplitude profile as probe is swept over weld flaw of various tilt angles.
56 the 0.20â³ Ã 0.20â³ embedded flaw with 45Â° tilt has a maximum amplitude of +9 dB from raster scanning, but a minimum peak amplitude during possible line scan locations of â12 dB even with full coverage from two crossing directions. Therefore, the amplitude of this flaw could be 21 dB below the maximum during the line scanning, even with providing full coverage in two crossing directions. This highlights the need for follow-up raster scanning rather than evaluating flaw rejection on the line scan results. The incidence angle ranges shown in Table 37 were chosen to investigate whether limits should be placed on the inci- dence angle used in the scan plan. Limiting the incidence angle range, while providing less sound coverage, may result in larger amplitudes for flaw detection. This would be simi- lar to the requirement in Annex K where the incidence angle used in the line scan must be within Â±10Â° of the weld fusion face. Rather than perform five different analyses for each flaw (one for each incidence angle range investigated), the indi- vidual incidence angle that had the largest amplitude across the 45Â°â70Â° angular range was documented for each index offset evaluated (0.24â³ increment). The minimum amplitude for each incidence angle range could then be determined from this data since, as the probe is swept over the flaw, the maximum amplitude at each index point will occur at a slightly different incidence angle. For example, referring back to the 0.20â³ Ã 0.20â³ flaw with 45Â° tilt, the maximum amplitude is +9 dB, but the minimum amplitude during line scanning using the 45Â°â70Â° incidence angle range was â12 dB. It is anticipated that the maximum amplitude would occur at an incidence angle of 45Â° since this would be perpendicular to the flaw. Therefore, limiting the incidence angle range closer during the line scan to 45Â° should result in a larger amplitude. This was confirmed in the CIVA analysis, where the minimum amplitude over the 45Â°â55Â° Flaw Parameters Maximum Amplitude (dB) Minimum Peak Line Scan Amplitude for Incidence Angle Range (dB) Location Height(in) Length (in) Plate Thickness (in) Tilt (deg) 45Â° - 70Â° 50Â° - 70Â° 55Â° - 65Â° 45Â° - 55Â° 45Â° - 50Â° Embedded 0.20 0.20 2 0 -5 -16 -16 -14 -15 -15 +/- 5 -6 -16 -12 -12 -16 -16 +/- 30 9 -10 -10 -9 -9 -9 +/- 45 9 -12 -12 -10 -6 4 0.15 0.15 2 0 -7 -17 -15 -15 -17 -17 +/- 5 -8 -17 -13 -13 -17 -17 +/- 30 5 -11 -10 -1 -11 -11 +/- 45 5 -15 -15 -12 -8 0 0.10 0.10 0.5 0 -6 -14 -14 -14 -10 -10 +/- 5 -10 -13 -13 -12 -13 -13 +/- 30 -4 -11 -11 -8 -11 NA +/- 45 -6 -15 -15 -15 -13 -6 0.10 0.10 2 0 -10 -21 -20 -17 -21 -21 +/- 5 -10 -20 -18 -17 -20 -20 +/- 30 -2 -14 -10 -7 -14 -14 +/- 45 -2 -17 -17 -17 -8 -5 Surface 0.10 0.15 0.5 0 3 -3 -3 1 -3 -3 +/- 5 -2 -8 -8 -8 -5 -4 +/- 30 0 -13 -13 -7 -10 -10 +/- 45 -2 -17 -17 -15 -6 -6 0.05 0.10 0.5 0 -3 -11 -9 -6 -11 -11 +/- 5 -6 -12 -12 -12 -11 -11 +/- 30 -8 -15 -15 -11 -15 -9 +/- 45 -10 -15 -15 -15 -11 -11 Table 37. CIVA flaw detection results for planar flaws.
57 incidence angular range was â6 dB and over the 45Â°â50Â° inci- dence angular range was +4 dB. Obviously, the amplitude of this flaw is very sensitive to the incidence angle, as is typical for tilted lack-of-fusion flaws. Along with limiting the sound coverage, limiting the inci- dence angle range assumes that the flaw tilt is known. Lack of fusion flaws are typically assumed to have the same tilt as the fusion face. However, one must also consider if the flaw is a vertical crack or tilted at Â±5Â°. Returning once again to the example of the 0.20â³ Ã 0.20â³ flaw, if the weld had a 45Â° bevel face and the incidence angular range was limited to either 45Â°â55Â° or 45Â°â50Â° but the weld flaw was tilted at Â±5Â°, the minimum peak line scan amplitude would be â16 dB for all possible index offsets. Therefore, using a flaw detection limit of â6 dB or +4 dB would be unconservative and the critical weld flaw would not be detected. While limiting the incidence angle range may be helpful for flaws with known tilt in order to better maximize the amplitude, it does not help when the flaw tilt is vertical or otherwise unknown. Table 38 includes the results from the CIVA flaw detection analysis for volumetric flaws. Since these flaws were modeled as spherical porosity, there is no flaw tilt or skew. As expected, spherical flaws are not as sensitive to probe location (i.e., index offset). For instance, the minimum amplitude during line scanning is relatively unaffected by the different com- binations of incidence angle ranges evaluated. Still, a loss of amplitude of 7 dB was typical for the line scan amplitude compared to the maximum possible during raster scanning. Similar to the raster scanning amplitude limits, the CIVA results could be further summarized by combining the results for flaw sizes comparable to the critical flaw sizes. The flaw detection limits for the 4 ksi and 8 ksi stress ranges for equal thickness and thickness transition welds is shown in Table 39. For planar flaws, limiting the incidence angle range to 55Â°â65Â° resulted in the largest line scan amplitudes, but only by a few decibels compared to using 45Â°â70Â°. Using an incidence angle range of 55Â°â65Â° would result in much less cover- age than using 45Â°â70Â°. Therefore, based on these results, it seems that using an incidence angle range from 45Â°â70Â° is jus- tified without the need for additional scan plan requirements beyond providing full coverage of the entire weld volume and HAZ in two crossing directions. As explained previously and shown in Figure 23, encoded line scanning is performed by keeping the probe perpen- dicular to the weld axis while probe movement is parallel to the weld axis. All of the previous CIVA analysis assumed Flaw Parameters Controlling Maximum Amplitude (dB) Minimum Line Scan Amplitude for Incidence Angle Range (dB) Location Height (in) Length (in) Plate Thickness (in) 45Â°â 70Â° 50Â°â 70Â° 55Â°â 65Â° 45Â°â 55Â° 45Â°â 50Â° Embedded 0.25 0.25 0.5 -7 -13 -13 -10 -13 -13 0.25 0.25 2 -6 -13 -13 -12 -10 -10 0.125 0.125 0.5 -13 -16 -16 -16 -15 -15 0.125 0.125 2 -13 -20 -20 -18 -17 -17 0.08 0.08 0.5 -18 -21 -21 -21 -21 -21 0.08 0.08 2 -18 -25 -25 -23 -22 -21 Surface 0.25 0.25 0.5 -7 -12 -9 -9 -12 -12 0.25 0.25 2 -5 -12 -12 -8 -8 -6 0.08 0.08 0.5 -13 -18 -18 -17 -15 -15 0.08 0.08 2 -11 -22 -22 -18 -15 -13 0.03 0.03 0.5 -18 -28 -28 -22 -18 -18 Table 38. CIVA flaw detection results for volumetric flaws. Planar Flaw Amplitude (dB) Volumetric Flaw Amplitude (dB) 45Â°â 70Â° 50Â°â 70Â° 55Â°â 65Â° 45Â°â 55Â° 45Â°â 50Â° 45Â°â 70Â° 50Â°â 70Â° 55Â°â 65Â° 45Â°â 55Â° 45Â°â 50Â° 4 ksi (Equal) -17 -17 -14 -16 -16 -13 -13 -12 -13 -13 4 ksi (Transition) -17 -16 -15 -17 -17 -22 -22 -18 -15 -15 8 ksi (Equal) -17 -16 -15 -17 -17 -22 -22 -18 -17 -17 8 ksi (Transitionâ Embedded) -21 -20 -17 -21 -21 -25 -25 -23 -22 -21 Table 39. Summary of CIVA flaw detection amplitude results without flaw skew.
58 that the flaw was aligned parallel to the weld axis and, there- fore, the probe was perfectly perpendicular to the weld flaw. While this is a valid assumption for raster scanning where the probe will be rotated as well as translated, it may be unconservative for line scanning. In order to account for this effect, the change in amplitude due to flaw skew was evaluated. CIVA analysis of embedded and surface vertical weld flaws in 0.5â³ and 2â³ plates with 5Â°, 10Â°, and 20Â° skew was performed and compared to the results with no skew. Since the sound is reflected to the side of skewed pla- nar flaws, as shown in Figure 28, lateral movement of the probe along the weld axis was performed as well as sweeping the probe over the flaw perpendicular to the weld axis. (Note: this figure represents the centerline of the sound beam, but the beam actu- ally has beam spread and width.) The maximum amplitude for flaws with skew is shown in Figure 29 for surface flaws and Figure 30 for embedded flaws. As the flaw skew is increased, the drop in amplitude increases, especially for larger flaws. This is due to the fact that the beam is hitting different parts of the flaw at different times, which causes the amplitude to drop more severely. Flaws with long sound paths also had larger drops in amplitude since the beam reflected off of the skewed flaw travels a further distance transverse to the probe. For instance, the embedded flaws in the 2â³ thick plate had greater drop in amplitude than the embedded or surface flaws in the 0.5â³ thick plate. The flaw and plate combinations chosen for this analysis correspond to those included in Table 37. In order to account for the effect of flaw skew on flaw detection, the drop in amplitude from the analysis of the ver- tical flaw with skew was added to the results of the tilted flaws without skew for each individual flaw. This assumes that the drop in amplitude from flaw skew will be similar for tilted and vertical flaws. This assumption was checked for the 0.20â³ Ã 0.20â³ embed- ded flaw in a 2â³ thick plate by modeling flaws with both skew and tilt and comparing the results to the estimated values. It was determined that this assumption was reasonable. For instance, in the case of 5Â° skew, the drop in amplitude due to skew on the vertical flaws was â4 dB. The drop in amplitude due to skew on the tilted flaws varied from â2 dB to â5 dB depending on the flaw tilt. In the case of a 10Â° skew where the drop in amplitude due to skew on the vertical flaws was â12 dB, the drop in amplitude for the tilted flaws varied from â7 dB to â15 dB. The flaw detection limits, including the effect of skew on the planar flaws, are given in Table 40 for 5Â° skew and Table 41 Figure 28. Flaw skew. -35 -30 -25 -20 -15 -10 -5 0 5 10 0 5 10 15 20 25 Ch an ge fr om R ef er en ce A m pl itu de (d B) Flaw Skew (deg) 0.05"x0.05" Surface Flaw in 0.5" Plate 0.05"x0.10" Surface Flaw in 0.5" Plate 0.05"x0.15" Surface Flaw in 0.5" Plate 0.10"x0.10" Surface Flaw in 0.5" Plate 0.10"x0.15" Surface Flaw in 0.5" Plate Figure 29. Flaw skew results for surface flaws.
59 -35 -30 -25 -20 -15 -10 -5 0 5 10 0 5 10 15 20 25 Ch an ge fr om R ef er en ce A m pl itu de (d B) Flaw Skew (deg) 0.10"x0.10" Embedded Flaw in 2" Plate 0.10"x0.30" Embedded Flaw in 2" Plate 0.20"x0.20" Embedded Flaw in 2" Plate 0.15"x0.15" Embedded Flaw in 2" Plate 0.10"x0.10" Embedded Flaw in 0.5" Plate Figure 30. Flaw skew results for embedded flaws. Planar Flaw Amplitude (dB) Volumetric Flaw Amplitude (dB) 45Â°â 70Â° 50Â°â 70Â° 55Â°â 65Â° 45Â°â 55Â° 45Â°â 50Â° 45Â°â 70Â° 50Â°â 70Â° 55Â°â 65Â° 45Â°â 55Â° 45Â°â 50Â° 4 ksi (Equal) -20 -20 -18 -20 -20 -13 -13 -12 -13 -13 4 ksi (Transition) -19 -17 -17 -19 -19 -22 -22 -18 -15 -15 8 ksi (Equal) -19 -17 -17 -19 -19 -22 -22 -18 -17 -17 8 ksi (Transitionâ Embedded) -23 -22 -19 -23 -23 -25 -25 -23 -22 -21 Table 40. Summary of CIVA flaw detection amplitude results with 5Ã§ flaw skew. Planar Flaw Amplitude (dB) Volumetric Flaw Amplitude (dB) 45Â°â 70Â° 50Â°â 70Â° 55Â°â 65Â° 45Â°â 55Â° 45Â°â 50Â° 45Â°â 70Â° 50Â°â 70Â° 55Â°â 65Â° 45Â°â 55Â° 45Â°â 50Â° 4 ksi (Equal) -28 -28 -26 -28 -28 -13 -13 -12 -13 -13 4 ksi (Transition) -26 -24 -24 -26 -26 -22 -22 -18 -15 -15 8 ksi (Equal) -26 -24 -24 -26 -26 -22 -22 -18 -17 -17 8 ksi (Transitionâ Embedded) -25 -24 -22 -25 -25 -25 -25 -23 -22 -21 Table 41. Summary of CIVA flaw detection amplitude results with 10Ã§ flaw skew.
60 for 10Â° skew. In general, including the 5Â° skew resulted in a â2 dB to â3 dB decrease in the flaw detection amplitude while the decrease was â9 dB to â12 dB for 10Â° skew. Since the drop in amplitude due to skew is greater for larger flaws, the 4 ksi equal thickness weld had lower flaw detection amplitude than the 4 ksi transition or 8 ksi equal thickness welds. Since the volumetric flaws were assumed to be spherical, the volumetric flaws were not affected by flaw tilt or skew; therefore, the volumetric flaw results are the same as those in Table 39. The proposed PAUT annex for AWS D1.1 was reviewed by the research team after performing the CIVA analysis. The proposed PAUT annex for D1.1 utilizes an incidence angle range from 40Â°â70Â° rather than the 45Â°â70Â° inci- dence angle range in D1.5 Annex K. One of the authors of the proposed D1.1 PAUT annex was contacted in order to obtain an explanation for the increase in the incidence angle range from 45Â° to 40Â°. The purpose of the incidence angle range increase was to enlarge the sound coverage area and to aid in verification of corner trap signals. The AWS D1.1 pro- posed PAUT annex also includes a requirement that the HAZ be covered with incidence angle range from 40Â°â60Â° in order to increase the detectability of corner trap signals from surface breaking HAZ cracks. The author of the proposed D1.1 PAUT annex stated that incidence angle was limited at 40Â° since standing wave signals were produced sometimes at 35Â° and this seemed risky since it was close to the first critical angle. Subsequently, additional CIVA analysis was performed by the research team on a subset of flaws with an extended inci- dence angle range of 40Â°â70Â°. This was used to verify that the previously determined flaw detection amplitude limits using the 45Â°â70Â° incidence angle range would be appropriate for use with a 40Â°â70Â° incidence angle range. All of these results with the 40Â°â70Â° range were within Â±1 dB of the results given in Table 39 for flaws without skew. Therefore, it was deter- mined that the incidence angle range could be extended from 45Â°â70Â° to 40Â°â70Â° without significantly altering the neces- sary flaw detection limits. 3.8.2 Verification Testing of Weld Flaw Samples In order to verify the CIVA results, physical testing was per- formed on weld samples with known weld flaws. These speci- mens included the round robin test plates as well as additional test plates with small weld flaws. None of the plates were fab- ricated with acoustically anisotropic material. This testing was performed in conjunction with final verification testing of the flaw rejection limits by an ASNT Level III UT/Level II PAUT technician. The weld samples were scanned with an Olympus Omniscan MX2 with a 2.25L16-AWS1 PAUT probe, which is a 2.25 MHz probe with 16 active elements (active aperture of 16 mm Ã 16 mm). Some supplemental line scanning was per- formed with a 5L64-A12 PAUT probe, which is a 5 MHz probe with 32 active elements (active aperture of 19.2 mm Ã 10 mm) in order to evaluate the differences between the 2.25 MHz and 5 MHz probes. The weld samples were scanned after performing TCG calibration on an AISI 1018 calibration block with 1.5 mm (0.06â³) diameter SDHs. The reference amplitude was set to 80% FSH, and +12 dB scanning gain was added. Line scanning was performed using an incidence angle range of 40Â°â70Â° with full coverage of the weld volume and HAZ in two crossing directions. No additional requirements were imposed on the scan plan. All indications with an amplitude greater than â20 dB were further investigated through raster scanning for flaw rejection verification testing, as previously explained. Since some of the plates had reinforcement on both faces of the plate, these plates were line scanned in the 1st leg from each side of the weld and each face of the plate (i.e., four line scans). This was also performed for the plates which were thicker than 0.75â³, since the TCG did not extend beyond 2â³ depth. Subse- quently, the 2nd leg portion of the scan would have extended beyond the last TCG point. Scanning of the 0.75â³ thick plates was performed with two line scans: one from each side of the weld using the 1st and 2nd leg to cover the entire weld, and HAZ in two crossing directions. Some of these plates had additional scans performed from the other face of the plate to verify whether all of the intended weld flaws would still have been detected if full coverage was provided from that face. The results of the verification testing are shown in Table 42 for the intended weld flaws. This table includes the maximum amplitude for each flaw after evaluating each of the line scans necessary for full coverage in two crossing directions. It also includes the maximum amplitude from the follow-up raster scanning that was previously reported. All of the results pro- vided in the table are from testing with the 2.25 MHz probe, but the amplitude with the supplemental line scans with the 5 MHz probe was found to be similar. Since +12 dB scanning gain was added to the line scans and the Olympus Omniscan MX2 instrument truncates the A-scan at 250% FSH, the maxi- mum amplitude during the line scan was truncated at â2.1 dB. As expected, the maximum amplitude from the line scan can be significantly lower than the maximum amplitude from the raster scanning. The largest difference between the line scan and the raster scan (excluding truncated line scan results) was for the slag, which was 0.16â³ Ã 0.18â³. This flaw had a maximum amplitude during the line scans from Face A of the plate of â9.9 dB, while the follow-up raster scan was +2.9 dB. This is a difference of 12.8 dB, approximately a factor of four times as much amplitude. The intended flaw with the smallest maximum amplitude after evaluating each line scan necessary for full coverage was â13.9 dB for the embedded vertical crack, which was 0.14â³ Ã 0.37â³. Therefore, all of the intended flaws would be detected
61 for any flaw detection limit of â14 dB or less. All of the flaw detection limits for planar flaws determined through the CIVA analysis and presented in Table 39 are below â14 dB, so use of the CIVA results would have resulted in all of the intended flaws to be detected. None of the weld flaws in the weld flaw samples had any skew relative to the weld axis. For this reason, the maximum line scan amplitude should be compared to the limits in Table 39 rather than those in Table 40 or Table 41. During the verification testing, there were many unintended indications that crossed the initial detection limit threshold of â20 dB, excluding geometric indications from weld reinforce- ment and surface roughness. While some of these unintended flaw detections seem to correlate to actual unintended flaws, most of these indications seemed to be spurious repeating sig- nals on the high incidence angle (70Â°), as shown in Figure 31 (left). Sometimes these repeating signals also appeared at the Drawing Details Line Scan Results Raster Scan Results Flaw Type Flaw Height (in) Flaw Length (in) Maximum Amplitude (dB) Maximum Amplitude (dB) HAZ Crack 0.18 0.52 -2.1* 8.0 Porosity 0.10 0.73 -10.5 -6.2 LOF 0.11 0.63 -2.1* 10.8 HAZ Crack 0.14 0.57 -2.1* -0.6 Slag 0.10 0.74 -6.4 2.3 Crack 0.14 0.37 -13.9 -12.1 LOF 0.18 0.50 -2.1* 11.8 LOF 0.12 0.64 -2.1* 15.5 Toe Crack 0.19 0.49 -2.1* 2.9 Toe Crack 0.11 0.61 -2.1* -1.0 Slag 0.09 0.92 -3.0 -4.2 HAZ Crack 0.14 0.45 -2.1* 3.7 IP 0.10 0.88 -2.1* 5.5 Slag 0.32 0.37 -2.1* (Face A) 1.7 -2.1* (Face B) Slag 0.16 0.18 -9.9 (Face A) 2.9 -2.1* (Face B) Slag 0.10 0.90 -3.4 (Face A) -1.8 -2.1* (Face B) Porosity 0.09 3.31 -8.7 (Face A) -8.0 -12.5 (Face B) Toe Crack 0.02 0.04 -8.3 -8.4 LOF 0.03 0.06 -2.8 -3.6 Slag 0.06 0.03 -12.1 -8.2 Slag 0.17 3.61 -2.1* 0.9 *A-scan was truncated at 250% FSH with +12 dB scanning gain, which correlates to -2.1 dB Table 42. Verification testing of flaw detection results. Figure 31. Spurious signals on high-incidence (left) and low-incidence (right) angles.
62 low incidence angle (40Â°), as shown in Figure 31 (right). These spurious indications also appeared during scanning of clean production welds that had been inspected with digital RT and had no noted indications. The spurious indications were only noticed in the scanning with the 2.25L16-AWS1 probe and not with the 5L64-A12 probe. After further investigation, it was determined that these indications were likely due to grating lobes due to the large pitch of the 2.25 MHz probe (1 mm) compared with the wavelength (1.45 mm). From the literature , grating lobes typically appear when the pitch is greater than the wavelength, although they may appear for a slightly smaller pitch as well. The spurious indications seemed sensitive to changes to the surface roughness of the plates, even though the surface roughness was typical of production welds. Grating lobes may reflect off of the surface roughness, which may explain why they appeared on locations with more roughness. Follow-up raster scanning on the detected locations of the spu- rious indications found no relevant indications greater than â20 dB, which confirms that they are noise. Due to the preva- lence of these grating lobe signals throughout the line scans, it was determined to exclude indications that were characterized as grating lobe indications from follow-up raster scanning. Table 43 summarizes the detection and rejection results of unintended indications during the verification testing. When the flaw detection limit was set at â20 dB, 36 unintended indi- cations (excluding geometric indications) were detected that required follow-up raster scanning. Of those 36 indications, 33 indications had a maximum amplitude during raster scan- ning that was less than â13 dB; these indications would be accepted under all of the flaw rejection limits previously dis- cussed. Three unintended indications were detected above â20 dB and would be rejected if the flaw rejection limit were set at â13 dB, while only one of these indications would be rejected if the flaw rejection limit were set at â10 dB or higher. As the flaw detection limit is shifted up (more positive), the number of unintended indications drops dramatically, but indications may be missed that would have been rejected during follow-up raster scanning. For instance, if the flaw detection limit were set at â18 dB, the number of accept- able unintended indications that are detected and require follow-up raster scanning decreases from 33 to 18. Increas- ing the flaw detection limit to â16 dB decreases the number of acceptable unintended indications to only 8, but one of the unintended indications that would have been rejectable if the rejection limit were set to â13 dB would not have been detected. Thus, that indication would now be accepted since it was effectively missed. Due to the high rate of unintended indications found during the testing of the weld flaw specimens, a clean pro- duction weld 32 inches long by 2 inches thick was obtained and scanned. This specimen was specifically obtained to get a feel for what level of ânoiseâ might be expected in clean shop production welds. Other than grating lobe signals, only two indications (excluding geometric indications) were noted as greater than â20 dB during four line scans with the 2.25 MHz probe, and no indications (excluding geometric indications) appeared above this threshold with the 5 MHz probe. Neither of the two indications detected with the 2.25 MHz probe exceeded the â18 dB detection limit. Therefore, excluding the grating lobe indications, it is anticipated that very few unintended indications would be identified above a flaw detection threshold of â18 dB in clean welds. 3.8.3 Recommendation for AWS A marked-up version of Annex K and associated commen- tary with the recommended changes is included in Appen- dix G. Based on the CIVA results for planar flaws along with the verification testing, it was determined to set the ampli- tude limit for flaw detection at 18 dB under the SSL (â18 dB). In Annex K, this is referred to as the disregard level (DRL) because indications with amplitudes lower than this limit during the line scans will not require additional raster scan- ning. As stated, this limit provided adequate sensitivity for all intended flaws in the verification testing, along with detecting all unintended indications that were â13 dB or greater when performing follow-up raster scanning. Therefore, this limit seems to set a good compromise between adequate detection of critical flaws and adequate sensitivity, so that the number Line Scan Amplitude Thresholds Raster Scanning Amplitude Threshold Detected at -20 dB Detected at -18 dB Detected at -16 dB Detected at -14 dB Accepted (<-13 dB) 33 18 8 2 Rejected > -13 dB 3 3 2 1 Rejected > -13 dB & 1" long 0 0 0 0 Rejected > -10 dB 1 1 1 1 Rejected > -8 dB 1 1 1 1 Rejected > -6 dB 1 1 1 1 Table 43. Number of unintended indications detected during verification testing.
63 of harmless indications which require follow-up raster scan- ning is minimal. According to the CIVA analysis, the â18 dB limit would overestimate the lowest possible amplitude from a critical flaw with skew. This may possibly result in missing a critical flaw, but based on the verification testing, it seems that the CIVA results for flaw detection were slightly conservative. For instance, the CIVA results for a flaw without skew showed that an amplitude limit of â17 dB would result in detection of all critical flaws, but the minimum amplitude measured during verification testing for a comparable flaw was actually â14 dB. Assuming the amplitude of flaws with 5Â° skew are comparably overestimated, a flaw detection limit of â18 dB should detect critical flaws with 5Â° skew as the CIVA analysis for these flaws gave an amplitude limit of â20 dB. In other words, no further reduction below SSL seems to be required. Flaws with 10Â° skew or greater may have much lower ampli- tude than the flaw detection limit proposed. For instance, the CIVA result for 10Â° skew was â28 dB and would be much lower for 20Â° skew. Setting an amplitude limit this low would result in a large amount of indications which would be acceptable and would likely result in manual raster scanning of most, if not all, of the weld. This would eliminate the eco- nomic advantage of encoded line scanning. The likelihood of planar flaws with skew is expected to be low since the LOF and incomplete penetration flaws will likely be aligned along a fusion face. Also, the current requirements for follow-up scanning for transverse flaws using scanning Pattern D or E will be retained. These requirements will allow for detection of flaws transverse to the weld axis and may aid in detecting highly skewed flaws. It is recommended that 12 dB over SSL be added to the scan- ning gain during encoded line scanning of tension welds in order to provide adequate screen height of indications greater than the flaw detection limit. After applying 12 dB of scanning gain, an indication which is greater than 40% FSH will require follow-up raster scanning, assuming that reference amplitude is set at 80% FSH. 3.9 Compression Weld Acceptance Criteria The flaw detection and rejection limits discussed previously were derived using the critical flaw sizes. These sizes were com- puted assuming that the weld is in a tension member for the FFS analysis. AWS traditionally has separate acceptance criteria for tension and compression welds for both conventional UT and Annex K. For conventional UT in accordance with Clause 6, the rejection limits for compression welds are â¼6 dB lower (less conservative) than for tension welds. There is also a slight modi- fication to the Class C length limits, as the maximum length for compression welds is 2â³ regardless of the through-thickness location, while there are tighter limits for tension welds in the top or bottom quarter of the weld thickness. For Annex K, the amplitude limits are exactly the same for compression and tension welds, but the maximum length limits are slightly different for compression and tension welds. The Class C length limits are carried over from Clause 6 with the smaller length for near surface flaws in tension welds. The Class B length limits are also slightly different, with compression welds having a maximum length of 0.75â³ and tension welds having a maxi- mum length of 0.5â³. Rather than include separate acceptance criteria for ten- sion and compression welds, the CSA W59:2018 code only includes separate acceptance criteria for statically loaded and cyclically loaded structures. Therefore, while bridges would fall under cyclically loaded, tension and compression welds would be evaluated using the same acceptance criteria. Since the critical flaw size of compression welds was not spe- cifically determined (and could not be using FFS), the accep- tance criteria for compression welds would either be based on workmanship criteria or on the results for tension welds. It is recommended that the flaw detection and rejection limits determined for tension welds be used to form the basis of compression weld rejection criteria. The research team believes that it is prudent to use the same scanning pro- cedure requirements in order to ensure that any critical flaws be detected. The proposed acceptance criteria for tension welds does not rely on length measurement for indications greater than â10 dB. Therefore, purely modifying the maxi- mum length of flaws similar to what is currently in Annex K does not seem reasonable as this would be a very low amplitude for automatic rejection of flaws in compression welds. Rather, any modification to the acceptance criteria rejection limits for compression welds should involve shifting the amplitude limit. One option would be to shift the amplitude limit +6 dB based on the shift that is in the existing Clause 6 amplitude tables. However, this would be contrary to Annex K and CSA W59:2018, which use the same amplitude limits for both com- pression and tension welds in cyclically loaded structures. Table 44 displays the raster scanned results from the raster scanned verification testing as shown in Table 36 but with flaw rejection amplitude limits of â4 dB, â2 dB, and 0 dB. The conventional UT rejection rate from the round robin testing is also shown using the AWS D1.5 Clause 6 compres- sion and tension criteria. Based on this data, setting the rejec- tion limit to â4 dB would result in rejection of all flaws which were rejectable by at least one technician per the round robin testing using the conventional UT compression tables. Set- ting the rejection limit to â2 dB would result in acceptance of a LOF flaw that was 0.03â³ Ã 0.06â³ and was rejectable by 60% of the technicians per the round robin testing using the conventional UT compression tables. Setting the rejection limit to 0 dB would result in acceptance of an HAZ crack,
64 Drawing Details Conv. UT Rejection Rate Raster Scan Results Max Amp (dB) Rejection Limit Flaw Type Flaw Height (in) Flaw Length (in) Ten Comp -10 dB -8 dB -6 dB -4 dB -2 dB 0 dB HAZ Crack 0.18 0.52 NA NA 8.0 Y Y Y Y Y Y Porosity 0.10 0.73 NA NA -6.2 Y Y N N N N LOF 0.11 0.63 NA NA 10.8 Y Y Y Y Y Y HAZ Crack 0.14 0.57 NA NA -0.6 Y Y Y Y Y N Slag 0.10 0.74 NA NA 2.3 Y Y Y Y Y Y Crack 0.14 0.37 NA NA -12.1 N N N N N N LOF 0.18 0.50 NA NA 11.8 Y Y Y Y Y Y LOF 0.12 0.64 NA NA 15.5 Y Y Y Y Y Y Toe Crack 0.19 0.49 NA NA 2.9 Y Y Y Y Y Y Toe Crack 0.11 0.61 NA NA -1.0 Y Y Y Y Y N Slag 0.09 0.92 NA NA -4.2 Y Y Y N N N HAZ Crack 0.14 0.45 NA NA 3.7 Y Y Y Y Y Y IP 0.10 0.88 NA NA 5.5 Y Y Y Y Y Y Slag 0.32 0.37 100% 60% 1.7 Y Y Y Y Y Y Slag 0.16 0.18 100% 100% 2.9 Y Y Y Y Y Y Slag 0.10 0.90 100% 100% -1.8 Y Y Y Y Y N Porosity 0.09 3.31 80% 0% -8.0 Y Y N N N N Toe Crack 0.02 0.04 40% 0% -8.4 Y N N N N N LOF 0.03 0.06 80% 60% -3.6 Y Y Y Y N N Slag 0.06 0.03 0% 0% -8.2 Y N N N N N Slag 0.17 3.61 100% 100% 0.9 Y Y Y Y Y Y Table 44. Verification testing of flaw rejection amplitude limit (compression welds). toe crack, and slag inclusion that were rejectable by all of the technicians per the round robin testing using the con- ventional UT compression tables. This is in addition to the acceptable flaws from the â2 dB rejection limit. The amplitude limits from the existing Clause 6 conven- tional UT acceptance criteria for compression welds vary depending on the incidence angle and plate thickness. Based on these amplitudes along with the results from the round robin, it seems reasonable to set the ARL for compression welds to 0 dB (i.e., equal to the SSL). While this would result in a shift of 10 dB from the tension criteria, it seems reason- able from a workmanship standpoint. It is also recommended that compression welds be evaluated based on the results from the encoded line scan rather than requiring follow-up manual raster scan of each indication. The criticality of flaws in compression welds is much lower than in tension welds and does not warrant the additional effort to maximize the signal. With these recommendations, compression welds will be essentially tested similar to how they are in the 2015 edi- tion of Annex K since evaluation would be performed on encoded line scans only. With this modification to the ARL for compression welds, the new version of Annex K would still be 5 dB more conservative than the old criteria, which sets the ARL at +5 dB. Based on conventional UT tables, it seems reasonable for the EVL to be set â¼â4 dB. This is close to the old Class C limit of â6 dB for flaws that are acceptable when 2â³ or less. There- fore, it is recommended that the EVL be set at â6 dB with a length limit of 2â³. This essentially duplicates the Class C criteria of Annex K, but are now labeled as Class B in the pro- posed version. Since follow-up raster scanning is not required for compression welds, the detection limit [i.e., disregard level (DRL)] would be set to the same amplitude as the EVL of â6 dB. Finally, the length measurements for compression
65 welds can be determined using the 6 dB drop method on the encoded line scan results similar to the current Annex K requirements. A marked-up version of Annex K and associ- ated commentary with the recommended changes is given in Appendix G. 3.10 Comparison to Radiographic Testing Currently, most tension bridge welds are only inspected with RT, except for fracture critical welds which are inspected with RT and UT. Compression welds may be tested with either RT or UT. There are select states that have replaced RT with UT for tension welds, but this is very unique. Bridge owners have traditionally preferred RT to UT due to the simple interpretation of an RT image, which can be saved and easily retrieved for permanent record. Conventional UT reports, on the other hand, are tabulated results of the indications which were detected by the UT technician. RT is known to be more sensitive to volumetric flaws such as slag inclusions and porosity while UT is more sensitive to planar flaws such as lack of fusion and cracks. This was apparent during this research project as most of the lack of fusion flaws in the round robin plates were not discern- able with RT while they were rejectable according to most conventional UT and PAUT inspections. Slag and porosity, while easily detectable with RT, had low rejection rates with PAUT according to Annex K. Since the rejection rates for these flaws were much greater for conventional UT, it is likely that the poor rejection rate for PAUT was due to differences between the acceptance criteria of PAUT and conventional UT as well as the lack of raster scanning with PAUT in order to maximize the amplitude. Therefore, while slag and porosity often have low ampli- tude with UT (conventional UT or PAUT) compared with planar flaws, it is not to say that volumetric flaws cannot be detected with UT. This is apparent in the previously reported verification testing results for flaw detection and rejection shown in Table 42 and Table 36, where setting rea- sonable amplitude limits resulted in detection and rejection of the volumetric flaws. Digital RT images of the round robin test plates are included in Appendix D. In addition to the intended flaws in the weld specimens, RT images are available for two of the three unintended flaws that were detected and rejected if the amplitude limit was set at â13 dB. One of these indica- tions (maximum amplitude of â10.8 dB during raster scan- ning) was apparent on the RT image, shown in Figure 32. It had a maximum length of 0.03â³ which would be acceptable for all thicknesses according to the RT acceptance criteria in AWS D1.5 Clause 6. The other unintended flaw (maxi- mum amplitude of â11.2 dB during raster scanning) was not apparent on the RT image since it was on the edge of the plate. If confronted with choosing either RT or UT, it seems that UT would be the preferred inspection method based on the increased sensitivity of planar flawsâwhich are more critical according to FFSâalong with the ability to detect and reject volumetric flaws. It is recommended that future research include a round robin testing program where the rejection rate using PAUT with the proposed revisions to AWS Annex K would be compared to the rejection rate for RT. Performance qualification of PAUT technicians requiring the detection and rejection of critical planar and volumetric weld flaws during an independently administered practical examina- tion would provide additional verification that PAUT is pro- viding adequate sensitivity to critical flaws of all types. 3.11 Technician Performance Qualification 3.11.1 Current AWS Requirements The round robin testing showed that there was a large amount of variability in inspection results using the PAUT Annex K and conventional UT codes. While some of this vari- ability may be due to differences in inspection equipment, equipment settings, and scanning procedures, a large portion of this variability is due to human factors and inconsistencies. For instance, it is likely that the large amount of variability in the reported location of the same flaw was primarily caused by poor calibration of encoders. It was noted that technicians Unintended Flaw Figure 32. RT image of unintended weld flaw.
66 would often consistently report multiple flaws either to the left or right of their actual location. This offset in flaw loca- tion was sometimes quite large, resulting in a large number of detected flaws which did not meet the API RP 2X require- ments for reported flaw location. The inspection variability due to human factors would likely be improved if PAUT technicians were required to pass practi- cal examinations that were administered by independent enti- ties. These proposed examinations would include inspection of flawed weld specimens using AWS D1.5 Annex K. The test plates should include known flaws that are comparable to the critical flaw size used to determine the acceptance criteria. Since AWS D1.5 and ASNT SNT-TC-1A do not currently have any guidance on critical weld flaw size, it is unknown what size flaws are included in the practical examinations performed in accordance with ASNT SNT-TC-1A . In addition, ASNT SNT-TC-1A only requires that a minimum of one flawed specimen be used for the practical examination without any guidance on the number of flaws, type of flaw, orientation of flaw, or requirement for blank specimens. It states that the flawed specimen should be representative of the component that would be tested, but that interpretation is left to the ANST Level III. Finally, no minimum require- ments for passing the practical examination are provided. It states that at least ten different checkpoints requiring an understanding of NDT variables and the employerâs proce- dural requirements should be included in the practical exam. It also states that the candidate should detect all discontinui- ties and conditions specified by the NDT Level III. Finally, it notes that while it is normal to score the practical on a per- centile basis, practical examinations should contain check- points or gateway tasks that failure to successfully complete would result in failure of the examination. The 2016 edition of SNT-TC-1A included a sample checklist for guidance on the development of practical examinations. This checklist is not specific to any method or level and may be modified as needed in accordance with the Level III. The sample checklist includes a possible breakdown of a scor- ing rubric and sample limits on flaw detection, false calls, and flaw evaluation. The ten categories listed in this sample checklist include 1. Knowledge of NDT Procedure 2. Equipment and Material 3. Test Specimen Care and Custody 4. Operations 5. Detection of Indications 6. Interpretation of Indications 7. Evaluation of Indications 8. Documentation and Records 9 General Health and Safety 10. General Observable Conduct Since there is no guidance on the flaw size in the practi- cal exam specimen, the weld flaws included in the specimen could be extremely large such that anyone who understands the very basics of PAUT inspection would be able to detect them and would pass the practical exam. In this case, the examination is not performing the intended function of testing the competency of the individual. This would be like giving a structural engineer a single question exam on steel design where they only had to compute the tension stress by dividing the given force by the given area and deeming that they are now qualified to design a bridge. ASNT SNT-TC-1A states that the technician should dem- onstrate familiarity with and ability to operate the NDT equip- ment, record, and analyze the resultant information to the degree required. The âdegree requiredâ statement is vague and leaves the decision on whether the technician has adequate performance completely to the ASNT Level III administering the exam. It also states that the Level II PAUT technician should detect all discon- tinuities and conditions specified by the NDT Level III. There are no requirements on the accuracy of flaw location measurements or limitations on the number of false calls. In the opinion of the authors of this report, discussion is merited on the self-policing of the NDT industry according to ASNT SNT-TC-1A. There are no specific requirements on the difficulty of the practical test or on the method for grading the practical test. It is the NDT firmâs advantage to have as many technicians pass the exam and be available for inspection duties. It is also difficult for NDT firms to have a large number of samples available or to have specimens which have not been used for previous tests. The test specimens may be reused from technician to technician within an NDT inspection firm or even reused for reexamination of candi- dates who previously failed (or passed) the exam. ASNT SNT-TC-1A includes no discussion on characteriza- tion of flaw type, yet AWS D1.5 Annex K requires that flaws characterized as cracks be rejected. In other words, the cur- rent training and certification program for PAUT technicians does not include any requirements on the ability to character- ize flaws, but it is expected that these technicians will be able to accurately characterize the flaw type when they perform weld inspections per AWS D1.5 Annex K. (As noted, the round robin phase of the research showed the current workforce has limited reliability in this skill.) The AWS D1.1 proposal for PAUT inspection has modifi- cations to the personnel qualification requirements, includ- ing doubling the minimum number of hours of work time experience in PAUT from 160 hours to 320 hours and requir- ing that the practical exam consist of at least two flawed specimens representing joint types to be examined with each specimen containing a minimum of two flaws. The research team believes that doubling the minimum number of hours of work time experience is unlikely to result in a
67 large improvement in PAUT inspection quality. In fact, it may actually have the opposite effect as a technician who is not properly performing any given task will become more entrenched in the wrong practice and become more confident that he or she is actually doing it correctly. In short, requiring a PAUT technician (or any individual performing any task) to perform additional work time experience does not necessarily mean that the technician will perform âbetter,â as it is unclear whether the technician is performing the inspection correctly in the first place. For example, a technician who has not prop- erly demonstrated that they can detect and reject critical weld flaws but has 320 hours of work experience incorrectly per- forming PAUT inspections is unlikely to be any better than he was after he completed the first 160 hours of incorrect PAUT inspections. Rather than doubling the required work time experience, improving the meaningfulness of the practical examination and setting more defined requirements for pass- ing the practical examination would likely result in greater improvement in technician performance. While there is merit to increasing the number of flawed specimens and the number of flaws tested during the prac- tical examination, it does not seem reasonable to require a minimum number of flaws per flawed specimen as this only provides the candidate with a minimum number of hits per plate that they need to find. It would be much better to have a random set of specimens with various number of weld flaws mixed with blank specimens which do not have any flaws. This way the candidate does not know how many flaws there are per plate and does not expect that there should be a flaw in each plate. This results in a much more realistic practical exam since the number of flaws in a weld is always unknown in an inspection. Other guidance on the development of a performance test to evaluate reliability of a nondestructive testing system is given in MIL-HDBK-1823A . Some states have started to recognize the need for improved practical examination of UT technicians. For example, New York State Department of Transportation (NYSDOT) has an Ultrasonic Testing Technician Program included in their Steel Construction Manual . This program requires prospec- tive UT technicians who wish to be certified by NYSDOT to pass examinations provided by NYSDOT. These examinations include an open and closed book exam worth 25% each as well as a practical exam worth 50%. NYSDOT also keeps a list of Certified UT Technicians available on their web page. One of the administrators of this UT certification program stated that a large percentage of prospective UT technicians have failed their stateâs exam. In fact, they are lessening the required time between initial test and the retest in order to make it easier for retesting. This case study highlights the need for independent examination of UT technicians. Texas Department of Transportation (TxDOT)  also includes a hands-on examination administered in-house in addition to the requirements of the AWS code and employerâs Written Practice. A TxDOT official who administers this exam was contacted to confirm the type and size of flaws used in this exam. TxDOT has a specific plate for each geometry, including T and corner joints and a thickness transition butt weld. Each plate has multiple flaws that are of various types, including both planar and volumetric flaws. These flaws are typically Class A rejectable defects in accordance with Clause 6 tension weld conventional UT tables. While the NYSDOT and TxDOT programs are for addi- tional practical examination of conventional UT technicians, Florida Department of Transportation (FDOT) is currently implementing a program to test PAUT technicians. This pro- gram involves practical examination of each PAUT techni- cian using an in-house test block and QA inspector trained in PAUT. The flaws chosen for this block are specific to the response of FDOTâs internal research. An FDOT official stated that this qualification was deemed necessary based on the level of training that technicians were receiving in typical PAUT training courses. Since FDOT is interested in replacing RT with PAUT, the selection of flaws for the practical exami- nation was determined to be critical and was one of the rea- sons that this qualification is performed in-house. 3.11.2 AWS Recommendations As PAUT use becomes more prevalent, it will be important for the PAUT technician qualification requirements to be stan- dardized across individual agencies. In speaking with these states, it became clear that having an independent central orga- nization that would administer a reasonable practical exami- nation using Annex K would be beneficial. By standardizing the practical examination, PAUT technicians would not need to take a separate exam to satisfy the individual requirements of each agency. Rather than having many separate examina- tions with one flawed plate, a standardized examination could be much more thorough and still take less time overall. A good example of a thorough, standardized qualification requirement is the Certified Welding Inspector (CWI) exami- nation. Instead of having separate exams for each agency, pro- spective CWIs only need to qualify through AWS. The AWS CWI examination process includes a three-part examination that extends over a full day. Two parts of the exam cover weld- ing background (closed book) and code requirements (open book) while the third part is a practical examination using weld specimens. The CWI examination is set by AWS and proctored at defined locations throughout the year. The CWI examination has a controlled level of difficulty and clearly defined expecta- tions that covers visual examination of welds for all agencies. The authors recognize that requiring independent practi- cal examination of PAUT technicians will result in additional cost but believe that this may be necessary to implement the
68 removal of RT requirements in lieu of in-depth PAUT inspec- tion. The researchers are confident that a meaningful practi- cal examination could be performed in a single day. The research team believes that additional research should be performed to develop a PAUT technician qualification program which utilizes independent and consistent evalu- ation of technicians. This study could compare practical examination results to shop performance in order to develop a quantitative practical examination that will adequately evaluate candidates for bridge weld inspections. Based on the research and what was discussed previously, the perfor- mance qualification requirements in the following sections are proposed. 220.127.116.11 Practical Examination Parameters The practical examination should involve meaningful per- formance testing that is conducted by an independent party. The research team recommends that, optimally, the test would be administered by a third party that has developed a realistic test procedure that is acceptable to all agencies. The practical examination should evaluate detection, loca- tion, and rejection of realistic weld specimens with critical flaws. In addition, the practical examination should include evaluation of following parameters: â¢ Familiarity and application of Annex K requirements â¢ Development of adequate scan plan and documentation of essential variables â¢ Use of proper equipment â¢ Proper calibration for reference sensitivity and acoustic properties â¢ Proper application of the two-part inspection procedure (line scan and follow-up raster scanning) â¢ Proper documentation and reporting 18.104.22.168 Specimen Details The specimens used in the practical examination should be fabricated from steel that have acoustic properties representa- tive of the typical steels encountered in bridge weld inspec- tion but also include some specimens in which calibration adjustments are required to account for variation in acoustic properties. The specimens should be representative of the con- figuration of the welds that will be inspected by the technicians during future applications of Annex K. This includes thickness transitions and weld reinforcement if both will be encountered. The number of flaw specimens shall be large enough such that the number of flaws per plate will be varied, and blank specimens should be included in the lineup of plates to be tested. The technicians should be instructed to detect and report all relevant indications and should be informed that a plate may not necessarily contain any flaws. The research team believes that, based on the previous research, including the round robin, at least four weld specimens should be used with a weld length of approximately 18 inches. Weld specimens that are too short may not adequately capture errors from encoded line scanning. More weld specimens may be necessary in order to adequately include possible weld configurations and acous- tic properties (proper calibration practices) that technicians may encounter during weld inspection. The number of flaws included in the weld specimens should be adequate to verify satisfactory performance. Both volumet- ric and planar discontinuities should be included in the set of weld flaw specimens. It is recommended that at least one slag inclusion, porosity grouping, lack-of-fusion flaw, and verti- cal planar flaw (e.g., crack or incomplete penetration) should be included. Variations in the flaw size and through-thickness location will also be necessary in order to verify that the scan- ning procedures for flaw detection and rejection are prop- erly being followed. The absolute minimum number of flaws included in the practical exam is recommended to be five. Use of more flaws and weld specimens would allow for variation in flaw size and location (through-thickness depth, tilt, and location along weld axis), which would provide additional information on the scatter of results from human factors. 22.214.171.124 Pass/Fail Criteria It is believed that the formulas and minimum performance levels included in API RP 2X are appropriate for the bridge industry for evaluating flaw detection, sizing, and location during the practical examination. API RP 2X includes mini- mum recommended performance levels for UT technicians using the following two formulas: 100 Formula 1 R 1 100 Formula 2 1 1 P L L L L L L c a c f = Ã = ï£«ï£ ï£¶ ï£¸ â ï£« ï£ ï£¶ ï£¸ Ã Where: P = percentage of actual reflectors correctly detected and sized, 0 to 100 R = overall rating including penalty for false calls, 0 to 100 La = length of actual reflector contained in the test plate Lc = credited length for indications that have been cor- rectly sized and located (credit is given for the lesser of the reported length or actual length of the flaw.) L1 = accumulative length of all indications by the techni- cian, right or wrong Lf = accumulative length of indications above the stated disregard level where no reflector exists
69 To be correctly sized, this document recommends that the reported dimensions be within a factor of two of true dimensions (one-half to twice the actual dimension). To be correctly located, this document recommends that the centerline of the reported indication be within the boundary of the actual indication or within Â½ inch of the actual center- line of the indication (whichever is greater). API RP 2X sug- gests that a score of 70 or above for Formula 1 and a score of 50 or above for Formula 2 be used as minimum performance for ultrasonic technicians. When the round robin test results were used to calcu- late the performance of each conventional UT and PAUT technician, two of the five conventional UT technicians and zero of the four PAUT technicians met these levels. The primary reason for the PAUT technicians not meeting the minimum performance was due to errors in flaw location. As stated, this is thought to be due to issues with the use of encoders. Since the flaw has to be detected in order to count towards the performance requirements, the lower sensitiv- ity with Annex K prior to the proposed modifications (no separate flaw detection limit for the line scans) will have also contributed to some of the low scores. The primary reason for the conventional UT technicians not meeting the minimum performance was due to poor accuracy for the flaw length measurement and some misses of long scattered porosity. The issues with flaw length measurement are likely due to the manual method for length measurement with conventional UT, which lends itself to overestimating the actual length. While the minimum scores provided by API RP 2X could be used to evaluate flaw detection and flaw location, the pass/fail criteria of the practical examination should also evaluate the following items in determining an overall practical examination score: â¢ Familiarity and application of Annex K requirements â¢ Development of adequate scan plan and documentation of essential variables â¢ Use of proper equipment â¢ Proper calibration for reference sensitivity and acoustic properties â¢ Proper application of the two-part inspection procedure (i.e., line scan and follow-up raster scanning) â¢ Evaluation of indications using proper acceptance criteria â¢ Proper documentation and reporting Using a checklist similar to that provided in the 2016 edition of ASNT SNT-TC-1A, a rubric should be developed that incor- porates each of these items. By assigning points to each item, along with flaw detection and location criteria, technicians passing the practical examination will demonstrate adequate comprehension and application of the Annex K inspection procedure. Suggested performance testing guidance has been added to the proposed modifications to Annex K found in Appendix G of this report. 3.12 Flaw Sizing Acceptance Criteria for Alternative UT Methods An acceptance criteria based on flaw height and length sizing is provided as an alternative method to the amplitude approach provided previously. The alternative method requires that the PAUT technician develop a written procedure according to specified requirements as well as demonstration of the procedure through performance testing on samples of simi- lar material and with flaws similar to the rejectable size. This allows for other advanced ultrasonic methods such as FMC/TFM PAUT or TOFD to be used to inspect bridge welds provided that they can detect and reject critical weld flaws. It is envisioned that the requirements for alternative ultra- sonic systems and the accompanying acceptance criteria based on flaw sizing would be included in the main body of AWS D1.5, rather than included in Annex K. Annex K includes all necessary requirements and acceptance criteria for application of encoded line scanned PAUT. Annex K does not rely on accurate flaw height sizing and, therefore, does not require development and qualification of an individual flaw sizing procedure for the specific equipment and applica- tion method. CSA W59-18  includes minimum requirements for alternative ultrasonic systems in the main body of the code. Alternative ultrasonic systems include encoded PAUT and TOFD in W59, as only manual raster scanned PAUT are allowed to replace conventional UT without additional performance testing. Because AWS D1.5 Annex K covers encoded PAUT using amplitude and length acceptance cri- teria, it is recommended that the additional requirements for alternative UT methods in AWS D1.5 apply to any methods that do not fit within the requirements of conventional UT Clause 6 or Annex K. 3.12.1 Minimum Requirements The following requirements from CSA W59 are recom- mended for incorporation into AWS D1.5: â¢ Written agreement by the engineer and contractor prior to the examination allowing the use of the alternative inspection method â¢ Certification of Level II or III in accordance with ASNT SNT-TC-1A for the specific method, if applicable; for instance, ASNT SNT-TC-1A includes qualification require- ments for TOFD, but does not include additional qualifica- tion requirements for FMC/TFM PAUT
70 â¢ Documentation of inspection procedures in writing in accordance with recognized standards and accepted in writing by the engineer â¢ Written procedures that contain at minimum the follow- ing information: â Specific operator training requirements â Types of weld joint configurations to be examined â Acceptance criteria â Type of UT equipment (manufacturer and model number) â Type of transducer, including frequency, size, shape, angle, and type of wedge â Scanning surface preparation and couplant requirements â Type of calibration test block(s) with appropriate refer- ence reflectors â Method of calibration and calibration interval â Method for examining for laminations prior to weld evaluation â Scanning pattern and sensitivity requirements â Methods for determining discontinuity location, height, length, and amplitude level â Transfer correction methods for surface roughness, surface coatings, and part curvature, if applicable â Method of verifying the accuracy of the completed examination; this verification may be by reexamination using UT by others, other NDE methods, macroetch specimen, gouging, or other visual techniques accepted by the engineer â Documentation requirements for examinations, includ- ing any verification performed â Documentation retention requirements â Demonstration of the system in order to achieve engi- neerâs approval ï¿½ System demonstration on one or more demonstra- tion blocks simulating the weld ï¿½ Demonstration of adequate and repeatable detec- tion of typical weld flaws and useful to set threshold parameters â Acceptable performance of system demonstration should include: ï¿½ Detection of all of the flaws in the demonstration block(s) ï¿½ Recorded flaw sizes for critical weld flaws with reported flaw size that exceeds the acceptance criteria limits ï¿½ Recorded flaw sizes for subcritical weld flaws with reported flaw size that is reasonably accurate 3.12.2 Acceptance Criteria The acceptance criteria used for the alternative UT meth- ods, which uses measurements of flaw height and length, should be comparable to the acceptance criteria used in the recommendations for Annex K acceptance criteria for planar and volumetric flaws. As described in Section 3.7, the accep- tance criteria for Annex K was developed using the critical flaw sizes computed using fracture mechanics. As shown in Table 33, the â10 dB amplitude limit for flaw rejection was based on CIVA analysis of a 0.05â³ Ã 0.10â³ planar surface flaw. This flaw was comparable to the 0.06â³ Ã 0.08â³ critical planar surface flaw for a 4 ksi stress range thickness transition weld and an 8 ksi stress range equal thickness weld. The controlling planar embedded flaw had an amplitude of â8 dB, according to CIVA, using a 0.15â³ Ã 0.15â³ planar embed- ded flaw. This flaw was comparable to the 0.17â³ Ã 0.17â³ critical planar embedded flaw for a 4 ksi stress range thickness transi- tion weld and the 0.14â³ Ã 0.14â³ critical planar embedded flaw for an 8 ksi stress range equal thickness weld. For the acceptance criteria used for the alternative UT methods, it is recommended that the flaw height and length limits be provided for various a/c ratios. The results for the critical planar flaws noted above were summarized into Tables 45 and 46. Table 45 includes a recommended accep- tance criteria for planar embedded flaws measured using alternative UT methods. Table 46 includes a recommended acceptance criteria for planar surface flaws. As shown in Table 34, the â13 dB amplitude limit for rejec- tion of flaws 1â³ long or greater was based on CIVA analysis of a 0.125â³ diameter embedded volumetric flaw and a 0.08â³ diameter surface volumetric flaw. The embedded flaw was comparable to the 0.14â³ diameter critical volumetric embed- ded flaw for an 8 ksi stress range equal thickness weld. The surface volumetric flaw was comparable to the 0.08â³ Ã 0.11â³ critical volumetric surface flaw for a 4 ksi stress range in a thickness transition weld and a 0.06â³ Ã 0.08â³ critical volu- Flaw Height (in) Flaw Length (in) Interpolated Length (in) for Intermediate Height (in) 0.02 1.00 - 0.03 0.20 For 0.03<H<0.06, L=0.31-3.67*H 0.06 0.09 Table 46. Alternative UT methods acceptance criteria for planar surface flaws. Flaw Height (in) Flaw Length (in) Interpolated Length (in) for Intermediate Height (in) 0.06 1.00 - 0.07 0.28 For 0.07<H<0.10, L=0.56-4.0*H 0.10 0.16 For 0.10<H<0.14, L=0.21-0.5*H 0.14 0.14 Table 45. Alternative UT methods acceptance criteria for planar embedded flaws.
71 metric surface flaw for an 8 ksi stress range in an equal thickness weld. Since these flaws are the same sizes as the critical pla- nar flaws, it is recommended that the limits provided in Table 45 and Table 46 also be used for the maximum size of an individual volumetric flaw (i.e., maximum slag inclu- sion or pore). In addition, a maximum length of 1â³ is recom- mended for scattered indications such as a group of porosity. This is similar to the length requirement recommended for Annex K for flaws greater than â13 dB but less than â10 dB. Since the inputs to the FFS study were for tension welds. The limits given in Table 45 and Table 46 will be overly conservative for compression welds. The recommendation for the Annex K amplitude limits for the compression acceptance criteria uses a 0 dB amplitude. This amplitude approximately correlated with the maximum amplitude of vertical 0.10â³ Ã 0.10â³ planar surface flaw and a vertical 0.20â³ Ã 0.20â³ planar embedded flaw from CIVA analysis. Note that this amplitude was based on tra- ditional conventional UT amplitude limits rather than fracture mechanics. For compression welds, a 50% increase in the criti- cal flaw height and length given in Table 45 and Table 46 seems reasonable to compare with the Annex K acceptance criteria. Another strategy would be to set this limit based on a strength requirement and an acceptable amount of unfused material.