Description of Welded Joints and Inspection Process
Non-destructive testing (NDT) inspector performance demonstration qualification (PDQ) is now required by most major structural codes including American Petroleum Institute (API), American Society for Mechanical Engineer Boiler and Pressure Vessel Code (BPVC), and American Welding Society (AWS) Bridge Welding Code (BWC). The objective of the PDQ for an NDT inspector is to confirm their proficiency in applying an NDT procedure, from calibration through reporting, to detect, size, and characterize material defects at a sufficient performance level, or confidence level, for the results to be compliant with desired risk assessment and probabilistic analyses.
Figure 1: Non-destructive testing of bridge poses physical challenges that may impact NDT inspector performance.
PDQ exams, and probability of detection (POD) studies, are therefore designed to quantify an NDT inspector’s ability to detect, size, and characterize at a compliance level with the relevant risk based inspection strategy. The NDT industry now has tremendous performance data on inspector performance as PDQs are now required and POD studies are becoming more frequent. But what about inspector performance in uncontrolled environments? For example, what kind of performance variation may be observed, or expected, on structures with unknown anomalies inspected by different inspectors over many years. This article explores how inspector performance varies on the same steel bridge butt welds over a 20-year period. The article analyzes the performance of 10 different inspector performances over this time frame in terms of detection and sizing.
Description of Welded Joints and Inspection Process
The inspected welded joints were complete joint penetration (CJP) welds connecting components loaded in tension. Most of the joints transitioned from a thicker to thinner plate with individual plates ranging from 1.0” to 2.5”. Double V joints were assumed. The weld lengths were approximately 24” long.
The inspection process followed the guidelines outlined in AWS D1.5 BWC which are summarized in detail in an earlier article (https://www.techknowserv.com/post/ultrasonic-testing-of-welds). A 2.25 MHZ ultrasonic transducer was used in conjunction with a 70 degree weld for the shear wave inspections. The inspector was responsible for calibrating the instrument, scanning the joints in compliance with the procedures and documenting the indication rating (IR), indication length and determining if the indication passed or failed based on BWC guidelines. The surface preparation was consistent for each inspection. A wire wheel brush was used to remove any coating from the weld metal, heat affected zone, and base metal permitting ultrasonic inspection in the second leg.
Comparison of Inspector Performance Across All Welds
Over a 20-year period the same 8 welds were inspected by 10 different inspectors at different time intervals ranging from 1 year to 5 years. Over this 20-year period, there was extreme variation in the total number of indications detected across all inspectors as shown in Figure 1. The total number of weld indications reported by each inspector varies from 3 to 33. The total number of failed indications reported varied from 1 to 21 demonstrating diverse results for that parameter as well.
Figure 1: Summary of total indications found by each inspector on tension joints.
The Agreement Matrix summarizes how often two different inspectors detected the same weld indications. For example, Inspector 3’s reported indications agreed with Inspector 1’s over only 34% of the total indications reported by both companies. Like the total number of weld indications detected across inspectors, a large variation in inspector agreement is observed between inspectors. Figure 2 presents the Agreement Matrix with the results color-coded green and red for over and under 50% agreement, respectively. The results show consistent disagreement versus agreement.
Figure 2: Summary of total indications found by each inspector on tension joints.
Comparison of Inspector Performance Across Specific Welds
Inspector performance across all welds in terms of total indications reported and agreement between specific weld indication varied greatly as presented above. This section analyses how inspector performance varies across weld indications that were detected by almost all inspectors. Over the 20-year period, a total of 93 different weld indications were reported by all inspectors. The 93 welds were analyzed further to determine their exact location in the weld relative to a known reference point. Of these 93 weld indications, 3 indications were detected across all inspectors at comparable locations in the welds. These indications were analyzed further to determine inspector performance variance on Indication Rating and indication length.
The Indication Rating results for these 3 indications are shown in Figure 3. Indication 1 was reported on by Inspectors 1 through 7 with an IR ranging from 2 to 10. Indication 2 was also detected by Inspectors 1 through 7 with and IR between 6 and 13. Indication 3 was detected by all inspectors except for Inspector 5. Again, a large variation in reported IR is observed. The inspector performance variation in measuring and calculating IR impacts the inspection pass/fail results. In AWS BWC, the Indication Rating and measured length are the deciding factors to pass or fail a weld. The outcome of the AWS BWC evaluation is color coded in the Figure 3 with green and red indicating pass and fail, respectively. Figure 3 shows a significant disagreement across all inspectors on IR and AWS BWC pass/fail results.
Figure 3: Variation in Indication Ratings reported by inspectors
A similar analysis was applied to the measured length of the same indications and is shown in Figure 4. Indication 1 was reported on by Inspectors 1 through 7 with an IR ranging from 2 to 10. Like the IR results in Figure 3, a large variation in reported lengths is observed. The length of Indication 1, 2, and 3 range from 0.25” - 1.00, 2.5 – 11.5”, and 0.25” – 4.5” respectively. Figure 4 shows a significant disagreement across all inspectors on length and AWS BWC pass/fail results.
Figure 4: Variation in Indication Length reported by inspectors
Further analysis was applied to the convergence of inspector performance with respect to indication severity, or Indication Rating and indication length, in AWS BWC. The outcome of this analysis is shown in Figure 5. First the indication detection rate of the was determined and presented in the left most column of Figure 5. The value in this column represents how many different inspectors detected the identical indication. The average IR and length were calculated for each detection rate level. The Average IR and Length, for example, were 8.38” and 0.83” for weld indications detected by only single inspector. As the detection rate increases from 1 to 7, there is no discernible pattern observed in the Average IR. This result may be considered counterintuitive since an increase in IR would suggest an increase in weld indication physical size or preferential orientation relative to the impinging shear wave. In contact ultrasonic testing, however, the amplitude is often an unreliable and inconsistent feature due to many factors including ultrasonic wedge alignment, surface roughness, and uniformity of couplant between the ultrasonic transducer and test surface.
Figure 5, however, shows that as the average length of the reported indication increases the detection rate also increases. For indications with detection rates 3 to 7, the average length increases from 0.13 to 2.98”. The likelihood of detection increases with weld indication length and finally some inspector performance convergence is noted.
This article evaluates non-destructive testing technician performance in an uncontrolled environment in contrast to controlled non-destructive testing probability of detection (POD) or performance demonstration qualification (PDQ) processes. Inspector performance on steel bridge CJP welds in tension member was analyzed in terms of indication detection rates, indication ratings, and indication lengths. The analysis shows significant inspector performance variability across all these variables and pass/fail results. Some inspector performance convergence was observed for detection rates on the longer weld indications.