Automated ultrasonic testing (AUT) data processing techniques were evaluated and computer simulation was used for improving flaw detection, sizing, and justifying the inspection procedures. Probability of detection (POD) and accuracy of flaw sizing was determined for six representative AUT systems with representative girth weld implanted flaws. Quantification methodology guidance was built and measures to improve the AUT performance and quantification process were recommended for future consideration.
Advanced and quantitative nondestructive evaluation (NDE) allows the use of acceptance criteria based on experimental destructive testing (DT) or fracture mechanics or other fitness-for-purpose assessments. DT or engineering critical assessment (ECA) approaches are used to determine allowable flaw sizes depending on material/weld properties and loading conditions during pipeline installation and operation. Typically, DT, or an ECA, gives the allowable flaw sizes (smaller than critical flaw sizes, as safety factors are used) that can be expressed as an allowable flaw height correlated to flaw length. In general, measurement of flaw height and length using NDE is uncertain, and the sizing errors encountered must be taken into account when calculating the acceptable flaw sizes from the allowable ones, as an assurance against accepting non-allowable flaws with a sufficiently high confidence. Quantitative performance data specific to girth weld automated ultrasonic testing (AUT) tools is unavailable in public domain from independent investigators. This leads to ECA and strain-based approaches being excessively or insufficiently conservative in assumed AUT uncertainty and detection capabilities. Further, the NDE reliability determination is an integral part of the applied research efforts to validate the capabilities of new NDE techniques, instruments, and procedures for an intended application. For practical purposes, the main criteria used in UT performance quantification and reliability are the probability of flaw detection (POD) and accuracy of sizing during flaw characterization. An often quoted requirement for POD is that it should be at least 90%, when relating to allowable flaw sizes at a 95% confidence level (90/95 rule).