Generating a tensile strain capacity (TSC) prediction model is a difficult challenge in applied mechanics. Because current models are relatively new and extensive strain-based design (SBD) pipeline service experience does not exist, rigorous model validation using full-scale tests (FSTs) is paramount. The lessons learned from 159 FSTs were presented previously and the data base has grown to 173 tests. This data base is used to assess the accuracy of a relatively new TSC prediction model. The new model simulates a single, surface breaking weld flaw; however, some of the FSTs contained interacting or embedded flaws or unintentional weld defects, while others failed by brittle fracture, and still others experienced welding problems rendering them unsuitable for model validation. Of 173 tests, a smaller number (122, 101, or 89 depending on the goal) is used for comparison to the new model. This paper describes (1) the importance of reliable FSTs, (2) how the 173 tests were judged for suitability in model accuracy assessment, and (3) the use of the FST data to develop a safety factor for strain-based engineering critical assessment (SBECA). The safety factor is generated from a 95% upper confidence limit on the ratio of predicted-to-measured TSC. The safety factor is 1.88. Using the new model and this safety factor, a TSC prediction equation is provided for use in SBECA. The practical meaning of this is that if either TSC or tolerable defect size is calculated using the new model, then the probability of being non-conservative is estimated to be 5%.
A BERT-Based Approach for Extracting Prerequisite Relations among Wikipedia Concepts
