193

Federal regulations governing gas transmission pipelines allow nondestructive (NDT) or destructive testing (DT) techniques to verify the material properties of pipe joints. As part of its material property verification program, the Pacific Gas & Electric Company (PG&E) routinely performs pipe grade determination using NDT strength and chemical composition measurements with associated uncertainty values. Over the course of several years, PG&E has also assembled a database of DT data for pipe joints, which includes data for pipe joints both with and without known pipe grade. For the pipe joints without known grade, it is desired to determine the pipe grade to update PG&E’s system of record. Any DT measurements have some degree of uncertainty associated with them that should be accounted for in the pipe grade determination process. However, DT data obtained for pipe joints may not include replicate measurements that could be used to estimate uncertainty (i.e., standard deviation), usually due to the cost and labor required to obtain that data. Therefore, we developed an alternative process to determine reasonably conservative uncertainty values for DT strength and chemical composition measurements.

For this process, we reviewed scientific and engineering literature, testing standards, and instrument manufacturer technical documents. We then obtained or calculated uncertainty values based on the literature and chose reasonably conservative uncertainty values within the element concentration ranges typically observed in low carbon pipeline steels. Additionally, we analyzed PG&E’s historic database of DT measurements and used a subset of the database, which consisted of DT chemical composition and strength data for 633 pipe joints with known grades, to validate our proposed uncertainty values. We did so by comparing the predicted pipe grades calculated using the DT measurements and the proposed uncertainty values with the known pipe grades. This validation process was also used to assess the sensitivity of pipe grade predictions to uncertainty values.

Read/Download PaperGradeAvailable after the event