39 Identifying irregular and erroneous chemical composition data

Abstract

Nondestructive testing (NDT) of chemical composition is a critical component of the Pacific Gas and Electric Company’s (PG&E) materials verification program. Additionally, 49 CFR § 192.607 states that the operator must “conservatively account for measurement inaccuracy and uncertainty using reliable engineering tests and analyses.” Accurate and precise NDT composition data can be used to determine or verify certain characteristics of a pipe, for example vintage, grade, or manufacturing process. However, it has been observed that composition measurements may at times be inconsistent across various field analytical tools, possibly due to the variability of experimental, environmental, and other statistical (random) factors. Irregular or erroneous field NDT measurements are problematic because they could lead to the mischaracterization of pipe features during the materials verification process. 

In this paper, we present a systematic procedure rooted in data science for the analysis of field NDT chemical composition data. First, NDT composition data for a set of pipe features are collected using field analytical techniques, such as optical emission spectroscopy (OES), laser induced breakdown spectroscopy (LIBS), X-ray fluorescence (XRF), and laboratory analysis of filings by atomic absorption (AA) and combustion. Next, the data are statistically analyzed using only the measurements in the incoming dataset and, when possible, are compared to measurements obtained for the same pipe features using other NDT composition analysis techniques. Following this analysis, the data are compared to historical composition data previously obtained for features with similar attributes—such as outer diameter (OD), seam type, and nominal wall thickness (NWT)—to identify potentially erroneous measurements. Finally, we present case studies illustrating the application of the proposed process to data obtained for pipe features at PG&E with abnormally high manganese measurements, and we demonstrate how the identification of the elevated manganese values as anomalous mitigated potential downstream challenges during the materials verification process. 

Read/Download PaperGradeAvailable after the event