Ultrasonic thickness testing of aging offshore structures

Abstract

The objectives of this thesis concern the use of ultrasonic thickness (UT) testing for use in the offshore industry. Evidence from prior studies conducted at Texas A&M University suggests that the corrosion on the surface of offshore structural members is not distributed in a random fashion. It was therefore desired to study the matter more extensively to 1) determine the feasibility of using ultrasonic thickness measurements for assessment of corroded members in offshore structures, 2) determine the amount of data needed for meaningful assessment, and 3) identify any common patterns of corrosion in offshore structural members, which might be used in designing a more effective assessment protocol. First, three specimens from an earlier study were available for use here. These "Riverside Specimens'' were subjected to extensive UT measurements. An important statistical tool, Analysis of Variance (ANOVA), was used to determine the probability that thickness variations along the length and around the circumference of each member could be a random event. Both longitudinal and circumferential non-random variations were found in some of these members. A study of reduced sample size confirmed that reduction of data caused more uncertainty in the results. Next, a field study was conducted on recently salvaged offshore jackets in Morgan City, LA. Six tubular bracing specimens (three horizontal, three diagonal) near the splash zone were evaluated. The statistical evidence for corrosion trends was found to be rather weak. Nonetheless, it was possible to make certain generalizations. In particular, the outside facing sections tend to be the most corroded (thinnest), while the surfaces toward the center of the structure tend to be the least corroded (thickest). This also agrees with industry observations. Finally, the effect of sample size on the detection of strength loss was observed. Using some basic assumptions regarding sample mean and standard deviation, it was shown that even with greatly reduced numbers of measurement points, one may still obtain reasonable estimates of critical thickness values corresponding to certain strength loss ratios.