Guided ultrasonic wave tomography of a pipe bend exposed to environmental conditions: A long-term monitoring experiment

Abstract

Continuous monitoring of corrosion damage in pipelines requires sensor systems that can achieve high precision (repeatability) to detect and size subtle wall thickness (WT) losses. In fact, corrosion typically progresses at a rate of 1 mm per year or less which poses very demanding performance requirements if the state of the pipe has to be assessed on a weekly or even monthly basis. Guided ultrasonic wave tomography (GUWT) has emerged as an attractive approach for continuous monitoring owing to its ability to map WT losses over an extended pipe section. This is possible because GUWT employs ultrasonic waves that are guided by the pipe wall to travel a large distance from the source transducer. Defects along the path of the guided wave cause a perturbation of the signal which is then interpreted by model-based inversion schemes to determine the WT loss. In addition to damage, there is a vast number of time-dependent operational and environmental (O & E) factors that can perturb the signal also when damage is not present. To achieve high precision, it is therefore essential that GUWT can detect the changes in the signal due to damage and suppress those caused by other benign factors such as temperature variations. Although numerous studies have been conducted to estimate the accuracy of various implementations of GUWT, precision has received less attention due to the challenges associated with reproducing realistic O & E conditions in the lab. This work focuses on the effect of temperature and presents the results of a continuous monitoring experiment conducted on a pipe bend kept outdoors and exposed to weather conditions for a period 21 months. It is shown that the pipe undergoes a relatively severe temperature cycling with typical daily peak-to-trough variations of 20 ∘ C that cover the range from -15 ∘ C to +51 ∘ C as the seasons alternate. The effect of temperature is suppressed by exploiting the spatial diversity of array measurements and combining multiple datasets measured during the monitoring period. The maximum WT loss is estimated with high precision exhibiting a standard deviation not exceeding 0.4% of the nominal WT.