Abstract

Recently, there has been a move toward the use of ultrasonic-based permanently installed systems to monitor the health status of a number of structures, such as pipes. A big advantage of this approach compared to the typical one-off inspection configuration is the potential for detection of damage at earlier stages. This is enabled by the frequent collection and analysis of data. Once the signals are corrected for the effects of temperature, the trends of the measurements taken at each structural location can be monitored, and deviations from the expected behavior can be related to the occurrence of damage. In this setting, fault-detection methods such as the ones widely used in the field of statistical process control can be used to detect those deviations. In this work, two methods based on the Generalized Likelihood Ratio are investigated and are applied to a set of T(0,1) guided wave signals collected by a pipe monitoring system. Receiver operating characteristics curves are used to identify a suitable threshold for the method to operate based on the desired levels of probabilities of false alarm and of true detection at given defect sizes. The method promises a substantial improvement in the detectability of small defects, while minimizing the human workload needed to actually inspect the signals.

How to Cite:

Mariani, S. . & Cawley, P. ., (2019) “Generalized likelihood ratio tests for defect detection in SHM of pipes”, Review of Progress in Quantitative Nondestructive Evaluation .