Low frequency ultrasonic torsional T(0,1) guided waves allow for the full volume of tens of meters of a pipe to be inspected from a single measurement position. Currently, these inspections are primarily carried out with detachable transducer rings and data collected is evaluated manually by the operator. Permanently installed guided wave sensors have been introduced by Guided Ultrasonics Ltd. in 2005 and since have been deployed on thousands of pipes. Measurements collected with these sensors are, in the current framework, evaluated in the same manual fashion as those obtained from detachable transducer rings. The contents of this thesis focus on the development and evaluation of a novel SHM procedure for guided waves in pipes utilising existing permanently attached transducer ring technology. This work can be subdivided into four main fields of investigation.
Firstly, a blind trial was conducted in order to test the sensitivity of a structural health monitoring (SHM) algorithm based on independent component analysis. It was found that compared to standard one-off inspections an improvement of approximately a factor of 5 could be obtained. Defects introduced on straight sections of a test pipe either before or after a 1.5D bend were identified with cross-sectional area losses between 0.5% and 1.5%. The sensitivity of the SHM algorithm to a defect located on a pipe bend was found to be significantly lower with a 3.5% cross-sectional area loss at the time of first detection.
The reduced sensitivity to the defect located on the pipe bend was investigated in the second part of this thesis utilising finite element (FE) simulations. It was shown that the sensitivity of guided wave inspections using the T(0,1) mode is highly dependent on the location of a defect both in circumferential and axial position, as well as the size of the bend itself. This sensitivity is proportional to the distribution of the squared von Mises stress across the structure. The results of this study allowed for the conclusion to be made that the reduced sensitivity during the blind trial was caused by the geometry of the inspected structure and the propagation of the T(0,1) wave mode rather than a shortcoming of the SHM algorithm.
The third main part of this thesis investigates the potential of using defect reflections obtained from FE simulations in combination with real measurements of a structure in a constant state in order to estimate the sensitivity of an installed SHM system. The validity of this method was verified by the use of measurements obtained from the blind trial setup. The growth patterns of the defects introduced into the pipe setup were replicated using FE simulations and superimposed onto measurements of the trial pipe in its baseline state. The applied SHM algorithm was found to have the same sensitivity to these synthetic datasets as to those containing the real defect growth.
The final part of this thesis discusses a novel temperature compensation approach developed for application to results obtained from the SHM procedure. It was found that, even if a stretch based temperature compensation algorithm is applied to guided wave measurements, temperature related variations in the coherent noise floor of the inspected measurements as well as frequency response changes in the signals transmitted by the transducer ring will continue to be present. A newly developed compensation procedure allows for these variations to be significantly reduced enabling lower rates of false calls as well as an improved probability of detection for small defects.