Nondestructive evaluation of ferromagnetic critical water pipes using pulsed eddy current testing

Publication Type:
Issue Date:
Full metadata record
Files in This Item:
Filename Description Size
01front.pdf1.1 MB
Adobe PDF
Modern day maintenance of infrastructure demands significant attention to structural health monitoring. Assessment of surface condition alone is insufficient for health and strength assessment, creating the necessity to evaluate the integrity of subsurface regions through Nondestructive Evaluation (NDE). This thesis focuses on approaches to solving the problem of condition assessment of critical pipes, i.e., large diameter high-pressure pipes owned and managed by water utilities to distribute consumable fresh water to customers, by developing techniques for representing the geometry of electrically conductive ferromagnetic materials via Pulsed Eddy Current (PEC) sensors. The main contribution of this thesis is a novel detector coil voltage decay rate based PEC signal feature, the fundamental behavior of the feature is analytically described and experimentally validated. The feature has a convenient advantage in practical application since it is directly extractable from raw PEC signals and demonstrates significant invariance to sensor shape, size, and lift-off. The feature behavior is exploited in two estimation approaches, in situ measurements on pipes are performed and pipe wall thickness is inferred with uncertainty. Firstly, an analytical approach to learning a function mapping the decay rate feature to test piece thickness with the aid of signals captured on calibration blocks is presented. The requirement of fabricating calibration blocks to have material properties matching those of pipes is extremely challenging. Thus, combining ultrasound measurements together with PEC is proposed to address material variations. Secondly, a numerical NDE semi-parametric estimation approach is presented, PEC sensor signals are simulated taking into account measured electrical and magnetic properties of materials being tested. The thickness-feature function is learned probabilistically using Gaussian Process. Unlike in the analytical approach, the function is learned non-parametrically, therefore, variations and marginal nonlinearities are captured. The advantages over the analytical approach are demonstrated in terms of improved accuracy of inferred material thickness. Finally, the resolution of commercial PEC sensors employed on pipes is identified as a limiting factor for structural integrity assessment. A numerical study on optimizing PEC sensor architecture to achieve higher resolution while maintaining sufficient penetration capability is carried out and a framework which can be used to perform 3D profiling by means of joint inference of thickness and lift-off is proposed.
Please use this identifier to cite or link to this item: