Development of an eddy current inversion algorithm for characterization of surface defects

Abstract

There is an increasing concern in aerospace industry regarding the integrity and performance of ageing metallic structures in aircraft that have been in service for several years. Cyclic loads applied to these structures during taxiing, pressurization, taking off, depressurization, and landing promote the nucleation and growth of many fatigue cracks from the locations of stress concentration. These cracks shall be detected and sized reliably by non-destructive evaluation (NDE) methods before they reach a critical size, which may lead to disastrous failure of components. Not only the in-service inspections benefit from a reliable NDE system, but also the role of such systems is becoming vital during design stages. Nowadays, it is possible to use life estimation models developed based on fracture mechanics to estimate the maximum allowable defect size located in critical components for an anticipated life span. Moreover, risk assessment models take the estimated maximum allowable flaw size and the flaw size that can be reliably detected as inputs to plan periodical inspections. Bearing these in mind, it is evident that defect detection is not the only concern of NDE studies anymore. Hence, the recent studies try to focus on defect sizing and the reliability of the sizing as well. Eddy current testing (ECT) is a well-established electromagnetic method of NDE. For several years, it has been employed to inspect the fuselage, wings, engine and wheel components of aircraft for tiny surface flaws. For these purposes, ECT split-D reflection differential surface probes are one of the best nominees since they provide a high signal-tonoise ratio. The complexity of electromagnetic theory governing ECT along with the very delicate geometrical details of such probes have slowed down the advances in analytical/semi-analytical modelling of the interaction between such probes and surface defects. Therefore, development of models that could provide quantitative measures of inspected defects has been accompanied with difficulties. However, with the growth of computer science and resources in recent decades, modelling of these kinds of problems using numerical methods such as finite element modelling has become faster than before. Furthermore, the soft computing techniques have become sophisticated enough to effectively imitate the pattern recognition and decision making abilities of human brain. These offerings, brought by the new technology era, can be integrated in order to develop new methodologies for interpretation of ECT signals and sizing of detected defects. The present research aims at developing an eddy current inversion algorithm for characterization of surface defects. The research work consists of two main parts. The first part aims to develop a FEM tool to simulate scanning of surface notches with ECT probe. In this context, the capability of Comsol Multiphysics in modelling the interaction between a split-D probe and various surface notches was investigated and the model was validated by experimental results. Subsequently, an adaptive neuro-fuzzy inference system was designed and trained, using model-generated signals of various notches, for notch length estimation. Furthermore, the sizing system was tested and its performance was assessed. As a second part to this research, reliability of the manual and automated ECT system was investigated through comparing probability of detection curves obtained by testing defective ferromagnetic steel components. The study involved a split-D surface probe, different inspectors and variation of influential test parameters. The sizing concept proposed in this research can be used for any type of ECT probe, material and defect.

Date	20 Aug 2018
Original language	American English
Awarding Institution	École de technologie supérieure
Supervisor	Martin Viens (Supervisor) & Wen Fang Xie (Co-supervisor)