Characterisation of surface and sub-surface discontinuities in metals using pulsed eddy current sensors

Due primarily to today's rigorous safety standards the focus of non-destructive testing
(NDT) has shifted from flaw detection to quantitative NIDT, where characterisation of
flaws is the objective. This means information such as the type of flaw and its size is
desired. The Pulsed Eddy Current (PEC) technique has been acknowledged as one of the
potential contenders for providing this additional functionality, due to the potential
richness of the information that it provides. The parameters mainly used to obtain
information about the detected flaws are the signal's peak height and arrival time.
However, it has been recognised that these features are not sufficient for defect
classification. In this research, based on a comprehensive literature survey, the design of
PEC systems and the interpretation of PEC signals, mainly for flaw classification, are
studied.
A PEC system consisting of both hardware and software components has been designed
and constructed to facilitate the research work on PEC signal interpretation. After a
comparative study of several magnetic sensing devices, probes using Hall device magnetic
sensors have also been constructed. Some aspects related to probe design, such as coil
dimensions and the use of ferrite core and shielding have also been studied.
A new interpretation technique that uses the whole part of PEC responses and is able to
produce more features has been proposed. The technique uses Principal Component
Analysis (PCA) and Wavelet Transforms, and attempts to find the best features for
discrimination from extracted time and frequency domain data. The simultaneous use of
both temporal and spectral data is a logically promising extension to the use of time
domain only with the signal-peak-based technique. Experiments show that the new 1 technique is promising as it performs significantly better than the conventional technique using peak value and peak time of PEC signals in the classification of flaws. A hierarchical
structure for defect classification and quantification has been presented.
Experiments in the project have also shown that the signal-peak-based technique cannot be
used for flaw detection and characterisation in steels, both with and without magnetisation.
The new proposed technique has shown to have potential for this purpose when
magnetisation is used. The new technique proposed in the report has been successfully
used for ferromagnetic and non-ferromagnetic materials. It has also been demonstrated that
the new proposed technique performs better in dynamic behaviour tests, which shows its
better potential for on-line dynamic NDT inspection which is required in many industrial
applications. In addition to testing calibrated samples with different discontinuities, a study
case using an aircraft lap joint sample from industry has further supported the statement
regarding the potential of the new technique.