Defect detection and classification using a SQUID based multiple frequency eddy current NDE system

The probability of detection (POD) of hidden fatigue defects in riveted multilayer joints, e.g. aircraft fuselage, can be improved by using sophisticated eddy-current systems which provide more information than conventional NDE equipment. In order to collect this information, sensor arrays or multi-frequency excitation schemes can be used. We have performed simulations and measurements with an eddy current NDE system based on a SQUID magnetometer. To distinguish between signals caused by material defects and those caused by structures in the sample, such as bolts or rivets, a high signal-to-noise ratio is required. Our system provides a large analog dynamic range of more than 140 dB/√Hz in unshielded environment, a digital dynamics of the ADC of more than 25 bit (>150 dB) and multiple frequency excitation. A large number of stacked aluminum samples resembling aircraft fuselage were measured, containing titanium rivets and hidden defects in different depths in order to obtain sufficient statistical information for classification of the defect geometry. We report on flaw reconstruction using adapted feature extraction and neural network techniques