High frequency eddy current testing has been reported that this method is applicable to detecting the electrical properties change in non-electrical conductive materials such as glass fiber reinforced plastics (GFRPs) [1-2]. This method induces displacement current into the specimen and detects the change of the electromagnetic field from this current. In this study, this method is called as electromagnetic induction testing because this method does not detect the change of the electromagnetic filed from eddy current.

Electromagnetic induction testing can conduct measurements at high speed and non-contact. Furthermore, in the previous research from our group, we proposed the new method to use Driver Field Lens (DFL) with electromagnetic induction testing to measure the crack angle for GFRP. Although DFL has the advantage of detecting the crack angle, it can reduce the sensitivity of detecting the crack existence.

In order to overcome this disadvantage of DFL, autoencoder is combined with electromagnetic induction testing. Autoencoder is unsupervised learning in artificial neural network and it can extract features of input data. Autoencoder can decode by using the former extracted features and it can detect abnormal signals comparing the decoded data. We construct the autoencoder to judge the defect existence in GFRP and possibility to use of it combined with electromagnetic induction testing for non-destructive testing is discussed.

[1] Koichi Mizukami, et al., "Desigin of eddy current-based dielectric constant meter for defect detection in glass fiber reinforsed plastics " NDT&E International, Vol. 74 (2015), pp. 24–32.

[2] S. Gäbler, et al., "Measuring and Imaging Permittivity of Insulators Using High-Frequency Eddy-Current Devices”, IEEE Transaction on Instrumentation and Measurement, Vol. 64, No. 8 (2015), pp. 2227-2238.