Support vector machine and optimised feature extraction in integrated eddy current instrument

Eddy current, especially pulsed eddy current (PEC), is known as an effective tool to detect defects in aircraft structures. Current PEC defect classification methods require highly trained personnel and the results are usually influenced by human subjectivity. Therefore, automated defect classification is desirable in a PEC instrument. In this work, five eddy current based methods are integrated into an instrument using a universal model and modular structure. Then, a Support Vector Machine (SVM) is used to build the classifier model and predict the type of defect. Principal component analysis (PCA) and independent component analysis (ICA) are investigated for feature extraction and compared for classification results using SVM. Two-layer Al–Mn alloy specimens with four kinds of defects are used for classification. The experimental results show that the proposed methods have great potential for in-situ defect inspection of multi-layer aircraft structures.

► Integrating pulsed eddy current (PEC) and other EC methods in an instrument.
► Investigating support vector machine based PEC defect automated classification.
► PCA and ICA are compared in the features extraction for SVM-based method.