Rapid Identification of Material Defects Based on Pulsed Multifrequency Eddy Current Testing and the k-Nearest...

Area of Science:

• Materials Science
• Non-Destructive Testing
• Machine Learning

Background:

• Eddy Current Testing (ECT) is a common non-destructive method.
• Accurate defect parameter estimation is crucial for material integrity.
• Supervised learning offers potential for enhancing ECT analysis.

Purpose of the Study:

• To develop and evaluate a method for estimating defect parameters in conductive materials.
• To integrate Pulsed Multifrequency Excitation and Spectrogram Eddy Current Testing (PMFES-ECT) with machine learning.
• To assess the classification accuracy of the proposed approach.

Main Methods:

• A three-dimensional finite element method (FEM) model was created for sensor-specimen simulation.
• Simulation outcomes were used as training data for the k-Nearest Neighbors (k-NN) algorithm.
• The k-NN algorithm was applied to measurement data for defect parameter estimation.

Main Results:

• The study successfully employed PMFES-ECT and k-NN for defect parameter estimation.
• FEM simulations provided effective training data for the machine learning model.
• Classification accuracy was evaluated for various predictor combinations.

Conclusions:

• The combined PMFES-ECT and k-NN approach is effective for defect characterization.
• This method offers a promising avenue for automated defect analysis in conductive materials.
• Further research can explore different machine learning algorithms and feature sets.