Frequency Optimization for Enhancement of Surface Defect Classification Using the Eddy Current Technique.

Area of Science:

• Materials Science
• Non-Destructive Testing
• Signal Processing

Background:

• Eddy current testing (ECT) is a widely used non-contact method for evaluating material integrity.
• Excitation frequency is a critical parameter influencing ECT performance, particularly for defect characterization.
• Existing research often focuses on detection sensitivity, with less emphasis on frequency optimization for classification accuracy.

Purpose of the Study:

• To investigate the impact of excitation frequency on surface defect classification performance in ECT.
• To identify optimal excitation frequencies for enhanced defect characterization, considering multiple defects simultaneously.
• To propose experiment-based procedures for determining optimal excitation frequencies.

Main Methods:

• Utilized eddy current testing with varying excitation frequencies.
• Analyzed defect signals for sensitivity, feature contrast, and classification accuracy.
• Employed kernel principal component analysis (KPCA) for feature space optimization.
• Applied support vector machine (SVM) for defect classification.

Main Results:

• Detection sensitivity for a group of defects is maximized when excitation frequency is near the optimal frequency for the largest defect.
• KPCA effectively optimized feature margins for SVM classification.
• The proposed method achieved the best classification accuracy by selecting optimal excitation frequencies.

Conclusions:

• Excitation frequency significantly influences the characterization of surface defects in ECT.
• An experiment-based approach for determining optimal excitation frequencies for multiple defects improves classification performance.
• This research provides a framework for enhancing ECT defect characterization beyond simple detection.