Classification of Corrosion Severity in SPCC Steels Using Eddy Current Testing and Supervised Machine Learning...

Area of Science:

• Materials Science
• Non-destructive Testing
• Machine Learning Applications

Background:

• Steel Plate Cold-Rolled Commercial (SPCC) steel exhibits durability but is susceptible to corrosion in harsh environments.
• Accurate assessment of corrosion levels is crucial for maintaining material integrity and predicting service life.
• Existing methods may lack the precision required for nuanced corrosion state classification.

Purpose of the Study:

• To develop and evaluate a novel method for assessing SPCC steel corrosion using eddy current testing (ECT).
• To classify corrosion into two distinct states: less corroded (state-1) and highly corroded (state-2).
• To compare the efficacy of generative (GMM) and discriminative (logistic regression) machine learning models for this classification task.

Main Methods:

• Implementation of eddy current testing (ECT) to gather data from SPCC steel samples.
• Feature extraction based on the peaks of perturbed magnetic fields at two distinct frequencies.
• Application of Gaussian Mixture Model (GMM) as a generative classifier and logistic regression as a discriminative classifier.
• Performance evaluation using metrics including absolute error, accuracy, precision, recall, and F1 score.

Main Results:

• The Gaussian Mixture Model (GMM) demonstrated superior performance in categorizing highly corroded states (state-2).
• The logistic regression model proved effective for estimating less corroded states (state-1).
• Both machine learning approaches, when combined with ECT, achieved high classification accuracy for SPCC steel corrosion levels.

Conclusions:

• Eddy current testing coupled with machine learning offers a viable and accurate method for evaluating SPCC steel corrosion.
• The choice between GMM and logistic regression depends on the specific corrosion state being assessed, offering flexibility in application.
• This approach enhances the non-destructive evaluation capabilities for materials exposed to corrosive conditions.