Answers are grounded in this paper and cite their sources.
Demonstrates classifying carburization stages in high-pressure steel using ultrasonic features, indicating potential for improved structural integrity.
Carburization is a critical degradation mechanism in HP steel alloys used in pyrolysis furnaces, affecting structural integrity and operational reliability. This study evaluates the feasibility of combining ultrasonic A-scan signal processing and Fourier Transform–based spectral descriptors with machine learning to classify four carburization stages in HP steel tube specimens. A total of 160 A-scan waveforms were acquired under controlled laboratory conditions, each containing 2500 sampled points. Frequency-domain features derived from the Discrete Fourier Transform were used as inputs to decision-tree and k-nearest neighbor classifiers. Model performance was assessed using confusion matrices, accuracy, precision, recall, F1-score, and ROC-AUC in a multiclass framework. Ensemble Bagged Trees achieved the highest within-dataset classification accuracy (>99%) under the adopted cross-validation framework, whereas KNN showed lower classification performance despite higher inference speed. The results indicate strong separability among carburization stages under the evaluated acquisition conditions. Given that multiple acquisitions originated from the same tube specimens, the findings should be interpreted as a feasibility-level assessment. Further validation using independent specimens and expanded datasets is required to assess generalization under industrial conditions.
No takes yet. Share an insight, caveat, or question.
Silva et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: