High-Sensitivity Identification of Micro-Voids at Thick Steel Shell-Concrete Interfaces Using Elastic Wave Analysis...

Area of Science:

• Structural Engineering
• Materials Science
• Non-Destructive Testing

Background:

• Steel-concrete composite structures are prone to interfacial void defects, threatening structural integrity.
• Conventional methods struggle to detect early-stage micro-voids (exceeding 2 mm) due to signal ambiguity.
• Missed detections of these voids pose a significant risk to structural load-bearing capacity.

Purpose of the Study:

• To develop a high-sensitivity method for early warning of micro-voids in steel-concrete interfaces.
• To integrate impact elastic wave response features with an advanced machine learning classifier.
• To address the limitations of conventional detection criteria for small interfacial voids.

Main Methods:

• Analysis of temporal and spectral evolution of impact elastic wave responses from full-scale model experiments.
• Construction of a comprehensive feature system including time-domain descriptors, spectral peaks, and sub-band energy.
• Development of a Feature-attention Multi-Layer Perceptron (MLP) classifier with adaptive feature weighting.

Main Results:

• The proposed lightweight classifier achieved a Void recall of 0.978 and a weighted F1-score of 0.780 under the original split.
• Stratified five-fold internal validation showed a balanced accuracy of 0.682 ± 0.021 and a Void recall of 0.845 ± 0.035.
• The classifier demonstrated rapid execution with an average CPU inference time of approximately 1.24 ms per sample.

Conclusions:

• The developed framework shows preliminary potential for engineering-oriented micro-void screening in steel-concrete structures.
• The integration of elastic wave features and attention-based MLP enhances sensitivity for early-stage defect detection.
• The lightweight design enables efficient and rapid defect identification in practical engineering applications.