Data-Driven Assessment of Concrete Slab Integrity via Impact-Echo Signals and Neural Networks

TL;DR

This paper tackles the unreliable detection of subsurface concrete defects by automating Impact Echo (IE) data interpretation through a data-driven pipeline. It converts IE time-series to dominant peak frequencies via FFT, maps them spatially, uses $k$-means clustering guided by Ground Truth Masks (GTM), and classifies four CCDs with a stacked LSTM using sequences of 20 peak-frequency values. Key contributions include a complete FFT-to-LSTM workflow, GTM-aligned defect labeling, and robust field validation showing 72.6% accuracy and IoU around 0.704 across eight laboratory slabs and in-service decks. The framework enhances objectivity, scalability, and repeatability of NDE for network-scale bridge health monitoring and supports proactive maintenance decisions. Practically, it enables automated, region-aware defect localization and multi-class classification, robust to coupling, noise, and environmental variability, with potential integration into asset-management platforms for near-real-time assessment.

Abstract

Subsurface defects such as delamination, voids, and honeycombing critically affect the durability of concrete bridge decks but are difficult to detect reliably using visual inspection or manual sounding. This paper presents a machine learning based Impact Echo (IE) framework that automates both defect localization and multi-class classification of common concrete defects. Raw IE signals from Federal Highway Administration (FHWA) laboratory slabs and in-service bridge decks are transformed via Fast Fourier Transform (FFT) into dominant peak-frequency features and interpolated into spatial maps for defect zone visualization. Unsupervised k-means clustering highlights low-frequency, defect-prone regions, while Ground Truth Masks (GTMs) derived from seeded lab defects are used to validate spatial accuracy and generate high-confidence training labels. From these validated regions, spatially ordered peak-frequency sequences are constructed and fed into a stacked Long Short-Term Memory (LSTM) network that classifies four defect types shallow delamination, deep delamination, voids, and honeycombing with 73% overall accuracy. Field validation on the bridge deck demonstrates that models trained on laboratory data generalize under realistic coupling, noise, and environmental variability. The proposed framework enhances the objectivity, scalability, and repeatability of Non-Destructive Evaluation (NDE), supporting intelligent, data-driven bridge health monitoring at a network scale.

Figures (13)

• Figure 1: Proposed methodology workflow integrating FFT, interpolation, clustering, and LSTM-based classification.
• Figure 2: Binary GTM illustrating the known CCDs region embedded within the $120 \times 40$-inch concrete slab.
• Figure 3: Binary GTM illustrating the segmented defect zones across the concrete slab.
• Figure 4: Peak frequency (kHz) distribution from IE signals across laboratory slab 1.
• Figure 5: Peak frequency (kHz) distribution from IE signals across laboratory slab 2.