Dynamic Optimization Framework for Real-Time Nondestructive Inspection and Quality Evaluation of Embankments

This study aims to establish a dynamic framework that integrates nondestructive sensing, real-time prediction, and route optimization for autonomous embankment quality evaluation. In automated earthwork, assessing the internal condition of compacted embankments without destructive sampling is essential for ensuring construction reliability. Following the nondestructive testing method described, spatial distributions of dynamic resistance are measured in real time during post-construction inspection. These data are used to identify the permeability and elastic wave velocity fields through inverse modeling, which are subsequently transformed into spatial distributions of flow rate and seismic response using a hybrid learning approach that combines the Neural Chaos framework (Bahador Bahmani et al., 2025). The proposed framework dynamically links real-time sensing with predictive modeling by continuously updating the inferred physical fields and their associated uncertainties. Building on information-theoretic operator analysis, the spatial distribution of information content is quantified at each update step. This enables the adaptive optimization of both observation points and inspection routes, guiding measurements toward regions that maximally reduce predictive uncertainty. The framework thus forms a closed-loop system that couples dynamic observation and learning-based inference, achieving interpretable and self-adaptive quality assessment for automated embankment construction.