GPR-Based Quality Assessment for Ultra-Thin Grouting Layers in Box Culverts Integrating Rheological Features and Time-Varying Electromagnetic Features

During the construction and maintenance of prefabricated box culverts in subsea tunnels, accurate assessment of the compactness of the ultra-thin bottom grouting layer is essential for ensuring overall structural stability. Ground Penetrating Radar (GPR) technology could allow non-destructive evaluation of grouting quality by capturing dielectric contrasts between media to identify hidden voids. However, real-time assessment of this 50-mm ultra-thin layer faces significant challenges, as traditional detection methods struggle to adapt to the drastic variations in dielectric properties during the rapid setting process. Furthermore, existing numerical simulations are typically based on preset defect sizes and locations, failing to reproduce the random defect features induced by grout rheology. This limitation results in a lack of high-fidelity training data for intelligent monitoring algorithms.

In this study, a physics-driven dynamic grouting scene-generation and assessment framework was proposed to address the scarcity of monitoring data for the quality of box culvert grouting. Based on the time-varying evolution laws of the grouting material from fluid to solid states, full-cycle electromagnetic characteristic parameters were obtained to establish a dynamic mapping mechanism between grouting age and radar response signals. To address the grouting diffusion mechanism in the heterogeneous structural environment of the culvert bottom, a defect scene reconstruction method was developed to consider the grouting process and the coupling between slurry rheology and gravity. This method simulates non-homogeneous and irregular void morphologies under realistic working conditions, overcoming the limitations of traditional regular geometric modeling. A high-fidelity GPR forward simulation framework was constructed to generate standardized datasets covering different setting sequences and interface contact states. Furthermore, a stepped frequency continuous wave (SFCW) simulation framework was developed to standardize data processing across different frequency bands, enabling rapid screening and localization of weak grouting zones through target-detection algorithms.

Results demonstrate that the synthetic data generated by this method effectively reflect the signal evolution patterns across different curing stages, resolving the issue of sample scarcity caused by sparse field data. Compared to static models, synthetic datasets incorporating time-varying features and rheological constraints better capture the authentic signal characteristics of early-stage defects. This indicates that improving the data generation paradigm is crucial for achieving intelligent, real-time monitoring of box culvert grouting quality.