Characterizing water-bearing structures ahead of a tunnel using DC resistivity with combined homogeneous- and heterogeneous-source current injection devices

Water-bearing structures ahead of the tunnel face are major hazard sources that threaten construction safety. Owing to its high sensitivity to water-bearing targets, the direct current (DC) resistivity method has been widely used for advanced geological prediction during tunnel construction. However, the confined tunnel space and complex construction environment limit the amount of effective information acquired by conventional resistivity methods, thereby constraining inversion imaging performance. In particular, traditional methods mainly rely on homogeneous-source current injection, while restricted electrode layouts further reduce the effective detection distance. To address these limitations, this study proposes a tunnel DC resistivity detection method based on combined homogeneous- and heterogeneous-source current injection observations. Numerical simulations were performed using a three-dimensional finite element model with a current intensity of 1 A and 27 face electrodes arranged in a 3 × 9 grid, with a horizontal spacing of 1 m and a vertical spacing of 2 m. A total of 15 numerical models were constructed by combining three anomaly sizes (10 × 10 × 4 m, 10 × 10 × 8 m, and 10 × 10 × 12 m) with five axial distances (10–50 m) ahead of the tunnel face; the 40–50 m cases were further used to assess the potential for extending prediction beyond the conventional reliable detection range of approximately 30 m. The results show that the apparent resistivity curves obtained from the two source modes exhibit distinct intersection points and that the intersection positions vary systematically with the axial distance of the water-bearing structure. Field validation in an operating tunnel, with face electrodes arranged in a 5 × 2 array at 1 m spacing, further confirmed the existence of this intersection feature. The distance information extracted from the intersection points (e.g., x = 11.6 m when y = 1.325) was then incorporated as prior information into reference-model-constrained three-dimensional inversion, thereby improving the stability and effectiveness of anomaly imaging. In addition, the analysis of long-distance anomalies indicates that the intersection information remains responsive at greater detection distances, providing a basis for future studies aimed at extending the effective detection range.