Research on karst detection based on full waveform inversion imaging technology

The geological development of karst areas is highly complex, and accurately detecting underground karst caves presents a significant technical challenge to ensure the safety of construction projects. Karst features, including underground caves and subterranean rivers, are widely distributed across China. Characterized by high concealment and spatial randomness, these geological structures are highly prone to inducing disasters such as ground collapse and water or mud inrush. While traditional geophysical methods are effective at certain depths, they tend to rely heavily on wave velocity parameters but underutilize waveform information, such as amplitude and phase. These results in significant errors in identifying cave boundaries and a high rate of missed detections.To enhance detection accuracy, this study proposes a combined approach that integrates the surface wave method with the cross-hole elastic wave method and introduces full waveform inversion imaging technology to create a comprehensive detection system. The surface wave method is cost-effective and suitable for rapid surveys, though its interpretation is significantly affected by surface heterogeneity. The cross-hole elastic wave method, while effective in minimizing shallow interference by utilizing hole-based excitation and reception, offers superior resolution at greater depths but incurs higher costs and requires pre-drilling.Full waveform inversion technology utilizes all available waveform information-such as wave velocity, amplitude, and phase-based on numerical inversion of the wave equation. It overcomes the limitations of the traditional horizontal layered medium assumption, significantly improving the accuracy of karst cave boundary identification. A field test conducted in a karst area in Guizhou adopted a joint observation approach, yielding promising results. Full waveform inversion clearly delineated the spatial distribution of karst caves (depth of 6 to 12 m, lateral range from 15 to 25 m), with imaging resolution far surpassing that of first arrival wave analysis and Synchronous Iterative Reconstruction Technology (SIRT). A serial inversion strategy, progressing from low to high frequencies, effectively suppressed noise and ensured stable convergence to the global optimum. The conjugate gradient iterative algorithm was used for inversion, with an attenuation boundary width set at 2 m, a grid size of 0.2 m × 0.2 m, and serial optimization divided into three stages at 20 Hz, 30 Hz, and 40 Hz. The error decreased progressively with increasing iterations, demonstrating the method's numerical stability and convergence.The results further highlight that joint inversion of P-wave and S-wave data offers mutual verification, with a high degree of consistency in identifying the anomaly locations. The boundary of the S-wave inversion result was clearer, in contrast to the first arrival wave method, which only provides a one-dimensional wave velocity curve. SIRT reconstruction, on the other hand, is prone to boundary ambiguities and geometric distortion. In comparison to traditional methods that rely solely on first arrival travel time, full waveform inversion leverages the dynamic characteristics of the wave field, significantly enhancing the ability to identify small anomalous bodies and complex boundaries.Additionally, comparing the inversion results from the surface wave and cross-hole elastic wave methods reveals good consistency in identifying the main forms of the karst caves. However, cross-hole data provided more reliable deep structural details, whereas surface wave inversion exhibited local false anomalies due to shallow inhomogeneities. This underscores the importance of the combined surface and hole-based observation method. The fusion of these methods maximizes the complementary strengths of the surface wave method's efficiency and the cross-hole method's detailed depth resolution, offering a high-reliability, adaptable technical solution for engineering geological surveys in complex karst environments, with substantial engineering application value.