AI + Geophones Enable a New Paradigm for Slope Monitoring: A Look at a Sichuan Project

Slope safety is vital to transportation networks and the safety of people and property. In Sichuan, where geological hazards are common, Genvict has deployed its independently developed multi-source data fusion model for slope early warning together with passive-source geophone technology for subsurface detection. The solution enables precise and efficient slope safety monitoring across the province, strengthening slope protection through technology.

1.Project Background: A Safety Mission amid Geological Risks

A highway fill embankment in Ziyang, Sichuan, carries the road through this section as a sidehill embankment formed by partial cutting and filling. The section opened to traffic in 2017. Irregular longitudinal cracks subsequently began to appear on the pavement. On the Chongqing-bound carriageway, the cracking gradually extended into transverse cracks spanning Lanes 2, 3, and 4, forming a ring-like pattern with a maximum width of approximately 2 cm. On the Chengdu-bound carriageway, settlement of the high-fill subgrade caused cracks of less than 1.0 cm across Lanes 2 and 3, extending for approximately 70 m. Although the maintenance team repeatedly sealed the cracks by grouting during annual pavement maintenance, rainfall infiltration caused them to reopen during the following rainy season. The three-dimensional progression of the cracking indicates that the embankment is experiencing not only vertical settlement but also horizontal displacement, creating a risk of subgrade sliding and instability.

Genvict deployed geophone monitoring equipment built around passive elastic-wave monitoring technology, together with its independently developed multi-source data fusion model for slope early warning. Integrated with GNSS receivers, rain gauges, and other instruments, the system provides long-term dynamic monitoring and early warning for the deformation zone. It enables accurate slope safety monitoring and evidence-based warnings, providing robust technical support for geological hazard prevention and control in the area.

Figure 1. Current Condition of the Monitored Pavement

2. Core Technology: AI-Based Early Warning through Multi-Source Fusion

Unlike conventional single-source monitoring, Genvict’s multi-source data fusion forecasting model integrates three core datasets: rainfall, GNSS displacement, and crack width. Together, they create an efficient slope early-warning framework.

By standardizing the three data types, the model achieves deep fusion of multidimensional information. It combines the overall deformation trends captured by GNSS with the high-precision local measurements from crack gauges, while identifying how rainfall influences slope deformation. This enables coordinated monitoring of external triggers and slope response.

Through systematic feature extraction and data integration, the model distinguishes long-term slope evolution from short-term fluctuations. It captures small routine deformations and identifies abnormal changes promptly, making warnings more targeted and reliable.

3. Core Technology: Geophones for Passive-Source Detection and Broader Coverage

The geophones use weak surface and subsurface elastic waves generated by highway traffic and ambient vibrations. Because wave propagation velocity varies with the density, moisture content, and mechanical strength of rock and soil, the devices collect natural vibration signals at different times and apply cross-correlation interferometry to extract changes in elastic-wave travel time. Combined analysis of dv/v, natural frequency, reflected-wave response, and other multidimensional data enables accurate assessment of moisture-induced soil softening, changes in the geological medium, overall slope stability, and internal structural changes. The passive, non-invasive monitoring approach is easy to deploy and supports continuous long-term operation, enabling comprehensive and accurate detection of changes in overall slope condition.

Figure 2. Geophone Sensor

4. Field Deployment: Efficient, Convenient, and Adapted to Engineering Requirements

At the equipment layer, a seismic survey line was deployed along the highway on the side closest to the slope. The line consists of 88 nodal seismographs spaced 5 m apart, with a total length of 435 m. Each seismograph is configured with a sampling rate of 1000 Hz to capture high-frequency elastic-wave information in near-surface materials. During monitoring, instrument data is transmitted to the server in real time over a 4G network.

At the AI model layer, a time-series forecasting model was developed for slope safety monitoring and displacement prediction using GNSS three-dimensional displacement data, rain-gauge data, and other sources. Multimodal fusion is used to model slope deformation trends. Data was collected from September 2025 to January 2026 at a sampling frequency of 1 sample per hour. Each monitoring point therefore generated more than 7,000 time-series records, forming a large-scale, multidimensional time-series dataset. Preprocessing included missing-value handling, outlier filtering, time alignment, and normalization.

Figure 3. Geophone Deployment

Figure 4. Software Deployment

5. Operational Results: Accurate AI-Based Early Warning for Stronger Safety Protection

A comparison of measured monitoring-point data and model predictions over nearly 6 months showed strong overall model performance and effective early-warning results.

During periods of stable slope conditions, the model predictions closely matched the measured data and accurately reproduced routine displacement patterns. The model provides reliable medium- and long-term trend forecasting that meets the core requirements of routine engineering monitoring. In sudden-deformation scenarios, it showed a slight lag at the instant of abrupt change but still captured the abnormal trend effectively, providing valuable time for emergency response.

Overall, the model is well suited to highway slope monitoring. Its accuracy and stability meet engineering requirements, providing an efficient and practical solution for slope safety monitoring.

Figure 5. Comparison of GNSS Displacement Measurements and Predictions

6. Operational Results: Slope Tomography for Early Identification of Internal Conditions and Potential Risks

The tomographic imaging results show an effective detection depth of approximately 40 m and a subsurface P-wave velocity range of 530-4500 m/s in the surveyed section. Three stratigraphic horizons are visible in the seismic profile. Along the highway alignment, the velocity distribution shows a degree of heterogeneity, reflecting spatial variation in the subsurface depositional environment and the properties of the rock and soil beneath this embankment section.

The velocity profile within the upper 20 m reveals more pronounced lateral heterogeneity. A continuous low-velocity body is present at one location in the profile. It extends over a relatively large area and maintains a stable thickness. Based on its velocity characteristics, it is interpreted as a loosely structured sedimentary layer with relatively weak mechanical properties. The presence of this weak layer indicates poor integrity and high deformability in the shallow strata beneath this embankment section.

Below the sedimentary layer, subsurface velocities remain lower than in the surrounding area, and the interface forms a downward depression near the center. Such structural features are commonly associated with fault-fracture zones, loosened rock and soil caused by tectonic activity, or shear-weakened zones near an old slip surface. The area is therefore likely to contain a fault-fracture zone of appreciable extent that could affect the stability of the overlying embankment.

Overall, the survey identified a shallow weak sedimentary layer located 105-176 m from the start of the survey line. The underlying interface has a downward-concave shape, showing characteristics typical of a fault-fracture zone or a strongly weathered weak zone. This structure is a potential weak point in the foundation. Rainfall infiltration, traffic loads, and seasonal changes may concentrate deformation and pose a risk to the long-term stability of this embankment section. To reduce the potential hazard, Genvict recommends prioritizing reinforcement of this area in subsequent remediation and conducting the necessary drilling and in-situ testing. These investigations should further determine the thickness, lateral extent, water-bearing characteristics, and mechanical parameters of the fracture zone, providing reliable evidence and quantitative parameters for reinforcement design.

Figure 6. Monitoring Results for Subsurface Elastic-Wave Velocity Change (dv/v) between Two Stations along the Slope-Crest Station Line

Figure 7. Resonant-Frequency Monitoring Results from Stations at the Slope Toe

Figure 8. P-Wave Tomographic Imaging Results (0-40 m Depth)

7. Industry Significance: Advancing a New Paradigm for Slope Monitoring through Technology

Genvict’s dual-technology architecture combines a multi-source data fusion model for slope early warning with nodal passive-source geophones, overcoming the inherent limitations of conventional single-source monitoring. It deeply integrates multidimensional data, including subsurface microtremors, GNSS displacement, crack deformation, and rainfall, to support efficient aggregation, unified assessment, and intelligent analysis of site-wide sensing data. This refined integration and scenario-specific application of monitoring data strengthen the scientific basis, accuracy, and timeliness of slope risk assessment.

By combining the geophones’ distributed subsurface sensing coverage with AI forecasting algorithms, the system detects hidden risk signals such as deep microtremors in rock and soil, movement along rock fractures, and precursors to slope sliding. It creates an end-to-end sensing chain from underlying geological anomalies to surface deformation, shifting slope safety management from reactive post-event response to proactive early warning and enabling earlier detection, assessment, and alerts.

This innovative solution provides a robust technical foundation for comprehensive geological hazard prevention and the long-term safe operation and maintenance of infrastructure such as highways, railways, and mining roads. It supports the slope monitoring industry’s transition toward more intelligent and precise practices and strengthens safety protection for high-risk slopes, mountain subgrades, and projects adjacent to mountainous terrain.