Sandvik Automates Underground Rock Bolt Analysis Using AI And LiDAR

Maintaining geotechnical stability in underground mining and civil tunneling relies fundamentally on the integrity of installed ground support systems. Rock reinforcement elements such as friction bolts, mechanically anchored bolts, and grouted tendons stabilize the rock mass by pinning loose strata and controlling convergence around openings. Geotechnical engineers face significant operational challenges when assessing installed reinforcement: physical inspection under unsupported or recently blasted ground exposes personnel to environmental hazards, while manual measurement of bolt patterns across extensive tunnel networks introduces measurement variance, slow documentation turnaround, and high operational overhead.

Ground Support Quality Control Automation
Sandvik launched xCell Vision, a cloud-based software platform engineered to automate rock bolt identification and verification for quality assurance and quality control (QA/QC) workflows. Operating as an extension of Sandvik's Digital Ground Support Platform, the system converts raw spatial measurements into validated geotechnical installation metrics. By replacing subjective manual auditing with systematic algorithm-driven evaluations, engineering teams can identify out-of-spec installations and prioritize physical inspections strictly where anomalous ground reinforcement patterns occur.

LiDAR Point Cloud Ingestion and Processing Architecture
The core architecture of the software processes 3D point cloud data acquired from multiple underground mobile and fixed surveying configurations. The data ingestion pipeline supports point clouds captured via:

Handheld mobile mapping scanners deployed during routine geology walk-throughs

Tripod-mounted stationary laser scanners used for high-density geotechnical surveys

Machine-mounted scanning systems integrated directly onto production drill rigs, bolting equipment, or utility haulage fleets

Once spatial point clouds are uploaded to the cloud environment, machine learning classification models isolate individual reinforcement fixtures from rough, shotcreted, or fractured rock mass geometries. The model detects discrete bolt types, filters out background surface irregularities, and computes vector orientations across the scanned excavation profile.

Automated Geotechnical Parameter Extraction
Manual documentation of reinforcement quality requires recording discrete geometric variables across thousands of installation points. The platform automates the extraction and quantitative assessment of four primary geometric parameters:

Total bolt count: Verifies total installed reinforcement against design density requirements across predefined excavation intervals.

Bolt spacing: Computes spatial distance between adjacent collar locations to verify adherence to approved geotechnical bolting grids and ring spacing.

Bolt orientation: Measures the 3D insertion angle and vector trajectory relative to the excavation profile and anticipated local joint sets.

Exposed tail length: Quantifies protruding steel length past the collar and faceplate to verify proper anchoring depth and plate contact.

The extracted parameters populate an interactive 3D digital model of the scanned underground infrastructure. Within this spatial viewer, geotechnical engineers can cross-reference physical installations against planned support design standards, visually isolate substandard bolting densities, and export analytical datasets directly into broader mine planning and ground control registers.

Improving Operational Safety and Planning
Eliminating direct manual tape measurements and physical collar counting under active headings decreases the exposure of survey personnel to unsupported ground, vehicle interaction hazards, and airborne particulates. From an engineering workflow perspective, automated spatial processing shifts geotechnical human resources away from repetitive data collection toward high-level risk assessment and reinforcement remediation.

The captured spatial data provides historical baseline verification for long-term excavation stability. When geotechnical departments integrate verified bolt installations into broader digital ground support workflows, structural discrepancies or ground deformation trends can be cross-correlated against initial installation quality.

Additional Context
This section details technical specifications and competitive benchmarking not included in the original product announcement.

Digital quality assurance for rock reinforcement forms an expanding discipline within digital mining operations. In conventional underground workflows, QA/QC audits follow industry guidelines such as those defined by international rock mechanics associations (e.g., ISRM and national mine safety inspectorates), which mandate regular verification of support installation compliance against approved Ground Support Management Plans (GSMP). Historically, compliance audits relied on paper checklists, mechanical pull tests on a tiny percentage of bolts, and manual tape-and-inclinometer surveys.

Automated analysis tools are evolving alongside modern terrestrial and mobile LiDAR technology. Comparable approaches in the geotechnical and tunnel inspection sector include specialized point cloud processing packages from software providers such as Deswik, Maptek, and Bentley Systems, as well as mobile-scanning OEM analytics like those paired with Emesent Hovermap systems. While general mining software suites often require manual point cloud classification or specialized CAD scripting to extract individual bolt vectors, dedicated machine learning models target near-instantaneous automated segmentation of small-scale hardware fixtures directly from noisy underground environments.

Standard terrestrial LiDAR hardware typically achieves range accuracy within 2 to 6 millimeters, whereas SLAM-based (Simultaneous Localization and Mapping) mobile units generally capture point clouds with accuracy between 10 and 30 millimeters depending on trajectory drift and scan speed. Automated bolt identification software must accommodate these resolution variances, recognizing bolt plates and protruding threaded tails that often measure under 100 millimeters in total dimension against irregular shotcrete surfaces. As underground mobile machinery increasingly incorporates real-time LiDAR payloads, automated analytical platforms are shifting from post-survey desktop evaluations toward closed-loop automated verification directly following the bolting cycle.

Edited by Evgeny Churilov, Induportals Media - Adapted by AI.

www.mining.sandvik