AGO - Automated Grouting Operations-Paper

Paper

Pdf Summary

The paper presents an automated system for improving grouting operations by combining the established Grouting Intensity Number (GIN) approach with machine learning. GIN limits grouting through maximum pressure, target grout volume, and an intensity value, but selecting suitable project-specific parameters is difficult because geological conditions are variable and often uncertain. Reliance on practitioner judgment can lead to overly cautious, inefficient work or excessive pressure that damages the ground.<br /><br />The proposed system analyzes pressure and flow-rate data in real time. It identifies standard and atypical grouting behavior by examining changes in both variables and recognizing characteristic patterns associated with fracture widening or creation. Experts label historical data to train a supervised-learning model that distinguishes ordinary grouting from potentially harmful jacking or fracturing. Reported pattern-recognition accuracy is 88–92%.<br /><br />Early detection enables operators to respond to adverse behavior and helps determine the maximum safe injection pressure under actual site conditions, including the influence of rock-mass saturation. This supports the iterative, site-specific adjustment of the GIN curve rather than relying solely on assumptions, generalized rules of thumb, or rigid parameters copied from other projects. The authors argue that the approach can improve efficiency by reducing unnecessary material use while lowering the risk of ground damage.

Keywords

grouting

GIN approach

machine learning

pressure monitoring

flow-rate analysis

pattern recognition

fracture detection

jacking

rock mass saturation

supervised learning