Deep Learning Algorithms for the Monitoring and Control of Drilling and Grouting Works-Paper

Paper

Pdf Summary

The study explores using deep learning to improve monitoring and control of drilling and grouting, operations critical to dam safety and other infrastructure projects. It draws on large operational datasets from Mosul Dam in Iraq and Rogun Dam in Tajikistan, totaling roughly 63 million recorded data points. These data were collected through T-Grout, an automated system that records grouting conditions and controls equipment.<br /><br />The researchers trained a Long Short-Term Memory (LSTM) recurrent neural network to predict effective grout pressure from recent time-series measurements. After cleaning and scaling the data, they used sequences of 10 time steps as model inputs and reserved 20% of the data for testing. The model’s test performance was promising: one reported dataset achieved a mean squared error of 0.00183, and the LSTM outperformed the simpler linear regression and ARIMA models evaluated.<br /><br />The results suggest that LSTMs can capture complex temporal patterns in grouting operations and could support more precise, potentially automated regulation of pressure and flow. Integrating predictive models into T-Grout may improve process efficiency, quality control, and safety, including in geologically challenging settings.<br /><br />The authors describe the findings as preliminary. Future work could test the approach in other engineering applications, incorporate drilling information collected before grouting, and connect predictions with real-time geotechnical monitoring so that operations can respond dynamically to ground conditions.

Keywords

deep learning

LSTM

grouting

drilling control

dam safety

time-series prediction

Mosul Dam

Rogun Dam

T-Grout

pressure forecasting