Using Sonic MWD to Predict CPT Values-Paper

Paper

Pdf Summary

The paper investigates whether sonic measurement-while-drilling (MWD) data can be used to estimate cone penetration test (CPT) measurements, particularly where hard layers prevent CPT equipment from penetrating. If reliable, the approach could extend subsurface profiles and make construction-phase drilling data useful for updating soil models.<br /><br />The authors first tested whether sonic drilling signals reflect material strength. In controlled drilling through concrete and granite, an accelerometer recorded vibration at high frequency. Signal distortion, measured as total harmonic distortion plus noise (THDN), showed a positive correlation with uniaxial compressive strength (UCS), with a Pearson coefficient of 0.77. This suggests MWD contains information about ground conditions, though THDN alone is not sufficient for accurate UCS prediction.<br /><br />A random forest classifier applied to MWD measurements from one borehole classified six soil or rock categories with 92% accuracy. A random forest regressor was then used to predict CPT cone resistance (qc) and sleeve friction (fs) from MWD parameters. The reported R² scores were 0.96 for qc and 0.88 for fs, with sonic frequency the most influential feature in the tested model.<br /><br />The findings are promising but preliminary. The models were trained and tested on data from the same borehole, so performance across different locations, rigs, operators, and drilling practices remains unknown. Results may also be affected by depth alignment between CPT and MWD data and by unusual measurements at the ends of drilling runs. The authors recommend collecting more diverse borehole data to test generalizability and improve robustness.

Keywords

sonic MWD

cone penetration test

CPT estimation

measurement-while-drilling

soil classification

random forest

cone resistance

sleeve friction

uniaxial compressive strength

subsurface profiling