We analysed over 15 million 30-second drilling records from the Equinor Volve open dataset to answer a question that matters to every drilling engineer: what parameters separate fast drilling from slow?
In 2018, Equinor released the complete production history of the Volve field in the Norwegian North Sea. 5,000 documents, 40,000 files, 24GB of operational data — all public domain.
Real-time 30-second sensor logs: weight on bit, rotary speed, standpipe pressure, flow rate, torque, hookload, mud weight and rate of penetration.
EDM XML geology records mapped 13 distinct formations from Utsira (shallow) to Skagerrak (deep reservoir) for each well in the campaign.
Two complementary approaches: a within-well XGBoost efficiency model to detect underperformance, and quartile benchmarking to find optimal parameter windows per formation.
An XGBoost model trained on the first 80% of each well's timeline predicts what ROP should have been given the drilling parameters. The efficiency score — actual ÷ predicted — reveals where performance deviated from expectations.
Sustained underperformance. Likely causes: bit wear, hard unexpectedly hard formation, or suboptimal parameter selection.
Well-calibrated drilling. Actual ROP matches what the model expects from the applied parameters and conditions.
Outperforming the model. Favourable formation conditions, good parameter selection, or both working in the driller's favour.
Median efficiency score for the held-out 20% of each well's timeline. Only active on-bottom rows (WOB > 2 kN, RPM > 20, ROP > 1 m/hr) are included.
F-10 drilled chalk at 4,500–5,300m — the deepest section in the campaign — where tight formation limited ROP regardless of parameters. F-15S is a directional sidetrack: high RPM and light WOB deliberately sacrificed ROP to maintain trajectory control.
F-14's test period hit formations where the actual ROP consistently exceeded model predictions. The well drilled its test section ~40% faster than the fleet-average model expected given those drilling parameters.
For each formation we split all on-bottom drilling records into ROP quartiles. Comparing median parameters in Q1 (slowest) versus Q4 (fastest) reveals which levers move the needle — and which don't.
Median ROP in the bottom and top quartile for each formation. The gap tells you how much headroom exists.
Push harder, rotate slower. High WOB crushes the soft sandstone; excessive RPM wastes energy without adding penetration.
| Parameter | Q1 (slow) | Q4 (fast) |
|---|---|---|
| WOB (kN) | 7.3 | 15.5 ↑ |
| RPM | 179 | 50 ↓ |
| Torque (kNm) | 15.5 | 15.2 |
| Flow in (L/min) | 1,963 | 1,963 |
Spin faster, push lighter. Ty responds to high rotation rates; heavy WOB in this section likely caused weight transfer issues or bit balling.
| Parameter | Q1 (slow) | Q4 (fast) |
|---|---|---|
| WOB (kN) | 7.6 | 5.1 ↓ |
| RPM | 30 | 160 ↑ |
| Torque (kNm) | 13.9 | 19.7 ↑ |
| Flow in (L/min) | 3,518 | 3,516 |
In Aasgard, Draupne, Heather and Hugin formations the fastest-drilling rows used less mud circulation than the slowest. High ROP in these sections reflects natural fractures and porosity, not aggressive parameters — the rock drills itself when conditions are right.
For each (well, formation) pair we compute the ratio of that well's median ROP to the field-wide median. Values above 1.0 mean this well drilled faster than the campaign average in that formation.
Colour intensity shows deviation from field average (1.0 = field median). Only wells with ≥ 50 on-bottom rows in that formation are shown.
Above field median in every formation with adequate data, reaching 2.3× in Hugin and 2.2× in Skagerrak. Strong reservoir placement throughout the campaign.
Strong in shallow sections but catastrophically slow in Heather (0.09× field median). Likely wellbore instability or severe bit wear in the deep formation transition.
Consistently below field median across formations. Not a parameter problem — F-10 encountered tighter, harder rock throughout its trajectory, confirmed by its 4,500–5,300m chalk section.