Drilling dysfunctions rarely have unique surface signatures. Probabilistic reasoning keeps several hypotheses alive, updates them as context and evidence arrive, and exposes diagnostic separation instead of converting one noisy KPI into a verdict. The probabilities describe a model's evidence, not direct observation of the downhole mechanism.
Overlapping signatures require probabilities
Torque variability may be generally:
- Healthy drilling: lower.
- Stick-slip: higher.
If the two populations never overlap, a threshold works perfectly. Real drilling is rarely that clean. Healthy drilling can occasionally produce large torque variation, stick-slip can begin gradually, and formation changes can increase variability without severe torsional dysfunction. The resulting feature distributions overlap:
A measured value in that overlap does not belong unquestionably to either state; the evidence may simply support one more strongly than the other.
Overlapping signatures make binary thresholds a lossy representation of evidence.
A model output of stick-slip probability = 0.65 does not mean:
The bit is physically experiencing 65% stick-slip.
The value represents the model's assessment of how the available evidence supports the hypothesis under its assumptions. Conceptually, Bayesian reasoning can be written:
$$P(D|E) = \frac{P(E|D)P(D)} {P(E)}$$
where:
- \(D\) = a possible dysfunction,
- \(E\) = observed evidence.
In practical language:
Start with what was believed before the latest evidence, then update that belief based on how consistent the new evidence is with each possible condition.
That update is useful precisely because uncertainty is preserved. If two dysfunctions produce similar signatures but one is historically common in the current configuration, the common condition may reasonably have a higher prior probability. Strong evidence for the rarer condition should still move the posterior. This balances historical expectation with current-well evidence, but also exposes a limitation: poorly chosen priors can bias the result.
Probability is not objective simply because it is expressed numerically.
Combine different physical evidence
Torque illustrates three distinct features:
| Feature | Example | Information carried |
|---|---|---|
| Location | Current torque = 18,000 ft-lbf | Whether the value is low, normal, or high |
| Trend | 15,000 → 16,000 → 17,000 → 18,000 ft-lbf | Whether the signal is increasing or decreasing |
| Variability | Rapid fluctuation between 12,000 and 24,000 ft-lbf | Whether the response is stable or erratic |
SPE-186166 uses this location-versus-movement distinction explicitly in its public methodology. Now take a model that evaluates:
- ROP,
- MSE,
- DOC.
A reduction in ROP will:
- directly reduce ROP,
- tend to increase MSE,
- reduce DOC.
The model now sees three worsening features. But all three contain the same underlying measurement. That is not equivalent to three independent observations. Likewise:
- torque,
- a torque-variability metric,
- a torque-derived stick-slip index
share information. This does not mean derived features are useless. It means their dependence should be understood. A model can become overconfident if correlated evidence is implicitly treated as independent confirmation. Consider suspected torsional dysfunction. More useful corroboration may come from combining evidence that reflects different aspects of the system:
- torque dynamics,
- ROP response,
- MSE relative to formation,
- downhole vibration if available.
That is stronger than four different mathematical transformations of the same torque channel. The best additional measurement is often the one that reduces ambiguity most.
Let missing evidence broaden the result
A dysfunction model normally uses five relevant features, but one disappears. It should assume neither normal nor abnormal behavior; the evidence should become weaker. Explicit probabilistic reasoning lets missing evidence contribute less information instead of forcing the model into an artificial state. For example, a hypothetical rotary-drilling model produces:
- Stick-slip: 0.48
- Whirl: 0.31
- Bit bounce: 0.08
- Other: 0.07
- No dysfunction: 0.06
These values are purely illustrative. Should the user see:
Stick-slip detected.
Maybe not. A better interpretation may be:
Evidence currently favors torsional dysfunction, but whirl remains a meaningful competing explanation.
That communicates much more engineering information. The probability distribution itself can contain useful ambiguity.
Inspect separation and calibration
| Case | Stick-slip | Whirl | Other | No dysfunction |
|---|---|---|---|---|
| A | 0.90 | 0.04 | 0.03 | 0.03 |
| B | 0.36 | 0.34 | 0.20 | 0.10 |
Stick-slip ranks first in both cases, but only Case A is well separated from alternatives. An interface should not turn the largest probability into a confirmed root cause without additional logic.
The highest-ranked hypothesis is not necessarily well separated from alternatives. Values are illustrative only.
If a model assigns 70% probability to 100 comparable historical cases, a well-calibrated model should see the target outcome in roughly 70 of them, assuming reliable ground truth and a properly defined population. If only 30 correspond, the probability is misleading. A confidence-looking number is useful only when historical outcomes support its interpretation, which is difficult when drilling ground truth is itself imperfect.
Preserve diagnostic scope
Torque behavior may strongly indicate torsional oscillation while surface data cannot locate whether the dominant behavior occurs at the bit, BHA, or elsewhere in the string. SPE-205844 notes a case where surface evidence indicated stick-slip somewhere downhole but could not establish whether the bit or BHA itself was responsible without downhole data. That distinction matters.
- Torsional dysfunction is occurring can be a defensible conclusion.
- The bit is sticking every revolution: may not be.
Real drilling does not always respect our classifier categories. A bit can experience:
- torsional dysfunction,
- coupled lateral behavior,
- changing formation
at the same time. Similarly, a poor hydraulic condition can contribute to:
- bit balling,
- poor ROP,
- elevated MSE.
So a model architecture that forces every observation into exactly one mutually exclusive state may simplify reality too aggressively. Sometimes the correct result is multiple mechanisms may be active. Probabilistic systems make that possibility easier to represent, provided the model structure allows it.
A qualified surveillance workflow
A defensible workflow is:
| Stage | Requirement |
|---|---|
| Identify state | Separate rotary, slide, reaming, and other activity. |
| Validate data | Detect missing, faulty, and stale channels. |
| Add engineering context | Calculate MSE, DOC, formation-relative baselines, and relevant residuals. |
| Extract evidence | Represent current level, trend, variability, and persistence. |
| Maintain hypotheses | Keep plausible dysfunction, formation, hydraulic, and normal states in play. |
| Update probabilities | Let new evidence strengthen, weaken, or leave support unchanged. |
| Represent missing evidence | Do not substitute normal for unknown. |
| Check separation | Determine whether one explanation is materially better supported than alternatives. |
| Explain and act | Show supporting evidence and recommend action only at appropriate confidence. |
Real DrillingMetrics interval comparing the deteriorating pre-trip evidence pattern with the shorter BHA #3 recovery interval. Depth increases downward.
Leading to the bit trip near 9,150 ft, the synchronized traces show ROP declining, MSE remaining elevated and spiky, rotary stick-slip belief increasing, and rotary drilling efficiency falling. On the new BHA #3 run, rotary stick-slip belief remains low, drilling efficiency returns closer to 1, ROP and WOB recover, and MSE is lower. The value of the figure is not the belief track by itself. It is the reader's ability to see which measured and derived evidence changed with that belief.
Related technical resources
- When “Unable to Determine” Is the Correct Output in Drilling Analytics
- Distinguishing Sensor Faults from Real Drilling Process Changes
- Mechanical Specific Energy: Powerful Metric, Incomplete Diagnosis
References
-
Ambrus, A., Ashok, P., Chintapalli, A., Ramos, D., Behounek, M., Thetford, T. S., and Nelson, B. A Novel Probabilistic Rig Based Drilling Optimization Index to Improve Drilling Performance. SPE-186166-MS, SPE Offshore Europe Conference & Exhibition, Aberdeen, United Kingdom, 2017.
-
Witt-Doerring, Y., Pastusek, P. P., Ashok, P., and van Oort, E. Quantifying PDC Bit Wear in Real-Time and Establishing an Effective Bit Pull Criterion Using Surface Sensors. SPE-205844-MS, SPE Annual Technical Conference and Exhibition, 2021.