Anti-collision: separation factor, ellipses and travelling cylinder

On a cluster pad the next well starts 6 m from a producing one. Anti-collision is the arithmetic that says how close the new well may pass: where each well might really be, how much room the rule demands, and what the plots show when a plan fails it. This article works through a five-slot pad in the Llanos Basin with DrillIQ's anti-collision engine, from the error ellipses to the travelling cylinder, and ends with what the rig team watches while drilling.

What anti-collision protects against

Drilling into a live well can put reservoir pressure straight into the new hole, cut the offset's casing and cost both wells. On a pad the risk sits in the top hole, where the wells are a few metres apart. Deeper down, wells drilled to neighbouring drainage points can run side by side for a kilometre or more.

No well's position is known exactly. Every survey station carries errors: sensor bias and scale factor, misalignment, BHA sag, and the error in the magnetic reference that turns a compass reading into an azimuth. They add up along the hole. Anti-collision compares two clouds of possible positions and asks whether they are far enough apart, with a margin agreed in advance.

Error ellipses: where each well might be

The ISCWSA error models list every error source of a survey tool (MWD, MWD with in-field referencing, gyros) with its size and how it propagates. Run along a survey, the model gives each station a covariance matrix: an ellipsoid of possible positions, an ellipse when seen from above. DrillIQ draws it at two standard deviations (2σ).

Near surface the ellipse is small and round. Once the well is inclined, the azimuth error dominates and the ellipse stretches across the hole. At TD the example well's 2σ ellipse is 18.8 m by 11.9 m, its long axis on 147°, square to the well's 056° heading. Two wells drilled side by side have their largest uncertainty pointing straight at each other.

The region matters. An MWD tool takes its azimuth from the horizontal part of the earth's field, and near the equator that part is weak. At the pad the World Magnetic Model 2025 gives 29,595 nT at 25.2° dip, so the horizontal field is 29,595 × cos 25.2° = 26,776 nT. A site with 43,252 nT at 34.9° dip has 35,460 nT. Several magnetic terms of the error model grow as the horizontal field weakens: on the same plan the engine's 2σ semi-major axis at TD is 19.5 m in the Llanos field against 17.5 m in the stronger one, 11 % more.

On a pad the MWD also reads the steel of the neighbouring casings, which no error model covers. That is why pad wells run a gyro in the top hole: here a north-seeking gyro to 600 m, then MWD.

The separation factor

DrillIQ uses the ISCWSA rule published in SPE-187073:

C2C

centre-to-centre distance between the closest points of the two wells

R_r, R_o

hole radii of the new well and the offset there

S_m

surface margin, 0.3 m

σ_r, σ_o

each well's 1σ uncertainty along the line joining the two points

σ_pa

project-ahead uncertainty, 0.5 m: the bit is always ahead of the last survey

k

3.5, for offsets where a collision is a safety risk

The top line is the clear space between the hole walls, less a margin; the bottom line is the room the uncertainty needs. At SF 1.0 the two 3.5σ envelopes just touch. The standard stops drilling below 1.0, calls for an engineering review (and a management of change while drilling) below 1.25, and stops scanning an offset above 5. An operator may raise the floor, never lower it. The example pad plans to 1.5, so the drilled well keeps room to wander.

The closest points can be found three ways: the 3D closest approach (the default), the normal plane square to the new well (what the travelling cylinder uses) or the same TVD. On this pad the first two agree at every minimum to two decimals.

The example pad

The pad is set in conditions typical of the Llanos foreland: flat ground about 290 m above sea level, a sandstone reservoir at 2,450 m TVD, and deviated wells fanned out from one row of slots. Four wells are producing. The new well goes in the middle slot, with a target to the north-east.

A hand check at the closest point

Plan A is the obvious first draft. But slot 4's well, 6 m east, is still vertical at that depth, and a line on 055° from slot 3 passes close to it. The engine reports SF 1.08 at 430 m. Here it is by hand, from the terms the engine shows for that point.

1. Clear space. The centres are 3.504 m apart. Both holes are 17½″, radius 0.222 m. Less the margin: 3.504 − 0.445 − 0.3 = 2.760 m.
2. Uncertainty along the line. The new well's 1σ is 0.362 m (gyro), slot 4's 0.392 m (MWD): √(0.362² + 0.392²) = 0.534 m.
3. The room the rule needs. With the project-ahead term, √(0.534² + 0.5²) = 0.731 m, times 3.5 = 2.560 m. Adding the radii and margin gives the minimum allowable separation, 3.30 m.
4. Separation factor. 2.760 / 2.560 = 1.08: below the 1.25 review level, and plan A can't meet the pad's 1.5.
5. At surface. There is no survey error yet, only σ_pa: (6.0 − 2 × 0.330 − 0.3) / (3.5 × 0.5) = 2.88 for two slots 6 m apart in 26″ hole.

The engine gives 1.078 and 2.880. Step 3 holds the lesson for shallow hole: the fixed 0.5 m term is as large as both wells' survey uncertainty at 430 m, so a better survey tool barely helps. With MWD from surface instead of the gyro, plan A's minimum would be 1.057.

Reading the separation-factor plot

The ladder plots the SF against each offset along the new well's depth. The lowest point of each line is that offset's verdict.

While the wells are vertical, the SF to slots 2 and 4 falls slowly, from 2.88 at surface to 2.27 at 300 m, as the ellipses grow. Plan A's dashed line then drops from 2.19 at its 350 m kick-off to 1.08 at 430 m: half its value in less than three stands. The closest approach, 3.44 m, is at 426 m.

The problem is geometry, so the fix is the trajectory. Plan B kicks off 50 m higher, nudges north to 8° and only then turns towards the target, passing north of slot 4, which falls away to the south-east. Its worst top-hole values are 2.25 against slot 4 at 320 m and 2.27 against slot 2 at 300 m, at the kick-off. The price is 5 m more hole and a build-and-turn instead of a plain build.

Slot 5 behaves differently. Its SF peaks at 11.0 at 720 m, then falls along the hold to 1.67 at TD while the wells move apart: 64.8 m at 2,000 m (SF 2.10), 74.5 m at 2,500 m (1.76), 80.0 m at TD. The ellipses grow faster than the gap.

The travelling cylinder

The travelling cylinder looks down the new well. At each depth it plots where each offset's centre crosses the plane square to the hole: distance and bearing. Here the bearing is from grid north. Drawn from the high side instead, each bearing becomes the toolface that would steer towards that offset.

Each offset starts at its slot spacing and stays there while both wells are vertical. On plan A slot 4 swings from east to south and comes in to 3.5 m at 430 m, just outside the dashed SF 1.0 boundary at 3.3 m, then leaves to the south-west. On plan B it moves away to the south-east: 9.1 m at 400 m, 14.8 m at 450 m, 21.7 m at 500 m.

The cylinder shows what the ladder can't: direction. At 430 m on plan A slot 4 sits on a bearing of 156°, so a correction towards the south-east closes the gap and one towards the north-west opens it. When the SF is falling, the directional driller needs that bearing as much as the number.

Why 80 m apart is only SF 1.67

At TD the 2σ ellipses of the new well and slot 5, 18.8 m and 19.5 m semi-major, point at each other and leave about 42 m between them. The rule still gives 1.67, because it works at 3.5σ. Along the line between the wells σ_r is 9.41 m and σ_o 9.75 m, together 13.54 m; the rule needs 47.4 m, and (79.95 − 0.22 − 0.3) / 47.43 = 1.67.

That denominator sets the sensitivity: with the same ellipses, every metre of drift towards slot 5 costs 0.021 of SF. Drift 8.3 m and the well is at the pad's 1.5; drift 20 m and it is at 1.25. A walk of that size in an 8½″ hold is ordinary, and so is a steering correction made for the reservoir (see geosteering with a typewell). The gyro top hole helps a little even here: with MWD from surface, σ_r at TD is 9.72 m and the SF 1.65.

While drilling: monitoring practice

A plan's scan holds only while the well follows the plan. Once drilling starts, the reference becomes the definitive survey plus the plan below it, and every survey changes the answer.

Survey often where the ladder is steep. Plan A's SF halved in 80 m. Near an offset, survey every stand, project the last survey to the bit and rescan: the sensor sits behind the bit.

Trust the gyro near steel. Compare each MWD station's total field and dip with the reference field; a station that fails is interference until shown otherwise.

Treat the offset as live. Before a close approach, agree with production whether the offset is shut in and its annuli watched. Watch for signs of contact: cement or steel at the shakers or on the ditch magnets, a jump in torque or vibration, an unexpected drilling break.

Know the offsets' data. Give an old well surveyed with an unknown tool a pessimistic error model, and check its datum, north reference and coordinate system: a mix-up there moves a well by hundreds of metres.

Act on the levels. Below 1.25, review before drilling on: confirm the surveys, rescan with a projection, agree a correction with the people who own the offset. Below 1.0, stop.

A checklist for the rig team

• Scan every plan against every well within the search radius before approval.
• Check the lowest SF by hand once: clear space over 3.5 times the combined uncertainty.
• In shallow hole, fix a low SF with the trajectory; a better survey tool won't create the space.
• Run a gyro in the top hole of a pad, and QC every MWD station's total field and dip.
• Survey every stand near an offset, project to the bit, and rescan after each survey.
• Read the travelling cylinder for direction before choosing a steering correction.
• Know how many metres of drift take the deep SF to your floor.
• Agree the offset's status before the close approach.

How DrillIQ handles it

DrillIQ's Directional & Surveys module scans any survey or plan against every well within the search radius, path to path, with each survey run modelled by its own ISCWSA tool model and each hole section by its own size. It applies the SPE-187073 rule with an optional tighter floor, and offers the 3D, normal-plane and horizontal scans. The anti-collision page shows the offsets table, the SF ladder, the travelling cylinder from north or high side, and the plan-view ellipses with lease and hard lines, with every SF worked through term by term. A plan is scanned while it is being drawn.

While drilling, surveys accepted from the live feed are rescanned in the background, the cockpit shows the closest offset, and the separation factor goes to the Log Viewer beside the drilling data. Everything above was worked out with the same engine.

Book a demo → See all capabilities

References

1. Sodling, Clark and Allen. The Development and Testing of an Enhanced Anti-Collision Rule. SPE 187073, SPE Drilling & Completion, 2019.
2. Williamson, H.S. Accuracy Prediction for Directional Measurement While Drilling. SPE 67616, SPE Drilling & Completion, 2000.
3. Torkildsen, Håvardstein, Weston and Ekseth. Prediction of Wellbore Position Accuracy When Surveyed With Gyroscopic Tools. SPE 90408, SPE Drilling & Completion, 2008.
4. Wolff, C.J.M. and de Wardt, J.P. Borehole Position Uncertainty: Analysis of Measuring Methods and Derivation of Systematic Error Model. SPE 9223, Journal of Petroleum Technology, 1981.

About this example. The well in this article is a representative scenario set in typical Llanos Basin, Colombia conditions and worked through with DrillIQ's engines. It is not a client's well, and its figures are not field results.