Deoiling Hydrocyclone — Engineering Fundamentals
Vortex physics, d50 correlations (Colman-Thew / Plitt), reject ratio, deoiling vs desander cyclones.
1. Why a cyclone?
A CPI plate pack settles oil droplets by gravity (1 G). A deoiling hydrocyclone (DOC) accelerates them with vortex flow producing 1,000–2,000 G effective body force on the droplet. Per Stokes, separation velocity scales linearly with G, so a cyclone can capture droplets ~30× smaller than a plate pack of the same throughput.
The trade-off: cyclones need pressure (ΔP = 50–150 psi typical, supplied by a feed pump) where CPI operates at ~zero ΔP. For a 20,000 BWPD train, the pump duty is ~50–100 HP — meaningful OPEX. CPI runs on gravity head alone.
2. Vortex physics
Water enters the cyclone tangentially at the wide end. Conservation of angular momentum makes the swirling velocity grow as the fluid spirals toward the narrow apex. Centrifugal force on each droplet:
For typical 35–60 mm deoiling cyclones at 100 psi ΔP, the effective G near the apex reaches 2,000+ G. Oil droplets (less dense than water) migrate inward toward the central air core; water exits the apex as the "underflow" / reject stream; oil-rich core exits the top as the overflow.
For deoiling cyclones (oil less dense than water): oil → overflow, water → underflow. For desander cyclones (sand denser than water): sand → underflow, water → overflow. Same hardware family, opposite plumbing.
3. The d50 correlation
The 50%-cut droplet diameter — droplets larger than d50 are captured, smaller pass through — is the master sizing parameter. From Plitt (1976) adapted to deoiling by Colman & Thew (1980):
Knobs:
- Smaller Dc → smaller d50. A 35-mm cyclone catches 8 µm droplets; a 70-mm cyclone catches only ~15 µm. The trade-off is per-liner capacity drops with Dc1.8.
- Higher ΔP → smaller d50 (∝ 1/√ΔP). Going from 50 to 200 psi halves d50 — but doubles the pump duty and the wear rate.
- Higher T → lower μw → smaller d50. Warm produced water (150 °F) cyclones 50% better than cold (60 °F).
- Higher ΔSG → smaller d50. Heavy crude on dense brine helps; condensate on fresh water hurts.
4. Capture-efficiency curve
Plitt's log-normal fit:
where m is the "sharpness" exponent (≈ 4 for deoiling cyclones, ≈ 3 for desander). At d = d50, E = 50%. At d = 2·d50, E = 94%. At d = 0.5·d50, E = 6%.
Real inlet streams contain a distribution of droplet sizes (usually log-normal centered at ~50 µm post-CPI). Total bulk capture is the integral of E(d) weighted by the inlet droplet distribution — typically 70–95% for well-sized deoiling cyclones.
5. Reject management
The overflow (oil-rich reject) is typically 1–5% of inlet flow with ~10× the inlet oil concentration. For 20,000 BWPD at 3% reject + 300 mg/L inlet, the reject is 600 BWPD at ~10,000 mg/L oil. This stream typically goes to:
- A reject vessel (skim tank) for further oil-water separation; recovered oil goes to slop and reclaim, water back to the inlet of the deoiling train.
- The atmospheric tank battery as part of routine slop handling.
Larger reject ratio → better capture but more water cycled back through the system. The 3% default is the industry sweet spot.
6. Deoiling vs desander cyclones
| Deoiling (DOC) | Desander | |
|---|---|---|
| Target phase | Oil → overflow | Sand → underflow |
| Body OD | 35–70 mm | 150–500 mm |
| L/D | 10–20 (long) | 3–5 (short) |
| Reject | Overflow, ~3% vol | Underflow, ~5–10% vol |
| d50 target | 10–20 µm | 10–100 µm sand |
7. References
- Colman, D.A. & Thew, M.T. (1980). "Hydrocyclone to give a highly concentrated sample of a lighter dispersed phase." Conference paper — BHRA Cranfield.
- Plitt, L.R. (1976). "A mathematical model of the hydrocyclone classifier." CIM Bulletin, 69(776).
- Bradley, D. (1965). The Hydrocyclone. Pergamon Press, Oxford.
- Svarovsky, L. (1984). Hydrocyclones. Holt, Rinehart and Winston.
- API Publication 421 — companion / upstream context.
- Vendor literature: Cameron Vortoil, KOS Cyclonics, eProcess, Schinazi, FMC Technologies.