Two-Phase Flow Regime Selection Guide

When to worry about slug flow, when a Lockhart-Martinelli correlation is good enough, and when to call for transient multiphase simulation. A practical guide for designing and troubleshooting gas-liquid lines in process plants.

Why This Matters

Two-phase flow breaks single-phase intuition. Pressure drop is not linear with velocity, the flow can be stratified or annular or slugging depending on velocity ratios and pipe orientation, and the difference between regimes drives real equipment decisions: pipe diameter, slug catcher sizing, vibration mitigation, meter selection, control stability.

The dominant mistake is assuming "two-phase flow = slug flow = bad." Most gas-liquid systems don't slug. But ignoring regime maps altogether puts you at risk of designing a line that spends its life hammering flanges loose. This guide gives you a decision tree that matches the regime to the design response.

Regime Identification (Horizontal Lines)

Horizontal two-phase regimes, in rough order of increasing gas velocity:

Use Mandhane or Baker maps for horizontal lines, Taitel-Dukler for rigorous prediction, and Hewitt-Roberts for vertical lines.

Decision Tree

Step 1 — Collect inputs. You need phase mass flows, densities, viscosities, surface tension, pipe diameter, and orientation. Conditions at line conditions, not standard.

Step 2 — Compute superficial velocities.

U_sg = Q_gas / A_pipe    (gas at line conditions)
U_sl = Q_liq / A_pipe

Step 3 — Plot on Mandhane map. U_sg on x-axis (ft/s), U_sl on y-axis (ft/s), log-log. Identify regime.

Step 4 — Apply design response by regime:

If Stratified Smooth/Wavy

• Normal pressure drop methods work (Lockhart-Martinelli, Beggs-Brill).
• Design for liquid holdup — downstream separator volume must absorb transient changes in holdup (e.g., when gas rate drops).
• Watch for corrosion at the waterline in wet sour or wet CO2 service. CRA cladding or inhibitor strategy.

If Slug Flow

• This is the regime to avoid by design. If you can change pipe diameter, orientation, or inclination to move out of slug region, do it.
• If unavoidable: size downstream vessel for slug catcher duty. Use a transient simulator (OLGA, LedaFlow, or equivalent) to predict slug length at severe operating conditions.
• Restrain piping against slug forces. API 14E erosional velocity is not the relevant limit — mechanical slug forces on elbows dominate. Compute F = rho_slug * A * v² at each direction change.
• Avoid control valves in slug flow; move them to downstream of separation. If unavoidable, specify anti-cavitation trim and expect erratic control.

If Annular

• Erosion is the concern. Apply API 14E C-factor (100 for continuous service, 125 for non-continuous), but know it's conservative for clean service and non-conservative for sand-laden.
• Heat exchangers in annular flow: beware film dryout at high heat flux — dryout transition causes local tube overheating.
• Instrumentation: capacitance probes and Coriolis meters are the honest performers here; DP flow meters are deeply unreliable.

If Bubble / Dispersed Bubble

• Usually benign. Single-phase liquid design methods with homogeneous dP multiplier (~1.1-1.3) are adequate.
• Watch for gas accumulation in low-velocity zones (dead legs, instrument taps).

When to Escalate to Transient Simulation

Steady-state regime maps are not enough when any of the following apply:

Escalate to OLGA/LedaFlow if: (1) line is long (> 1 km) with elevation changes, (2) upstream slug flow expected, (3) terrain slugging is possible (pipeline with dips), (4) severe turndown is in the operating envelope, or (5) startup/shutdown dynamics drive the equipment design.

For short plant piping (< 100 m, relatively level), a Mandhane-map check plus Lockhart-Martinelli pressure drop is usually sufficient. For anything off-plot or offshore, you need transient.

Pressure Drop — Which Correlation

Common Mistakes

1. Using standard-condition flows instead of line-condition flows. Gas density changes the superficial velocity ratio dramatically.
2. Ignoring pipe inclination. A 1-degree uphill line behaves very differently than horizontal at low rates.
3. Trusting API 14E alone. C-factor is a screen, not a design. For sand service or multiphase, requires erosion modeling (DNV RP O501).
4. Assuming a 3:1 turndown works. In slug-prone service, the operating envelope at turndown may be a different regime than at design rate.
5. Skipping transient sim on long pipelines. Terrain slugging can generate slugs many times the pipeline inventory — no steady-state method predicts this.

Quick Reference — Erosional Velocity (API 14E)

v_max (ft/s) = C / sqrt(rho_mix_lb_per_ft3)
  C = 100  (continuous service, clean)
  C = 125  (intermittent, clean)
  C = 150-250  (continuous with corrosion inhibitor and clean)
Sand service: use DNV RP O501 instead.

References

• Mandhane, Gregory, Aziz (1974) — horizontal flow pattern map
• Taitel & Dukler (1976) — mechanistic flow regime transitions
• Beggs & Brill (1973) — pressure drop correlation for inclined flow
• API RP 14E — offshore piping design (erosional velocity)
• DNV RP O501 — erosive wear in piping for sand service
• Shoham (2006) — Mechanistic Modeling of Gas-Liquid Two-Phase Flow in Pipes, SPE
• Brill & Mukherjee (1999) — Multiphase Flow in Wells, SPE Monograph 17