Investigation of high-flux two-phase flows and entrainment mechanisms

Interfacial area concentration and critical heat flux predictions in high-void-fraction two-phase flow depend on correctly identifying the annular sub-regime but static flow regime maps do not resolve these sub-regimes or their transient behavior. A Kohonen self-organizing map (SOM) trained on chord length statistics from a two-sensor droplet-capable conductivity probe (DCCP-2) was previously used to classify these sub-regimes up to j g ≈ 16 m / s . Here the classifier is applied, without retraining, to 51 new test conditions in a 25.4 mm ID vertical pipe extending the range to j g = 27 m / s and j f = 3 m / s . The SOM separates the extended dataset into the same three classes with minimal inter-class overlap above j g ≈ 16 m / s : annular flow at j f = 0.1 − 0.2 m / s , rolling annular at intermediate j f , and wispy-annular above j f = 1 m / s . High-speed imaging at 2000 fps for representative conditions at high j g for each class is consistent with the classification. Annular flow is marked by a thin, ripple-dominated wall film and an optically clear gas core, with disturbance waves absent below the critical film Reynolds number. Rolling annular flow shows large-amplitude disturbance waves that propagate and merge across the film, with entrainment governed by ligament break-up at wave crests and wave undercut. Wispy-annular flow is distinguished by liquid ligaments that shed from the disturbed interface and penetrate deep into the gas core, carrying entrained gas bubbles rather than breaking down into discrete. These regime-specific mechanisms are consistent across both short ( 0 − 44 ms ) and long ( 0 − 180 ms ) observation windows, distinguishing transient wave passage from sustained interfacial disturbance. The imaging is corroborated by a Kohonen self-organizing map applied to DCCP-2 chord length statistics, which reproduces the same three-class structure with well-defined boundaries above j g = 16 m / s , confirming that the visually identified flow structures correspond to statistically distinct probe signatures.