Bubble-Plug transition. It has been traditional to regards this transition to occur as a result of bubble coalescence leading to gradual bubble growth and the formation of large Taylor-type bubbles which occupy the whole pipe cross-section. Typically, the transition to slug flow occurs when the void fraction is around 25-30%. In highly turbulent flows, break-up of the bubbles may be postulated to occur (through this is rarely seen in actual bubble flows) to offset the progression of the coalescence. However, recent information seems to indicate that this view of the transition may be quite wrong. It seems more likely that void waves are formed in the flow, and that, within these waves, the bubbles become closely packed and are better able to coalesce, leading to plug flow. Evidence for this is provided, for instance, by Beisheuvel and Gorissen (1990).

Plug-annular flow transition. This has been an area of major controversy. Probably the main difficulty has been a semantic one. Taitel et al. (1980) define a "churn flow" which is essentially a developing plug or slug flow. However, churn flow as defined here does exist in fully developed flow, and has the following unique characteristics:

The onset of churn flow is accompanied by a sharp increase in pressure gradient, as is illustrated by the results obtained by Owen (1986) and illustrated in Figure 5. A detailed evaluation of the plug to churn flow transition in terms of this flooding mechanism is given by Jayanti and Hewitt (1992).

Churn to annular flow transition. As the gas velocity is increased after the churn flow regime has been entered, the pressure gradient initially decreases and the passes through a minimum value (see Figure 5). The flooding waves (and their associated intensive gas-liquid interactions promote large pressure gradients, and as they disappear, the pressure gradient reduces.

Figure 5. Data for pressure gradient in fully developed air-water flux in a vertical tube (Owen, 1986).

Eventually, however, the pressure gradient increases again as the gas flow rate increases. Properly speaking, one might define the onset of true annular flow as that corresponding to the point as which there is no flow reversal within the liquid film. This might correspond approximately to the pressure drop minimum (though not accurately so). Another definition might be the flow reversal point (see Flooding and Flow Reversal). It is clear that, though both churn and annular flow have the characteristic of having a liquid layer at the wall and a gas core in the centre of the pipe, their flow behaviour is quite different. The definition of the exact point of transition is, nevertheless, quite difficult.

Annular to wispy-annular transition. This is supposed to occur approximately at a critical liquid momentum flux (see Figure 3) though, again, identification of the transition is to some extent subjective.

Stratified-slug transition. The Taitel and Ducker (1976) model ascribed this transition to the onset of Kelvin-Helmholz instability and applied an inviscid form of this instability to the prediction of the transition. However, it was found this did not agree well with the actual transition and it was necessary to introduce an arbitrary correction function to bring the data into line with the predictions. A closer approximation is presented in the analysis of Lin and Hanratty (1986) who present a theory taking account of viscous effects. However, the prediction of this transition still remains an area of uncertainty.

Slug-annular transition. In the Taitel and Dukler (1976) model, this transition was assumed to occur if the equilibrium liquid level at the onset of the Kelvin-Helmholz instability was less than half the channel diameter. However, this does not appear to fit the transition well, nor are there any generally satisfactory models. Figure 6 shows a comparison of the model predictions from various approaches and it will be seen that there is not only a discrepancy between these in magnitude, but also in trend!

Slug-dispersed bubble transition. In the Taitel and Dukler and related maps, this transition is ascribed to the capability of the turbulence in the liquid phase to suspend the bubbles. When the bubbles cannot be suspended, then they agglomerate to form the gas bubble regions in slug flow. In actual fact, however, the mechanisms of bubble entrainment and behaviour in slug flow are extremely complex and need much further investigation.