A mechanistic model has been developed for the rate of initiation of hydrodynamic slugs and has been coupled with improvements in the closure models for slug flow. The resulting Slug Tracking simulator is capable of predicting the slug frequency and slug length distribution, and their evolution along a long pipeline. The simulations are validated against published laboratory data, against data from the large scale Tiller loop, and against field data. For field cases, computed slug lengths and frequencies are typically within a factor of two of measured values; the frequencies compare favourably with the best available correlations.
Slug flow is a commonly occurring phenomenon in hydrocarbon production pipelines and many other process applications. Slugs can form for a variety of reasons, including operational slugging, terrain and severe slugging, and hydrodynamic slugging. Dynamic multiphase flow simulators for long flowlines typically operate with a coarse grid (spacing of order 100 times the pipe diameter or more) and correspondingly large time steps. Nevertheless, they can generally capture phenomena associated with operational, terrain, and severe slugging, since the associated time and length scales are large. In contrast, hydrodynamic slugs arise from short length-scale phenomena, developing from large waves with steep gradients over length scales of the order of the pipe diameter. Even though hydrodynamic slugs are initially quite short, the distribution of slug lengths can evolve substantially as the slugs propagate through long, undulating pipelines, sometimes leading to the formation of very long slugs, with associated problems at the receiving facilities. Furthermore hydrodynamic slugging can interact with terrain slugging in complex ways, leading to difficulties in predicting the onset and amplitude of large-scale flow instabilities. Additionally, all slugs, short or long, lead to unsteady loads on pipes and equipment, which can contribute to fatigue failure. For these reasons, it is very important to have accurate predictions of slug length and frequency.
Some progress has been made with "Slug Capturing" approaches, which attempt to resolve the two-fluid equations on a fine grid (1, 2, 3), 4, 5, 6). However, the underlying mathematical model is only conditionally well-posed, so that a mathematical solution may not exist, and the simulation results may not converge as the grid is refined. The model can be made well-posed by adding interfacial pressure or diffusion terms, but these modify the short-lengthscale features of the flow in a way that may not be physically realistic. Furthermore, the application of these simulations is severely limited by the very large computational cost for simulation of full-scale pipeline systems over operational timescales.