A predictive model for rollover in stratified LNG tanks

Abstract

The incubation period preceding "rollover" within a stratified LNG tank involves intensive heat and mass transfers between layers. Numerical integration of equations describing these processes leads to predicted time-history and boil-off characteristics which are in excellent agreement with Sarsten's (1972) even in well-insulated tanks, cause a slow boil-off of the LNG, and this requires removal of some vapor. During this "weathering" process the composition of the LNG changes because the small amount of nitrogen present is much more volatile than the methane and the heavier hydrocarbons are effectively nonvolatile. Obviously the temperature and density of the LNG change as well. Natural convection causes circulation of the LNG within the tank, maintaining a uniform liquid composition. The addition of new liquid, however, can result in the formation of strata of slightly different temperature and density. The energy and mass iransfer between these strata (or cells), given sufficient time, will result in an unstable interface and rapid mixing of the cells. This is called rollover. When the lower, more dense cell is also at a higher temperature, rollover causes the vapor pressure at the top of the LNG to rise very rapidly, and a rapid evolution of a large quantity of vapor results. At best, such a situation results in a loss of valuable fuel. At worst, it could be dangerous.In operating an LNG storage facility it would be helpful to have a reliable method for predicting the occurrence and severity of a rollover and for evaluation of countermeasures to prevent it. This in turn requires a good understanding of the energy and mass transfer between the cells and of the rate of boil-off. A mathematical (computer) model of the dynamics within the tank, with provision for addition and removal of LNG, would be a useful tool.Only one rollover incident, at LaSpezia in Italy, has documentation sufficient for testing a computer model of the rollover process (Sarsten, 1972). Geist (1972, 1976) and Germeles (1975) developed models which omitted the important contribution of nitrogen to the density and volatility of the LNG; they used the thermohaline equations of Turner (1965) and Huppert (1971); and introduced simplifications which required "adjustment" of the data supplied by Sarsten.The present paper gives a more adequate theoretical framework for rollover analysis and presents quantitative computer results for the simulation of the LaSpezia rollover. CONCLUSIONS AND SIGNIFICANCEA model has been developed which closely simulates the LaSpezia incident details using only input data from Sarsten (1972). It contains only two adjustable constants. The first is the fraction of the heat leak into the vapor space which is transferred to the vaporizing film; changing this fraction from 0.95 to 0.85 has no sensible effect on the dynamics of vaporization and rollover. The second is the proportionality constant in the expression for the intercellular heat transfer coefficient. The rollover time is acutely sensitive to this constant. A...

Optimization methods for the real-time inverse problem posed by modelling of liquefied natural gas storage

“…Heestand et al [6]; and Bates and Morrison [7]). Chaterjee and Geist [3] considered only two chemical species: methane and non-volatile heavy hydrocarbon and the rollover criterion considered in their approach was equal temperature and composition of the stratified layers.…”

“…Heestand et al [6] considered the five most common constituents of LNG namely methane, ethane, propane, n butane and nitrogen. Heestand et al [6] argued about the use of thermohaline heat and mass correlations of Turner [8] in the previous models, as those correlations were provided for salt-water experiments and claimed that these correlations significantly under-estimate mass transfer between the stratified layers.…”

“…Heestand et al [6] argued about the use of thermohaline heat and mass correlations of Turner [8] in the previous models, as those correlations were provided for salt-water experiments and claimed that these correlations significantly under-estimate mass transfer between the stratified layers. Instead, Heestand et al [6] assumed fully turbulent conditions inside the LNG storage tank and used the correlation of Globe and Dropkin [9] for heat transfer between two horizontal plates heated from the bottom. Heestand et al [6] also reported that rollover predictions are acutely sensitive to proportionality constant in empirical correlation and +20% change in this constant can lead to -15% change in predicted time to rollover.…”

“…Instead, Heestand et al [6] assumed fully turbulent conditions inside the LNG storage tank and used the correlation of Globe and Dropkin [9] for heat transfer between two horizontal plates heated from the bottom. Heestand et al [6] also reported that rollover predictions are acutely sensitive to proportionality constant in empirical correlation and +20% change in this constant can lead to -15% change in predicted time to rollover.…”

“…Heestand et al [6] assumed a thin film at the top surface of LNG, which is in thermodynamic equilibrium with the evolving vapours instead of the entire content in the upper layer. The same approach (Heestand et al [6]) is used in the present work, as it explains the peak in boil-off rate at the time of rollover.…”