These slides were presented at the NAFEMS World Congress 2025, held in Salzburg, Austria from May 19–22, 2025.

Abstract

Liquefied Natural Gas (LNG) is often stored in large tanks under cryogenic conditions. When new LNG is introduced into a tank that is already partially filled, it is important to understand how the new LNG will mix to mitigate the risk of rollover. This is when an initially stable stratification becomes weakened over time until it becomes unstable, at which point the contents of the tank can suddenly roll over, releasing a large amount of LNG vapour and presenting a potentially significant risk to the asset and personnel. This paper presents a real validation case study for the mixing behaviour within the tank prior to a potential rollover event. It does not consider the simulation of the rollover event itself, but rather the process needed to achieve confidence in CFD for exploring the effectiveness of alternative mitigation strategies within the tank for avoiding rollover. It considers the verification and validation aspects in line with the ASME V&V 10 diagram:

1) The development of a conceptual model.

2) Solution verification to understand the errors associated with spatial and temporal discretisation.

3) Validation by quantitative comparison with laboratory scale experimental data. Code verification and uncertainty quantification is not undertaken explicitly.

As a consequence of the validation study, several important aspects are identified for simulating this application with CFD:

1) The turbulence model needs to include a mechanism to allow flow to re laminarize. Away from the turbulent fountain where fluid is introduced, the flow within the tank is not fully turbulent. Without re laminarization enabled, the CFD predictions do not agree well with the experimental data, but with re laminarization the CFD predictions are in good quantitative agreement. However, the onset of re laminarization and how this is implemented within a CFD model requires further investigation.

2) The commercial CFD code used for the study (which will not be named) suffered from a restriction on the allowable time step. For very small time steps, good quantitative agreement was achieved. But as the time step was increased, it was found that at above some critical value, the simulated behaviour would change and would resemble fully turbulent flow (as if the turbulence model had been reinstated everywhere). Unfortunately the limitation on time step meant that this particular CFD code could not be used for the large storage tanks of interest, which may be around 80 m in diameter '“ a single axi symmetric simulation may take several weeks to complete.

Although alternative CFD codes were considered to overcome this limitation, subsequent discussions with industrial partners around the physics captured within the conceptual model led to a consensus that sufficient confidence could not be achieved in the CFD approach within the scope of the project. Almost exclusively, the case studies we see in the public domain are the success stories where CFD and other simulation methods have been used to great effect for some application or other, but very often these papers are very light on the VVUQ aspects. In this paper, although CFD was not used in the end, the validation study is an important real case with real lessons to be shared.