A verification and validation study of the application of computational fluid dynamics to the modelling of lateral sloshing

Abstract

An understanding of liquid sloshing is of primary concern to the design and operation of Liquefied Natural Gas (LNG) carriers. Safe operation of LNG carriers requires the knowledge of global and local pressures imposed by the sloshing liquid. The most general method available to quantify such sloshing loads is the solution of the Navier Stokes system of equations using Computational Fluid Dynamics (CFD). Given the wide variety of modelling options available, as yet there is no consensus on the best modelling practice for such sloshing flows.

This report seeks to address this issue, examining various models and identifying the most suitable combination. The work uses the commercial CFD code ANSYS superscript TM CFX-10.0 superscript TM but most of the findings are also relevant for similar other commercial codes. The physics of the sloshing problem are considered in order to identify the key modelling aspects. The correct application of CFD and how it can be used to model sloshing is considered. A suitable experimental dataset is described for use as a validation test case. The sloshing problem simulated is in a 1.2 m long and 0.6 m high tank with a 60 % filling level; excited at 95% of the first natural frequency with a maximum displacement of 1.25 % of the tank length.

A space and time discretisation independence study is carried out to ascertain the applicability of the results. Subsequently, the effect of including either a k ? ? or Reynolds stress turbulence model as opposed to forcing laminar flow is examined. The choice of fluid (water and air) compressibility is investigated to determine its effects on model accuracy as well as the associated computational cost. Results are compared to experimental data and a computational reference case.

It is found that a grid of 6000-7000 elements with an initial node wall offset of 1 mm is sufficient to achieve effective grid independence for sloshing in. The necessary time discretisation scheme was determined to be second order with a dynamic timestep adaptation scheme controlled by a root mean square Courant Number of 0.2. The flow regime should be considered as turbulent and the standard k ? ? turbulence model is suitable. Finally it is observed that a compressible-incompressible model combination for air and water respectively gives a near identical result to a fully compressible model with a 20% reduction in computational time.

More information

Identifiers

ORCID for S.R. Turnock: orcid.org/0000-0001-6288-0400

Catalogue record

Export record