Performance and implementation of the Launder-Sharma low-Reynolds number turbulence model

The Launder-Sharma model (Lett. Heat Mass Transfer 1, 1974, 131-138) is one of the earliest and most-widely used low-Reynolds number (LRN) models and has been shown to be in good agreement with experimental and DNS data for a range of turbulent flow problems, performing better than many other LRN k-ε models. However, some recent studies including those using commercial CFD solvers report the model to behave otherwise. In the present study, the LS model has been implemented in FLUENT using user-defined functions (UDFs), and its performance in predicting steady and unsteady turbulent flows has been tested and found to agree closely with those reported in the literature using 'in-house' CFD codes. However, this FLUENT-UDF LS model performs very differently from the in-built FLUENT LS model. The former agrees well with experimental and DNS data, whereas the latter does not. In addition, the UDF is used to demonstrate that the model predictions are very sensitive to the interpretation of the model formulation. Consequently, it is concluded that whereas the LS model is not sensitive to numerical method or method of coding, it is sensitive to changes in the interpretation of the formulation of the model.