Intercomparison of measurements and computations of flow through a contraction and a diffuser

Before embarking on a computer modelling project to simulate actual pipe installation conditions and their effects on flowmeter performance, it was agreed at an Experts Group meeting, held under the aegis of the European Union's Bureau of Community Reference, that a more fundamental intercomparison should be carried out first. The object of this intercomparison would be to obtain quantitative experimental data on velocity distributions and turbulence levels in simple pipeline configurations and, using these, to compare the results with a range of the computational methods currently in use, in order to assess the adequacy of the latter. Two basic configurations were chosen, a 20° contraction and a 5° diffuser, both installed in long straight pipes. Conditions at a station far upstream were to be measured to provide the initial conditions for the computations. The project was carried out between 1988 and 1992, eleven European organizations participating. Of these eleven, nine made computation predictions, seven made experimental measurements using water and six undertook tests with air as the working fluid. There were significantly wide variations between the data sets obtained both by the experimental and the computational groups. For the contraction the velocity profile development was predicted with reasonable accuracy, but changes in the turbulence were not captured by any of the numerical codes employed. The diffuser represented a more stringent test and all the schemes failed to predict either the mean flow or the turbulence intensities. Overall, therefore, it was concluded that the numerical simulation of the details of turbulent flow through geometrically simple pipe components like those tested in this intercomparison cannot be achieved with the commercial programs available (up to 1991). To obtain results consistent with the modelled equations requires more care (in terms of mesh arrangements, choice of differencing schemes etc.) than was sometimes used, and, likewise, comparisons between the laboratory data sets indicated varying levels of care and consistency in the undertaking of the experiments. Turbulence models more sophisticated than those generally employed in commercial CFD programs would undoubtedly do better, and some indications of the necessary improvements are given in the paper. The level of accuracy which would have to be achieved to determine the effect of disturbances on flowmeter performance was not established and the practicali