(113c) Using Simple CFD Models to Identify Efficient Baffle Arrangements for Shell-and-Tube Heat Exchangers

Chemical reaction fouling is principally controlled by two factors: wall temperature and local fluid shear stress. Computer programs for the design of shell-and-tube exchangers generally calculate a mean heat transfer coefficient for the tube bundle. In reality both heat transfer coefficient and fluid velocity vary significantly from one location within the bundle to another location. This variation is a function of bundle geometry.

 Commercial computer codes for the analysis and design of shell-and-tube heat exchangers generally allow the engineer to look at a very wide range of geometry. For instance, the prediction methods could cover baffle cuts ranging between 15 and 45% of shell diameter and virtually any baffle spacing to shell diameter ratio.

 Designers working before the advent of these programs took a more conservative approach to baffle configuration.

 It is possible to use CFD to determine the detailed flow field within a tube bundle. However, such models are difficult to develop and to modify. Furthermore, the solution of such models demands either a large amount of computer power or computation time.

 In this paper the authors have analysed the performance of a rectangular channel containing a row of tubes. The CFD modelling of such a system is relatively straightforward and the computer power needed to obtain a solution to the model is low. Most significantly, experimental data on the distribution of heat transfer in such a system is available for one geometry operating at two different velocities.

 The predictions of CFD model coupled with a heat transfer model are compared with the available experimental data and are found to compare well.

 The model has then been extended to cover geometry covering a range of baffle cuts and baffle spacing in order to obtain a “qualitative” measure on how geometry affects heat transfer performance.

 The presence of low velocity regions at the trailing edge of each baffle and in the corners formed by the edges of the channel and the baffle plate are clearly seen and confirmed by the experimental data.

 Limits on the better baffle geometry are identified.