Engineers at Solutia (St. Louis) had a dilemma: how to prove that a proprietary flow distributor would outperform and save money over a baffle in the design of a shell-and-tube heat exchanger, and how to communicate these results with a non-technical audience.

While heat exchanger design typically focuses on surface area requirements, the fluid flow patterns within the tubes can be of equal importance. One optimization target is the equal distribution of flow into each tube within the bundle; in an unmodified heat exchanger, this rarely happens. Rather, the tubes in the middle of the bundle receive the bulk of the fluid because of inlet fluid impingement. The tubes on the side of the bundle get the fluid that works its way over. This tube-side flow mal-distribution produces a condition in which much of the heat transfer surface area is wasted. Consequently larger, more costly heat exchangers are used to meet the original design specifications.

A common method of avoiding undesirable tube-side flow patterns involves putting a baffle up into the head of the heat exchanger. This causes more of the fluid to work its way around to the far reaches of the tube bundle and improves flow distribution. Solutia decided to try a new approach that would improve on the baffle idea. Its plan was to design a proprietary flow distributor that would fit into the head of the heat exchanger and direct fluid more precisely than a baffle.

CFD ANALYSIS AND DESIGN

In designing the flow distributor, Solutia engineers decided to use a commercial postprocessor for computational fluid dynamics (CFD) analysis to direct their optimization efforts. They used the CFX4 CFD code from AEA Technology (Bethel Park, PA) to quantify the baseline performance of an unmodified heat exchanger and a heat exchanger equipped with a baffle. They also used it to compare the performance of different flow distributor designs against each other. The optimal design was the one that distributed fluid most effectively to all tubes.

The CFD analysis provided fluid velocity, temperature, and concentration values throughout the heat exchanger model. The results guided Solutia engineers towards optimized fluid flow patterns by assessing changes to the flow distributor geometry. Although CFX4 included a postprocessor that could present the analysis results visually, Solutia opted to use a commercial post-postprocessing program called Fieldview from Intelligent Light (Lyndhurst, NJ) instead.

One reason Solutia preferred Fieldview for this project was the ease with which it provided information about the analysis. For example, engineers wanted to determine velocity components for normal and tangential flow directions instead of Cartesian or x, y, and z directions. Normal velocity components gave a direct measure of fluid distribution to all the tubes within the bundle. This normal velocity was used as an optimization criterion. Obtaining this value would have been difficult with the base CFD code or most post-processors. It would have first required the creation of a FORTRAN program to calculate the new normal flow component. Then the CFD analysis would have to be repeated, the new results stored, viewed, and finally used to evaluate the optimization criterion. With Fieldview, a tool called the CFD calculator was used. To calculate the normal velocity, the engineer selected "velocity" from the scrollable menu, applied the proper trigonometric function to obtain the normal component, and gave the new variable a name. The calculations needed to derive new values were done entirely through menu selections. Engineers were also able to perform other calculations on the fly, without having to rerun their simulations.

Knowing the normal component of velocity was valuable, but even more valuable was determining the total flow into the tube bundle at a particular location. The engineer determined this by using the Fieldview integration capability on the normal velocity component. Another command, called "sweep," automatically obtained the total flows by integrating the normal velocity across the entire tube sheet. This provided a quantitative assessment of the uniformity of the flow going into the tube bundle.

This information was also presented graphically in a slightly different but even more illuminating way. After making the assumption that the total flow rate in the system was constant, the engineer determined what fraction of the fluid flow each tube should get if the flow was distributed perfectly. The quantitative assessment of flow going into each subset of tubes was divided by its fair share value. Using the postprocessor's ability to make 2-D plots, the engineer created a plot of normalized flow vs. its position in the tube bundle. This plot was re-created for each different flow distributor configuration, as well as for the comparison configurations using Fieldview scripting. The ideal flow situation was a flat horizontal line, and it was immediately clear from these plots that the flow distributor did a better job of achieving this than an unmodified heat exchanger or one with a baffle.