Just ask the NASA scientists who worked on last month’s Mars probe mission: sometimes you have every reason to think a piece of equipment should run smoothly, but it just doesn’t. That’s the kind of mystery a German chemical company faced with regard to a major heat exchanger, and as with the Mars probe, setting eyes on the situation was difficult. The installed 324-tube exchanger was living up to just a fraction of its theoretical performance potential, but the explanation might as well have been written on the surface of another planet.

A Bundle of Questions

When all conventional approaches failed to yield answers, the company solicited help from across the English Channel. Cal Gavin (Birmingham, England) is a process-oriented chemical engineering company that specializes in the performance and economics of processing fluids. What it found was an 700mm-wide, 2,500mm-long vertical exchanger had 200-mm nozzles on the tubeside. It was designed to operate with a tubeside flow of 90,757 kg/hr at 69.5C with the fluid density at 873kg/cu meters and viscosity of 1cP, yet the heat exchanger was providing far less thermal duty than necessary for the application.

Cal Gavin engineers recognized that, in theory, the heat exchanger should have been able to meet required duty, and they suspected a fluid flow distribution problem. They also recognized that short of fixing this unit, the alternative solution to this problem would be to replace the exchanger with a larger unit, involving significant installation work and expense.

While heat exchanger design typically focuses on surface area requirements, the fluid flow within the tubes can be of equal importance. A shell-and-tube heat exchanger consists of a bundle of tubes through which one fluid flows while the other fluid flows around the tubes. Should most of the fluid flow through just a few tubes, the majority of the installed surface area is wasted.

The next important consideration is fluid flow within individual tubes. Frictional drag at the wall and viscous shear forces within the fluid create a velocity profile with maximum flow at the center of the tube and zero flow at the wall. Even where turbulent flow is fully established, a significant boundary layer still persists in both single and two-phase flow regimes.

Furthermore, flow conditions are typically assumed to be ideal because there is no simple way to make experimental measurements within an exchanger. The building of a model to assess the flow is possible but would be expensive, time consuming, and would not provide quantitative results without special instrumentation.

Stuck in the Middle

In the last few years, however, advances in computational fluid dynamics (CFD) simulation have made it possible to provide engineers with a graphical portrait of flow within a heat exchanger, with limited expense and time constraints. CFD involves the solution of the governing equations for fluid flow, heat transfer, and chemistry at several thousand discrete points on a computational grid in the defined flow domain.

The use of CFD enables engineers to obtain solutions for problems with complex geometries and boundary conditions. A CFD analysis yields inter alia values for fluid velocity and temperature throughout the solution domain. Based on the analysis, a designer or engineer is able to optimize fluid flow patterns or temperature distribution by adjusting either the geometry of the system or the boundary conditions such as inlet velocity/temperature, wall heat flux, and so on.

In this case, subsequent CFD simulation using software from Fluent, Inc. (Lebanon, NH, www.fluent.com), which also offered the solution-adaptive grid capability to refine the grid displayed, “magnifying” sections to allow for more specific analysis. It showed that nearly 70% of the exchanger pressure drop was lost across the nozzles. The remainder was insufficient to evenly distribute the fluid through the tube bundle. Further analysis indicated adding inserts to increase the flow resistance of the bundle could solve the problem.

Cal Gavin engineers prepared a model of the client’s exchanger, using a porous media to model the tube bundle. This greatly reduced modeling and execution time by eliminating the need to produce a geometry for each tube and thereby reducing the size of the domain. The analysis showed that as flow passed from the 200-mm nozzle, the momentum pressure loss of expanding into the header was greater than the frictional pressure loss across the bundle.

Hence, the majority of the fluid remained concentrated in a few tubes at the center of the bundle. The velocity in the center of the bundle was more than double that at the periphery, leaving about 75% of the surface area ineffective.

The Fix Is In

Cal Gavin engineers then modified the model to investigate the effects of adding wire matrix inserts. Their solution was to increase the resistance to flow across each tube and thereby even out the flow distribution in the bundle. At the same time, the inserts would create more ideal flow conditions within individual tubes by continuously removing stagnant fluid from the tube wall and replacing it with fluid from the centre of the tubes. This reduces the residence time of fluid in contact with the heat transfer surface and creates a flow regime where the velocity profile across the tube is nearly flat.

Although wire matrix inserts depend upon an increase in pressure drop to achieve these flow conditions, the enhancement in heat transfer is much larger than would be achieved by simply increasing tubeside velocities. In fact, heat transfer enhancement is usually achieved at much lower superficial velocities than would be employed in plain tube designs. Thus it is often possible to reconfigure a bundle to reduce the number of passes and maintain or reduce the original pressure drop while still dramatically increasing the performance of the unit.

Reanalyzing the heat exchanger with the inserts added showed that the flow distribution was now quite even across all of the exchanger. The result was a dramatic improvement in heat transfer efficiency with all the installed area now effective.

Based on known data from similar heat exchangers, Cal Gavin engineers estimated that the heat transfer of the exchanger would be increased by a factor of nine with the relatively minor cost of installing the wire matrix inserts.