When the microcomputer era started, we all had to develop our own software for specific calculations. Most programs written by engineers suffered from lack of documentation, difficulty of use and narrow applicability. On the plus side, we understood the calculation choices made and what assumptions were built in (at least we should have). Now, we generally rely on commercial software. Most programs on the market today are very good — but we do need to be careful about their underlying assumptions. Unfortunately, too many engineers don’t take time to understand what the software does.
A recent issue with shell-and-tube heat exchanger design and rating software — some calculated pressure drops looked too low — points up the importance of this. Because serious work on pressure drop in shell-and-tube exchangers has been ongoing since the mid-1930s, you might assume that a consensus has arisen on what the calculation methods should be and, further, that while specifics of the calculations might differ between software packages, the values should be relatively close.
To evaluate this, we ran nearly 30 exchanger geometries and various operating conditions through five programs from reputable software vendors for heat exchanger design. We’ll discuss one small part of the evaluation here: two-phase-flow exchangers.
The good news: across the board, duty evaluation gave results in a narrow band. Variation was within 5% over the entire range of cases. With a standard approach of 10% extra area for new designs, exchangers designed with any of the programs will meet their design objectives.
The bad news: pressure drop values varied by up to 50%. The calculation methods and exact assumptions used make a big difference. As long as you know what you’re doing and understand the underlying assumptions, you should select the factors required for your situation — letting the program set them can create havoc in your design or evaluation efforts.
We found the major factors of concern to be pressure drop correlations, two-phase-flow handling, phase separation, friction factors, “bad” design restrictions and tube-count estimates.
Pressure drop correlations. A lot of correlations are available for two-phase flow, both in tubes and across tube banks. The programs used a combination of open-literature and proprietary correlations. Differences between these were large, up to 50%. Nearly half of the correlations were termed “proprietary,” with no information provided. Just because a correlation is proprietary doesn’t mean it’s wrong but some discussion of general method, range of applicability and assumptions included is needed. Without this, you never know how good your answers are.
Two-phase-flow handling. The major choices were no-slip flow versus rigorous two-phase handling. One product took no account of two-phase flow patterns; its calculation method assumes flow is always single phase. Calculations for two-phase flow in the other programs relied on the average fluid density and transport properties, which give a low number for pressure drop compared to using a flow pattern map to determine flow regime and its effect on pressure drop. This had more influence on tube-side than shell-side calculations.
Phase separation. After two phases form, the liquid and vapor can separate in the inlet, outlet or return heads (on the tube side). In this case, there aren’t equal amounts of vapor and liquid in each tube. Data has shown a range of vapor-to-liquid ratios in a single shell of more than 45:1. Phase separation tends to lower the exchanger pressure drop. Some programs assumed a uniform level of separation, ranging from none to full. Other software allowed you to specify separation at the inlet and outlet only or at the return heads as well. The same type of phase separation also can occur on the tube side.
Friction factors. Programs differed in their built-in friction factors. None allowed you to modify these. In general, this had a greater effect on the single-phase than the two-phase exchangers. Pressure drop differences of up to 15% stemmed from friction-factor values.
“Bad design” restrictions. We saw a real range. Here the issue seems to be that design programs often have limits built in to prevent bad design. In contrast, programs originally meant to rate exchangers will let you set up whatever you have, even if it’s a bad design. When a bad design limit was crossed results varied from refusing to run the problem to “fixing” the problem without telling the user.
Tube-count estimates. Can we all agree on the number of tubes that will fit in a specific diameter shell? Evidently not. One program gave a number 20% lower than the rest. Doing some manual layouts quickly showed that the tube counts it generated were too low.
So, have you checked your software’s assumptions lately? If not, you really don’t know the applicability of the results you are getting.