High velocity in heat exchanger tubes 2

[tim02]

In normal the operation of our cross flow heat exchanger the rate of heat transfer goes up proportionally to the cold side flow and we control exit temperature that way with a PID loop program operating the cold side throttle valve. But we seem to be able to reach a condition when the flow rate reaches a certain point the rate of heat transfer starts to drop, so the throttles open more and the rate goes up more so the rate of transfer drops even more and continues into an out of control situation.
I seem to remember something from an Advanced Fluid Flow course (about a zillion years ago) where high velocity flow drastically reduces efficiency but I don't remember any details or even what the phenomenon is called.
Can someone explain to me the mechanism by which cooling water flow above a certain velocity in the cold side tubes will drastically reduce the heat transfer ability? And what solutions either in design or controls do you suggest?

 

[vpl]

tim02

Getting back to the original question - heat exchanger efficiency and velocity are related through the Reynolds number, and if you normally operate at a very low flow rate (such that you're in the laminar region) and the increase is just enough to put you in the transition region rather than get you to turbulent flow, then you could experience a large drop in heat transfer. JOM asked you to give us some data. If you have information such as your normal operating temperatures, normal flow rate, the flow rate when things go whacky, and the type of heat exchanger you could check for this.

Patricia Lougheed

Please see FAQ731-376 for tips on how to make the best use of the Eng-Tips Forums.

 

[tim02]

More System Details and Description

The HX is has 4 cross flow tube bundles in a row separated into 2 parallel stages (a "hot end" stage and a "cold end" or process controlling stage. Each stage has its two tube bundles connected in series (1st bundle flows across and the other flows back).
Cold supply header water enters each stage into the bundle furthest downstream relative to the process flow and discharges out of the upstream bundles. The cold side fluid is relatively clean, chemically treated, closed loop water. The hot side process flow media is air-fluidized sand. The air serves predominantly to keep the sand bed of approx 7 tons fluidized so it will flow and only incidently provides 3-4% total cooling. Cooling water supply temperature ranges from 50 to 90F and discharges with a wide range of delta T greatly dependent on flow rate. The process sand flows at 45 tons +/-2 per hour at 650+/-25F and discharges at 120+/-5F.
Under normal steady operation the cooling water flowrate throttles in a range of 200 to 300 gpm per stage. But when the unit begins to lose controllability, it gradually opens throttles up to near 500 gpm and at that point seems to lose all control driving throttles completely open and achieving up to approx 1000-1200gpm per stage with a negligible to 10F Delta T, while sand exit temperatures continually climb at a slow rate. We then use the program to drive the throttles to close to about 400 gpm and the Process sand exit temp comes back down. We have since programmed a throttle max clamp position but the system still gets into a condition where it drives both stages throttles to the max positions and still won't cool.
I'm chasing my tail trying to figure out what's happening.

 

[JOM]

tim,

I think perhaps we all assumed it was a liquid/liquid shell and tube exchanger(s). Fluidised sand? That's certainly two-phase!

Is there any possible alternate path for the water to take at the high flowrates? Can it be bypassing the heat transfer surfaces?

"We then use the program to drive the throttles to close to about 400 gpm and the Process sand exit temp comes back down."

Bizzare. Cheers,
John.

 

[dooron]

I have not designed a heat exchanger for some time but referring to a book "Compact Heat Exchangers" by Kays and London, Fig. 10-1 you find that at Reyonlds No. 1500 to 3000, transition range, the heat transfer rate drops below the laminar or turbulant range-i.e heat transfer does not always increase with increased velocity. Maybe your flow works in this range, as some of the others have suggested.

 

[tim02]

Yes it is fluidized sand but...
... that doesn't make it a two-phase heat exchanger. It has 2 different fluids (the sand flows essentially as a fluid) but neither changes phase. Two-phase HX implies a phase change (i.e. a turbine exhaust condenser).
Dooron, thanks for the reference. I think that might put me on the trail of the phenomenon I was looking for.

Ciao for now,
Tim

 

[JOM]

tim02

i guess i have to eat my words - increasing velocity can result in a decreased hx co-efficient according to dooron.

You go from 500 to 1000 gpm, so the Reynolds no. is doubled. Dooron's transition zone is over a factor of 2 as well. If that is the problem, then is your solution to push the flow past 1000 gpm and ensure flow is turbulent? Cheers,
John.

 

[25362]

Using Sieder-Tate and other formulas and graphs I couldn't confirm the strange behaviour mentioned by Dooron on laminar or transitional flow regimes. The only possibility of getting lower HTC at higher Reynolds number was to reduce the Pr number. But on using cooler water Pr values go up, so does the viscosity to the 0.14 power. The only value that drops at lower (high velocity) water temperatures is the thermal conductivity, but it's too little to compensate the others going up. So, my conclusions were that the higher the Re No. the higher the HTC for the case in question.

Of course, there may be new formulas that contradict older conventional approaches. And I may be wrong.

Anyway, I'd like to raise two more options:

1. Is there an enthalpy-and-mass balance showing that one set doesn't steal water from, or receives more process fluid than, the other set ?

2. Are the inlet nozzle water linear velocities lower than the tube average velocities ? Water maldistribution happens when axial entrance velocities are greater than tube velocities.

tim02, please keep me posted. Thanks.

 

[tim02]

Whatever is causing this, 25362,
I believe it must be flow related on the cooling water side.
The fluidized sand has only one flow path available, entering down a chut at one end, through the heat exchanger shell horizontally across all 4 tube bundles in series, and out a chute a the other end. The water side, however, enters from an constant pressure (20psi) 8" header then splits into each of 2 independently throttled 6" parallel stages, and through 286 tubes per tube bundle. It seems that there may be some possibility that the is developing a very uneven flow distribution profile through the bundles above a certain velocity. My theory is that it is then effectively not using all the tubes, and therefore bypassing some heat transfer surface area, so the overall capacity of the unit effectively goes down as does the efficiency of the "working" portion of the HX. (i.e. using the basic heat tranfer equation Q'=UA(Tout-Tin)ave, the UA decreases in the equation). Even so, I don't know how to prove this is what's happening.

 

[25362]

I pressume you can not measure the water flow rates to each set and the temperatures around each set on both (process and water) sides to make the needed heat balances.

You must find a way of measuring these to corroborate your theory.
Good luck !

 

[25362]

tim02, I found an old (late 1960's) article in CE by Scaccia and Theoclitus (McGraw-Hill, New York) that deals on "types, performance and applications" of heat exhangers, in general, and cross-flow types, in particular.

They say that a cross-flow exchanger losses effectiveness when the fluids are "mixed". In "unmixed" tubeside fluids (water in this case) there would be temperature profiles across the bundles resulting from "cold" and "warm" corners in the exchanger. When the tubeside fluid becomes "mixed" because of high turbulence (larger flows in this case) the thermal effectiveness drops and the temperature profile disappears.

Strange, isn't it ?

In short, it appears that this seems to be a characteristic of cross flow exchangers. It would seem appropriate to look for specialized literature on this type of compact units.

 

[reactorshell]

25362,

This is one interesting finding. I'm not sure whether I get it intepreted correctly though. Are you saying that for cross-flow HX, it would not be advantageous to operate the shell side at higher RE because thermal effectiveness decreases? What do you mean by the temperature profile disappears? Do you mean that the profile is linear in this case?

 

[hacksaw]

tim02,

you've got a 560 deg F delta T across the tube surface.


that is a bit excessive don't you think?

water at 20 psi boils at 258-259 Deg F.

At low cooling water flow, everything is more or less okay, but you have a conditionally stable process.

Open the CW disch. valve and you drop the pressure, poof!, you immediately develop local hot spots in the tube that prevent nucleate boiling,local superheated zones).

The air/sand exit temperature immediately rises (sees an effective loss of surface area), at which point your controller asks for max CW water flow. You eventually drive the temperature down....

you need to put a pressure transmitter on the cw dsch. (up stream of the valve) to allow temperature modulation of the CW valve with a low pressure constraint.

you can also achieve same with electronic limits on the controller output or mechanical stops on the valve.

 

[25362]

to reactorshell, the article defines the temperature (or recuperative) effectiveness as the ratio of the temperature drop (cooling) on the process (hottest) fluid to the difference between the extremes: the hottest (fluid) and the coldest (water) inlet temperatures.

It also says that cross-flow units show their effectiveness to be in between those of counterflow and parallel flow heat exchangers.

A temperature profile or better a gradient of temperatures appears at the cold fluid outlet, when water is not "mixed" meaning that the water leaving the tubular shows a range of temperatures from coldest at the lowest point to warmest at the highest. According to the article, when this happens the thermal effectiveness is highest, when the streams mix (as by turbulence) this effectiveness drops and the oulet gradient disappears.

As the mixing takes place the effectiveness lowers and becomes nearer to that of a parallel flow exchanger.

 

[25362]

The CE July 1990 issue contains an article under Engineering Practice named: "Quick design and evaluation: H/E" by Bowman and Turton, that brings worked out examples on cross-flow exchangers.

Using their own data and graphs, when doubling the flow of cooling water, we obtain, of course, a drop in water outlet temperature, with a rise (!) in the hot fluid outlet temperature, contrary to expectation. The same as in the case in question.

They base themselves, among others, on Taborek's "Heat Exchanger Design Handbook" (ed. by E. Schlunder - Hemisphere Pub. Corp. Washington, D.C. 1983) and on "Design of Heat Exchangers" by Turton and others in CE Aug.18, 1986, both of which I don't have access to.

The trick offered in the article is based in using graphs (as usually used for LMTD correcting factors) showing precalculated values of NTU (number of transfer units) = UA/mc, where m,c represent mass flow rate, and specific heat, of the cold fluid, respectively. These graphs are meant to circumvent the hard work of calculating T₂ and t₂ via NTU.

Readers are invited to comment on the subject.

 

[vpl]

tim2

Do you know the diameter of the tubes? I have a spread sheet that does a quick and dirty calculation of Reynolds # for cross flow heat exchangers, and other heat transfer factors but need the tube diameter. Patricia Lougheed

Please see FAQ731-376 for tips on how to make the best use of the Eng-Tips Forums.

 

[tim02]

Tube dimensions:
4.5 foot long tubes with diameter 29/32".
Thanks,
Tim02

 

[hacksaw]

she'll need the wall thickness too

 

[25362]

If I got my figures right all water flow rates are in the laminar regime and
WT is 0.5(31-29)= 1". Right?

 

[hacksaw]

at nominal 1 gpm /tube you a nominal pipe reynolds of 5000, if i have consolidated the design details properly. it is definitely not in the regime of well developed turbulent flow.

tim02, what kind of cycles(temperatures) do you see in the cw exit of the hot stage?

 

[tim02]

Hacksaw-
CW supply temps cover a wide range (40 to 90F) as they are driven by ambient. Exit temps are nominally at a 50 to 70F delta. But I've seen exit excursions up as high as 175F and down to 90F during normal operations and almost regardless of inlet temp. There is no control on the water exit side temp and every cooler seems to settle into a different delta range even though they share a common supply header. When one starts losing controlability the delta drops to 10F or less.
And 25362, the tube I.D. is 29/32" with wall thickness .0183".