Interpreting a Flow Curve 3

[Jiks]

Hi, I have a question regarding flow curves of a series of cooling channels drilled into a thick cylinder. Water is pumped through the channels to cool the cylinder. I have theoretically calculated the flow curves for the cooling channels with different diameter holes. A sketch of the channels is shown in this link

http://img53.imageshack.us/my.php?image=flowpath6pw.jpg

When I overlap the pump curve for the pump we currently use the curve doesnt intersect the system curve. The current diameter of the hole we use is 25mm and I was hoping to increase the size of the hole to 32mm. How do I interpret the following curve

http://img53.imageshack.us/my.php?image=flowcurve0tu.jpg

What is currently happening in the system? How do I know if the holes are running full with water? What would happen if I move to a 32mm dia hole? How will it affect the heat transfer coefficient?
Could anyone please help me?
Thanks.
Jiks.

 

[bulkhandling]

Assume all system curves are accurate - pressure drop included all system losses such as friction loss in piping, loss in heat exchanger and static head (if any), the pump is running out. At the end of the curve, the pump will pump the maximum flow. Due to the constraint of pump configuration, when system pressure is further lower, the pump will still run that maximum flow- or may be little more. Running out is considered not a stable operation condition.

However, your pump curve looks like not complete. Some pump manufacturers do publish some imcomplete curves since they do not wish the customers run the pumps at the close-to-run-out condition.

You change the hole from 25mm to 32mm only make thing worse.

As for the heat exchanger coefficient, it's your job to check all the parameters as liquid temperatures in and out, calories or KJ transfered, total heat transfer coefficient, and cooling media flow rate. If you do need more cooling water for the process, a larger pump may be required.

 

[davefitz]

You can cause the system curve to intersect with the pump curve by adding pressure drop at the inlet to each circuit using inlet orifices or inlet capillary tubes.

If here are only 2 circuits and you expect to use the component for a long period of time, thenI would recommend using 2 capillary tubes, each would be made of stainless steel tubing , and the diameter of these tubes would be as required to yield the added pressure drop to intersect the pump curves.

If the fluid is not boiling, then the inside film heat transfer coeficient will increase by the 0.8 power of the water velocity, but the overall factor of (A,i*h,i) would be larger if a large diameter hole is used. You would only use a smaller diameter hole if you needed to avoid a boiling crisis in the tubes; gnerally you should strive for a fluid mass flow of 750,000 lb/hr/ft2 to avoid boiling issues.

 

[Jiks]

Thank you very much for your reply. The system is actually a bit more complex than that. I didnt want to confuse the original question which was why I thought I will explain this bit later. The pump is pumping to four parallel circuits each with the same number of passages. The theoretical flow curves would be identical for all 4 sections. If I superimpose the actual pump curve on one of these sections they would actually intersect and the pump is quite a good match for the section. However since the flow is divided by four, I divided the pump flow rate by 4 and plotted it against the same pressure drop. ( I read somewhere that if two pumps are pumping in parallel, the resulting pump curve is obtained by adding the respective flows at the same pressure drop. I did the reverse). What is shown in the picture is actually 1/4 of the origianl pump curve. Am I allowed to do this?
I am assuming that when someone sized this pump originally they only considered one section. Also how do I know if the holes are running full even when the pump is pumping at its run-out point?
I considered static head loss, friction and form loss for the flow curves. The curve is actually very similar to the test curve in a lab test.

 

[bulkhandling]

The way you did not only confused other people but also confused yourself.

The right way to do is to take the pump as the major object. Calculate the pump with full flow, friction loss in suction piping, discharge pipe header and individual circuits. Show the full flow and calculated total dynamic head on the REAL curve. And then discuss about the hole and circuit pipe size.

 

[JohnGP]

I agree with bulkhandling. It certainly looks like the pump is running right out on its curve (which is not completely shown). The fact that you have flow at all means that there is an intersection with the System Curve, but as also suggested above, you need to plot the actual flow curve for the pump against the actual System Curve (which includes inlet piping, outlet piping and factored flow (x4) for single channel losses).

It would be useful to get a fix on the actual flow in the system, by installing a flow meter or diverting into a calibrated container, as a check on where the pump is actually operating. Checking power draw may get you there, but accuracy can be a bit off sometimes with this method.

If the flow in the passages is "up" as suggested by the sketch, and if there is no evidence of boiling, then you can be fairly certain that the passages are running full. If I interpret your statement correctly, the fact that the theoretical channel flow curve is similar to a test curve in the lab suggests to me that passages are running full.

I guess I am puzzled though as to why you want to increase hole sizes if the pump is seemingly already running somewhere near its maximum capability, unless the need is for more flow and you intend putting in a different pump.

 

[bulkhandling]

Jiks might not quite get the method for a system curve calc.

Assume Total flow 400 gpm get in to the pump. 100 gpm gets into each of the four circuits. Friction loss in suction piping for 400 gpm is 5 ft in fluid head, in discarge header 10 feet, in each circuit for 100 gpm 100 ft. Static head 0 ft (suction level same as discharge level). Total pump head for 400 gpm is 115 ft = 5+10+100. Since the four circuits are in parallel, friction loss will be balanced, only friction loss in one circuit needs to be added to total head.

From there you can plot your system curves on the pump curve and make decision if you need to change pipe size or not. You may pay attention to the velocity in the piping and velocity into the cylinder. Too high velocity
will speed up erosion and corrosion in your piping and cylinders and shorten their life.

 

[Jiks]

John, bulkhandling, thank you for the reply. I will start with John's question as to why we want to increase the size of the hole. I work for a university as a manufacturing engineer and we were doing an exercise for a local company to reduce the manufacturing cost of one of their key components. Along with many improvements like changing the design details, weld sizes etc, we also suggested to reduce the number of drilled holes by increasing the size. As these holes take about half an hour each to drill on a deep hole drilling machine, any reduction in the number of holes means a massive saving in time and cost. But they wanted to know how it would affect the cooling and flow in their existing arrangement. I drew the short straw and have been reading fluid mechanics ever since! My worst dreams came true when I tried to plot the existing system curve with the pump curve and found that they do not intersect! Or it seemed like that. I could not go back and tell this before I could convince myself that I am right.
The problem here is that I have only half the information. From what I understand, the pump they use in their lab is different from the pump they use in service. Moreover the test curve is only for one of the four sections. One section is connected to the test pump and the outlet is throttled to get the flow at different pressure drops.I have used the pump curve from their service pump rather than the test pump. This might also explain John's point that the pipes are running full as the test pump might be suited to that section. If hypothetically we match a system with a pump that do not have an intersection point what is actually happening in that line? Would it still run full?
If I understood bulkhandling right, I will find the friction losses at 100gpm in each parallel section and plot this against 400gpm. Therefore even though the outlet from the pump is 400gpm the pressure drop is only that of 100gpm in each section. For all other lines the friction losses will be at 400gpm. I might be wrong but would this not create the same effect as taking just one section and plotting the pressure drop at 100gpm and then dividing the pump curve by four? I understand that this might be the wrong way to approach the problem but my initial assumption that they do not intersect ( or the pump is working at its run out point) may still be belivable. I will re-do the calcs.
Thanks a lot for the input. It has got me thinking. I also have the Crane Technical paper 410 here and have been reading it.

 

[bulkhandling]

Do not use the way of two pumps in parallel to think about this application.

Concentrate on the pump, not the individual circuits. Like a city water supply, if a pump supply water to a thousand homes, you do not need to break the pump to one thousand parts. For your pump, the flow is 400gpm. Again the friction loss in suction pipe is 5 ft, discharge header 10 ft. In Circuit One, flow is 100 gpm, friction loss 100ft; Circuit Two, flow 100 gpm, friction loss 100 ft; Circuit Three, flow 100 gpm, friction loss 100 ft; Circuit Four, flow 100 gpm, friction loss 100 ft. Combined total head is 5+10+100=115ft.
It is your job to check what is the static head - discharge elevation take away suction elevation (You show 0 static head on your curves).

So, the operating point put on your pump curve (again, DO NOT modify or divide the curve!) is 115ft at 400gpm.

 

[JohnGP]

Right on! Certain parts of the circuit see full pump flow so they should be analysed that way. Logical procedure, particularly where you have dividing and combining flows, is to break the circuit into sections, as bulkhandling has done, and analyse headloss in each section.

The Crane paper is a useful reference and will provide much of the basics. You may need to dig into a standard Fluid Mechanics text to find discussion on pipe networks though, but the problem that you have here is much simplified where you have the flow dividing into "identical" paths.

It would still be useful if it were possible to obtain an actual flow reading for the pump to check against your analysis for the circuit with 25mm holes. A reasonable correlation between the two would provide some confidence in the analysis of a modified circuit with a smaller number of larger holes.

Cheers,
John
PS, if the outlet header is higher than the inlet header, then the holes should be running full.

 

[Jiks]

Thank you very much for the input. I am working on these now. Will let you know how it goes.
Jiks.

 

[Jiks]

Hi everyone,
Its been a while since my last post. I have now managed to map the whole 4 lines in parallel with fittings and pipe lenghts. What I have found is that the fittings and the length of pipe in each of these 4 lines is different meaning the pressure dorp will be different for each of the 4 parallel line. If therefore as bulkhandling suggested the total flow is 400gpm how do I find out how much flow goes through each line? Is there an equation for finding the equivalent pressure drop in a parallel circuit? Further I have been told that it is a closed loop circuit. Therefore I assume static head does not come into play. Thanks.

 

[davefitz]

For each channel , you should plot the pressure drop ( y-axis ) vs flowrate for 4 different assumed flowrates. The effects of a changing pressure drop and/or flow regime should be included in each assumed calculation. This shuld yield a total of four curves plotted on a single shet of paper.

All four circuits are flowing in parrallel, which means that at any instant they have the saem overall pressure drop. So, on the sheet with 4 curves developed above, assume a pressure drop, draw a horizontal line across the page, and note where this DP line intersects each of the 4 curves- this will yield 4 flowrates . The sum of these 4 flowrates sill equal the known total flowrate only at one distinct pressure drop ( assuming no flow instability ) - just keep improving your guess of total pressure drop until the sum equals the known total flowrate.

 

[Jiks]

Thanks Davefitz, working on it now. Jiks.