Pump mechanical seal starts leaking with water temperature drop

[jimhokie]

The pump is a horizonally mounted, closed impeeler, single stage, end suction dual volute, top centerline discharge centrifugal pump, 1800 RPM, 12" suction and discharge.

The service is a chilled water pump that circulates cold water from an AC plant. It initially runs with no mechanical seal leakage, but once the chiller is turned on and the water drops below about 50°F, the seal starts to leak. When the chiller is turned off and the water rises back to ambient (~70°F), the leakage remains. Once the pump is left idle over night, it starts up with no leakage the next morning until the chillers are turned on and the water drops below 50°, then the leakage returns and repeats the pattern. Cycling the pump off/on after the water warms back up doesn't stop the leakage...only sitting overnight does. We’re puzzled as to why the leakage remains when the chiller is turned off and the water warms up above 50°--if it’s thermal pipe strain, it seems this should be relieved once the pipe warms back up to ambient. Perhaps there’s some hysteresis in the pipe strain that relieves itself over night due to structural vibrations or other longer term effect.

I suspect there is likely some static strain preexisting as a result of bad piping alignment, which is then exacerbated by the relatively small temperature change when the chillers are turned on, enough to distort the casing, causing shaft strain and the seal leakage.

At this point, it is impractical to break the piping connections to verify piping alignment. Any other ideas for confirming if static and/or thermal piping strain could be causing the leakage, or any thoughts on what else the problem could be? Both the seal and pump manufacturers are stumped.

 

[EnergyProfessional]

Can you tell what brand the pump is?
If you have Taco/Armstrong pumps you can expect leaks.

 

[jimhokie]

The leaking seal is a Crane type 3710 split mechanical seal. It is designed for a temperature range of 32°-122°F. Is there any explanation for how the seal leakage could be caused by the temperature variations within the seal itself? That's why I suspect a pump/piping interaction, but the experts here are discounting that and focusing on the seal.

 

[JJPellin]

Fully split mechanical seals are genearlly not considered as zero leakage seals. It is normal and expected for them to drip a bit. How much leakage is occuring?

Johnny Pellin

 

[jimhokie]

We specify an allowed leakage of up to 30 drops per minute to account for that. But in our experience with these seals on several different design pumps, they are generally visibly leak free but sometimes on an initial start on a new seal, there is some small leakage that goes away as the seal faces get run in together.

On this pump, there's zero leakage when the pump is first started for the day, then it first begins to drip when the water drops below about 50°F, then increases to a steady stream with further cooling to about 40°. When the chillers are turned off and the water warms up above 50°, the steady stream leakage continues, even with cycling the pump on and off. Then the next morning after the pump has been off over night, the seal is miraculously leak tight when the pump is started until the water drops below about 50°. This cycle has been repeatable for several days.

So the main questions are:
1) Why does the seal start to leak only at lower water temperatures (but within the seal's design range)?
2) If it leaks because of a temperature effect, why doesn't it stop leaking when the water temperture rises again?
3) What can explain why it seals again after sitting idle overnight?

 

[1gibson]

Is your seal drain port pointed straight down (6 o' clock) or is it on the side/top? Is it possible that it leaks constantly, and just takes some time to accumulate enough liquid for an external leak to become apparent? Then that leak somehow siphons or the rest of the fluid evaporates overnight?

Initial thought, you've basically added a lot of details and possible variables to the situation that "it runs for a little while before there is visible leakage."

Start the pump an hour early and leave the chiller off, see if it takes the same amount of time to see external leakage, or if it is really somehow dependent upon the chillers operating.

Just to mention it, is it possible that the leakage is actually condensation?

 

[LittleInch]

What is your coupling arrangment with the motor? Can the motor shaft / motor heat up and actually expand and cause issues which are actually independant of the cooling water on/off? Try a test like 1 gibson says and see if the start action really is the cold water or is it juust a function of time after you start?

It is beginning to sound as thought the chilled water is actually not the cause, but something else so perhaps look at what heats up reagardless of the pump temperature that then cools down overnight.

My motto: Learn something new every day

Also: There's usually a good reason why everyone does it that way

 

[ianhoag]

My first thought was that the shaft cools and shrinks more than the shaft sleeve, and takes longer than the water to warm up. Perhaps the friction between the seal faces keeps the shaft sleeve warm. There should be an o-ring between the shaft and shaft sleeve to prevent leaking, but it could be missing, broken, or scratched.

 

[jimhokie]

1gibson: Yes, the seal drain port is at the 6 o'clock position, so any leakage past the seal is immediately drained away. And during initial pump testing, it was run for many hours without the chiller on with no leakage. It wasn't until the chiller started making cold water that the seal began leaking. Condensation is a good question...probably doubtful, but something we need to rule out that I hadn't thought of.

LittleInch: The pump is close-coupled...the impeller is mounted directly on the motor shaft extention, with only a short portion of the shaft exposed to the cold water, and as mentioned above to 1gibson, the pump was run for extended periods previously without the chiller, during which it did not leak. So it has seen many cycles of the motor heating and cooling off without any leakage until pumping cold water. I can't get away from the piping system having some hysteresis in the thermal load it is putting on the casing that takes overnight to relieve itself somehow (gradual relaxation due to external structural vibrations maybe?), but nobody else thinks this is the case.

ianhoag: That's a good thought that I'll look into more. I can't imagine that any differential expansion/contraction over such a relatively small temperature variation can be an issue, but i'll try to vet that out to be sure.

Thanks for all the thoughts so far!

 

[NegativeZero]

How long has this been doing this? and has anyone tried to make adjustments? IMO they have to rebuild the seal. if that much water is coming out, the seal is shot. Any dirt in the water most likely compromised the seal at this point. If they cant tighten it anymore they most likely tighten to much while the water was warm on start up or someone tried later on. I would get a rebuild kit make sure the shaft is free of scratch's, chips etc.. and closely follow the install directions. when you have good cold water make your final adjustments.

 

[jimhokie]

NegativeZero: The pump has been run on and off for several weeks, but the chiller just came on a couple weeks ago, and that's when the leakage started. We don't think the seal is shot, otherwise it wouldn't run leak free each morning before the chiller is turned on. They are about to rebuild the seal to see if that fixes it, even though no one can explain the temperature effect. Maybe it was a poor initial seal installation. I will pass your thoughts on. Thanks!

 

[dcasto]

I suspect a difference in materials that have different coefficients of expansion. The inner is contracting faster than the outer and the gap is to much for the seal to make up.

 

[NegativeZero]

can you show me where you get those temp specs for the 3710 split? because I cant find those spreads you posted. Unless I'm looking at the wrong Crane type 3710 split mechanical seal.

 

[jimhokie]

It is a Crane type 3710, size 3.625", part number NSP-1000044-1. The pump specification stipulated that the seal must operate over a temperature range of 32-120°F. What temperature range did you find for this seal, and where? I'd be shocked to find out the pump manufacturer selected a seal that didn't meet the temperature requirement...but if so, that could explain the problem! This is a new design pump, but the qualification testing did not test the full temperature range. We are in a holding pattern right now awaiting a break in the system testing that will allow the seal to be replaced to rule out a bad seal.

 

[1gibson]

I wonder how much help an IR or laser temp gun would be? Getting a good reading will probably be difficult.

 

[spciesla]

Getting a "good" temperature measurement from the seal with an IR camera or gun will be tough because of the shiny surface / low emissivity. However, putting a piece of electrical or duct tape on the surface (safely) should help out.

 

[NegativeZero]

Spec sheet that's all it says. Their other seals will give a min/max.. I think you need to call Crane

http://www.johncrane.com/~/media/J/Johncrane_com/Files/Products/Specification Sheets/S-3710.ashx

■ Temperature:
Up to 121°C/250°F
■ Pressure:
Vacuum to 13.7 bar g/200 psig
■ Speed:
3600 rpm up to 62mm/2.500"
1800 rpm for seal sizes between 62mm/2.500"
and 135mm/5.250"
900 rpm for seal sizes greater than 135mm/5.250"

that was going to be my next suggestion. taking temp readings on the seal. laser gun will work.

 

[jimhokie]

Help me out here on where you guys are going with measuring the seal temperature. We do have a water temperature measurement, and I would think for all practical purposes, all the wetted seal parts are essentially at the water temperature (except for the seal faces themselves), and the external surfaces wouldn't be too much different given the thermal conductivity of the metal. What is the thought for what to be looking for with the external temperature measurement?

NegativeZero: It is interesting that the Crane spec sheet doesn't identify a low temperature limit. Both Crane and the pump manufacturer have been consulted on this problem, and both are at a loss to explain how a seal that begins leaking when the water temperature drops, and does not stop leaking when the water temperature is raised again, but miraculously seals again the next day after not running over night.

I found this article yesterday that gave me an idea to explain how the seal "fixes" itself over night:

http://en.wikipedia.org/wiki/O-ring

Skip down to the section about the Challenger disaster, that discusses Richard Feynman's demonstration where he drops an o-ring into ice water and then shows its loss of pliability. The key part that struck me is: "The material of the failed O-ring was FKM which was specified by the shuttle motor contractor, Morton-Thiokol. FKM is not a good material for cold temperature applications. When an O-ring is cooled below its Tg (glass transition temperature), it loses its elasticity and becomes brittle. More importantly, when an O-ring is cooled near, but not beyond, its Tg, the cold O-ring, once compressed, will take longer than normal to return to its original shape." Per the seal parts list, 2 o-rings and a gasket are fluorocarbon material (FKM). Does it sound plausible that the chilled water is cooling the fluorocarbon o-rings and/or gasket material to where it loses enough elasticity to begin leaking, and then “will take longer than normal to return to its original shape”, thus not resealing itself for several hours (overnight) later? Although it can't really "return to its original shape" as it is contrained within the seal parts, the sealing force my gradually restore itself over an extended period.

Maybe I'm clutching at straws here, but there has to be a physical explanation for something that is happening in such a repeatable fashion. It has stumped the seal and pump manufacturers as well as other engineers here, and the powers that be here have discounted my theory that thermal pipe strain distorting the pump casing that for some reason takes overnight to relieve itself explains the seal works again the next day, until the chiller is turned on again. Their best plan is to replace a $4000 seal when there is an available repair window to see if that fixes it.

 

[NegativeZero]

the rubber used in this seal is well within the temperature ranges. If you have glycol in the water that could mess with Fluorocarbon type of o-ring. Do you have glycol? Also if they ran the seal dry that will could damage the seal. (It don't take much time minute or less) or the seal came damaged from the factory. Its not like I never seen that. Good luck figuring it out, I will be curious what you find. I assume you have the installation instructions. keep us updated.

http://www.johncrane.com/~/media/J/Johncrane_com/Files/Products/Instruction Manuals/I-3710.ashx