Easiest Method to Estimate Losses Through Mechanical Seals?

I am puzzled by this. In my experience, mechanical seals in good condition have zero leakage. Occasionally a particle in the fluid can wedge into and traverse the seal faces. This may separate the faces for a (hopefully) short time, allowing some leakage. If the particle is softer than the seal parts, the seal should reseat. If harder, a sealing surface can be scratched, and a small leak will persist. If the leakage is large or unacceptable, the seal should be changed or repaired.

A couple weeks ago I had a brief refrigerant pump leak that reclosed on its own.

Tornado - Mechanical seals all leak. But not enough for you to see. Double seals with flushing stuffing boxes are common in chemical processing service where the liquid is actually hazardous in small quantities to keep this minute leakage out of the environment.

drift - The API standards are a good place to start. Been a long time since I spent time there. I suspect they will give maximum allowable leakage rates. But these will be a sanity check for your first trial calculations.

gussosa - I'm looking at the second edition of Stepanoff, Chapter 10 "Leakage, Disk Friction and Mechanical Losses", pg. 185, about 4 pages in from the beginning of the chapter. First off the flow through the seal is laminar, not turbulent. That simplifies things. But still what Stepanoff says there is that basically he doesn't have a good answer either mathematically based on test data. But remember that he wrote this work over 60 years ago when most of us in the pump industry left it to the mechanical seal manufacturers worry about such things and apply their tribal knowledge and black art to the selection of seals.

I have this suspicion that the ideal case of a constant clearance seal interface of known pressure, viscosity conditions and spring loading per unit area is solvable with a closed form one dimensional Navier-Stokes solution if one assumes a linear relationship of length to viscosity due to friction heating. But there are a lot of idealized assumptions there having to do with the actual geometry across the entire seal face. Add additional dimensions or non linear functions to the fluid characteristics and you get into differential equations that require numerical solutions on the computer.

The McGraw-Hill Pump Handbook (Karassik, Krutsch, Frazer and Messina from Worthington Pump) has some of this solution in section 2.2.3 Centrifugal Pump Mechanical Seals by James P. Netzel of John Crane Packing Co. The second edition, 1986, shows this on pg. 2.135 through 2.128.

You may have to go search through ASME transactions in an engineering university library to find more detailed exploration of this flow problem and the actual dynamic behavior of mechanical shaft seals in operation. Does Uruguay have a major University that has that resource. Or perhaps in Buenos Aires? A huge amount of research in the world of tribology has been published in the decades since Stepanoff did his work and computers became as commonplace as slide rules.

The other problem is that testing and analysis of seal performance has probably been confined primarily to the seal manufacturers who keep what they learn as trade secrets. Testing methods that involve measuring extremely low flow rates as well as the difficulty of actually measuring the precise dimensions of seal clearance in a miniature liquid filled dynamic environment would be a challenge. If seal company researcher worked out these problems and got good data it might be a long time before he could publish it.

Ed Weldon

I am prepared to accept that even "perfect" mechanical seals can have extraordinarily small leakages, such as occasional molecules. But if the interseal fluid reservoir between two mechanical seals remains at the same volume for several months, then the seal leakage is indistinguishable from zero. (Yeah, yeah, yeah, it might not really be exactly zero, but so fuqqing what?) My environment is ammonia refrigeration, in which harmful concentrations are readily smellable; the rationale would change for odorless pollutants.

There are sligh loses since the surfaces will not work properly without a "lubricant" to separate them.Here are a few sources for more information about.

gussosa,

if you're referring to the losses in say, barrier fluid of the auxiliary system, or maybe water for external flushing of the seal, for usual engineering practice you can simply estimate the losses based on the experience. Otherwise, try checking API 682 and API 610 standards for more details.

Leaks through the shaft seal aren't the main problem in centrifugal pumps. The biggest issue is with return of flow through the space between the walls of the rotor and the body. However, the Stepanoff equations are meant to be used in any of those situations. I finally found some lecture notes in Scribd, in Spanish, where they explain the application of the method, and they refer to some paper by Stepanoff from 1932. I will ask it through my university librarian. Maybe it is in Science Direct or Springer Link.