Age Effects On Performance Curve 1

[packdad]

In general, how would you expect age to affect the performance curve of a centrifugal pump? In order to perform bounding calculations which encompass future wear on a pump, I have typically assumed the curve to degrade a certain percent by head. In other words, for a given GPM value, I multiply the associated TDH value by a fixed percentage, and the curve shifts downward and "flattens out" somewhat. However, the curve could also be reduced by flow in a similar manner in order to simulate age degradation. That is, for a given TDH value, the GPM value could be reduced and the curve would shift to the left and "straighten up".

Are either of these methods reasonable simulations of normal pump wear over time?

 

[mucour]

Your method is quite interesting.

Honestly, I do not know of whether you can do this.
The only thing I know is that pump manufacturers are the ones that generate the pump curve based on running this pump using water at a given base temp,e.g. 20deg C, at different flow rates. Then the pump curve, H vs Q, is then generated.

Why don't you carry out similar test since, I presume, the pump is still in service by testing it for different flow rates starting from (zero) 0GPM to max. and checking whether the pump will deliver the corresponding pump head at the discharge nozzle as shown in the pump curve.

There are several questions that come to mind on reading your thread.

1. Why do you want to do this?
2. Has the pump been run to its "run-pout point"?
3. Is the pump not generating the max. head given in by the pump curve?

 

[electricpete]

To my understanding the primary effect of aging is increase in the internal recirculation.

It seems like this could reasonably be modeled by adding in parallel with the existing system a high-resistance parallel (bypass) path. The flow resistance of that parallel path decreases with age.

Construct the associated pump curve as follows:
For each dp: determine flow in imaginary bypass path.
The actual (aged pump) output flow for that dp will be the ideal pump output flow for that dp minus the bypass flow for that dp.

 

[packdad]

Most normal people (industries) would not need to do this!

However, I work in the nuclear industry, and when we evaluate system performance, we are required to take into account the potential future effects of performance degradation due to age on our safety-related pumps. Such a phenomenon will probably never be seen in the field due to preventive maintenance activities and in-service testing and trending, but we have to account for it in the design basis nonetheless.

With that said, I think I may have answered my own question through a little research, so I'll share: According to Karassik's "Engineer's Guide To Centrifugal Pumps", Pages 51-52, "the major potential cause of any reduction in capacity from the original condition...is the wear at the wearing rings, in other words the increased leakage through the increased clearances at these rings. The effect of an increased leakage on the pump head-capacity curve is...the net capacity of a pump at any given head is reduced by the increase in leakage." In other words, this says that effects of wear ring reduction over time are best accounted for by degrading the pump curve by flow.

Degrading a pump curve by flow is more physically realistic than by TDH. However, is degrading or adjusting a curve by flow necessarily more *conservative* than degrading a curve by TDH? Well, that depends entirely on the shape of the curve and where the pump normally wants to run. For a flat curve, or flat regions of a curve, it is probably less conservative to adjust by flow. For a steep curve, it may be more conservative to adjust by flow. It depends.

 

[butelja]

I think that ElectricPete's suggestion is saying the same thing as Karassik. If you model a bypass from discharge to suction with an orifice in it, you will capture the behavior of the system. In your system model, replace the pump with the pump/orifice combination. As the pump ages, the "orifice" enlarges. The $64 question is how big does the orifice get? You will need to work with the pump manufacturer on this.

Incidentally, this exact method can be used to accurately model the performance of a roots type positive displacement air blower, if the compressible flow behavior is properly modeled.

 

[BobPE]

Looking at your pump to determine future operation characteristics from wear is a good thing to do and not a lot of people look at it that way. Essentially you will lose head capacity of the pump from clearance wear. To model this just use the pump affinity laws to move the manufacturers curves down. Remember this is a non linear curve movement, so I think the way you did it wont work for you.

A lot of crazy things happen to the curves of older pumps like bumps in the curves at certain flows. This is typical of the conditions the pumps are in and how they fall off their manufacturers curves. Predicting the performance of a pump that is way of BEP becomes a field exercise but your use of the curve will still give you a good idea what to expect.

I hope I helped...

Bob

 

[packdad]

Yes, I think electricpete and Karassik agree on the cause (wear ring wear causing internal recirculation). However, the parallel-path (orifice) concept would, I think, tend to shift the curve down more at the low-flow end than the high-flow end, which is opposite from the actual effect according to Karassik. A parallel path with an orifice would pass more flow at a higher dP (and vice versa) which would tend to result in a larger difference between the "vendor" and "calculated" pump curves at the low-flow end rather than the high-flow end. Karassik's book suggests that wear ring wear will result in almost no effect over time at the low-flow (shutoff) end of the curve, with an increasingly larger drop in performance as flow is increased.

 

[electricpete]

I don't have your curve and may be missing your point. But there may perhaps be some insight gained by distinguishing between flow error for a given head and head error for a given flow.

Assume for the moment for simplicity that flow error as a function of dp is constant. i.e. constant bypass flow regardless of dp. Then the shape of the resulting corrected curve would have a much higher degradation when expressed as percent-hd for a given flow when at the high flow than at the low flow, simply due to the shape/slope of the curve.

Now return to the more accurate model where bypass flow increases with dp (flow error for given dp). Even though the flow error for a given dp is lower at low heads (high flow), it may show up as a higher error for a given flow expressed as percent-head... again simply due to the shape of the curve.

Once again I may not have understood your point correctly. If so I apologize.

 

[vanstoja]

On BobPE's observation of "crazy" bumps in the head-flow curves of older centrifugal pumps, be advised that the same kind of discontinuities, dips, bumps, etc. can be found in test curves of brand new pumps of sufficiently high specific speed. You may never find these discontinuities following "code" testing methods that only require test points at 10% intervals of the total flow range. Your chances of finding these discontinuities is much better if you use a 5% or less interval range and admonish your testers to report any preselected test points where conditions were too unstable for pump safety, eg. vibration amplitudes of 20 mils or velocities of 0.75 in./sec. You really want to know where these instability points are located and if necessary prohibit operations at or close to these flow instability points. This is how you pinpoint low flow rotating stalls and recirculation onsets in impeller and diffuser channels. There may also be instability flowates above best efficiency flow due to backflow onset, inter-channel flowstream shifting, etc. If you're going to build a number of pumps of this design you might test the first few with many points to cover the inevitable range of variations in instability locations after which you can revert to fewer test flows. If your head measurement pressure transducers are relatively far upstream and downstream of the pump impeller you may miss the flow channel action in the head-flow curve and would do better to monitor motor input power and current which will respond immediately to hydraulic component flow instabilities. Even more revealing are axial hydraulic thrust and radial hydraulic thrust curves if you can get them. They will sometimes respond violently to flow channel instabilities (with load excursions in the thousands of pounds across a flow range of less than 1% of total range) that barely ripple the remotely sensed head curve. You can construct flow instability "maps" that shows you precisely the flowrate fractions where you never want to run your pump at any speed for reasons of safety and longevity. Axial hydraulic thrust is most revealing when there are impeller balance holes directly connecting impeller flow channels to the back of the impeller where the thrust action occurs. If you retest the pumps in a plant with a different inlet piping configurations, you need to repeat this process so as to relocate the unstable flow ranges in the new facility.
They are generally connected with blade and vane flow incidence angles. Tests of many NACA airfoil profiles show that they mostly stall (lose lift) in the 12 to 18 degree incidence angle range. This is why near 60% rated flow has
been marked as "no man's land" in many papers dealing with centrifugal pump flow instabilities. Good hunting!

 

[electricpete]

Isn't the head vs flow curve restricted to monotonic decreasing? If not, then there must be at least one relative minimum (low-point) on the curve. How does the pump "know" which way to go if DP is increased from an initial condition at the relative minimum?

 

[electricpete]

Packdaddy - My second post of this thread ("I don't have your curve…&quot was somewhat rambling so I'd like to restate it as a question (maybe I'll learn something in the process).

If you look at the flow error for each dp (reading horizontally), does the highest flow error occur at high dp values (consistent with my proposed model) or low dp values (not consistent)?

 

[vgarzani]

May be too late in this thread, but the thrust curves mentioned by vanstoja were most interesting. Where do these come from? Force transducers in the mounting structure during performance testing? I have always thought that a step change improvement in condition assessment (either at the design phase or as installed) would be offered by knowledge or estimates of forces rather than just responses.

 

[vanstoja]

To vgarzani,
Axial hydraulic thrust curves are routinely obtained during engineering tests and early unit production tests of new designs for zero leakage, vertical, water-cooled pumps employing fluid film radial and thrust bearings. Two ways of measuring thrust are used. In the first, a special top closure assembly is installed having a threaded rod that screws into tapped threads in the top of the rotor shaft with rolling element bearings supporting the upper ends of the rod. Strain gages are placed on the rod to measure strain which is calibrated to force. The rotor has to be lifted off the thrust bearing (usually one-half the endplay clearance). The pump is run at normal cold water operating conditions over its entire flow range to provide a static thrust load vs flow plot. The second method uses a special balance piston installed as the top closure assembly and a measured pressure times piston area gives the force to balance the rotor shaft. This latter method enables dynamic axial thrust to be determined during startups and speed changes as well as steady state static thrust.
Radial hydraulic thrust measurement is a more difficult problem and is rarely done. Pressure transducers in the pump casing walls have been used but I don't know the detailed placements.
For conventional pumps driven by air cooled motors, the accessible pump or motor shaft either side of the coupling could probably be used for strain gage measurements of axial thrust and possibly also radial thrust. This is a worthwhile enterprise if you're dealing with relatively high specific speed pumps and need to know the pump hydraulic stability characteristics. It is useful to normalize hydraulic thrust, head and flow data to best efficiency flowrate and overplot the head-flow and thrust-flow curves. The relative slopes sometimes reveal unequal flow distributions in the hub to shroud plane of the impeller which can lead to premature runout of some flow streamtubes causing reverse flow through part of the impeller channel.

 

[SteveKW]

I have had experience where the internal diameters of an impeller wear causing an INCREASE in flow.

 

[rmasta]

I too have a case in which I need to determine flowrate given the head and amp readings for what looks to be a very worn pump. It makes sense to me that the curve gets its flows downgraded for a given head, much akin to a small bypass from the inlet to outlet. My question is will the horsepower/efficiency curves change?

 

[electricpete]

Steve - how do you explain that (increasing flow with wear). Can anyone explain why that would occur. I haven't heard of it.

 

[pumper]

electricpete,
More seal ring clearance will degrade head, look at any pump curve as the head drops the capacity increases. This however is not good for your pump, recirculation damages.

Pumper

 

[ccfowler]

I've seen some very old pumps in very rough condition providing essentially undiminished performance.

 

[ccfowler]

Sorry about the key stroke error and inadvertent submission!

Until reading this thread, I hadn't thought much about these observations in a long time. Most of these installations had relatively unfavorable suction and discharge piping configurations by modern standards (these were already very old pump installations many years ago), but the wear ring clearances were kept in fairly good order. I suspect that keeping leakage at the rings reasonable was probably responsible for keeping the pump performance in good order. The power consumption was not measured precisely, but motor amperage draws simply remained within very reasonable bounds.

In summary, I suspect that keeping wear ring clearances in good order may fairly well minimize performance degradation over an extended service life.