Estimate rate of reverse flow through centrifugal pump when it has stopped 2

[SNORGY]

Hi folks.

Is there an accepted methodology to estimate the reverse flow through a centrifugal pump assuming the discharge check valve has failed simultaneously with the pump not in operation? I have a situation where I am being asked to size a relief valve on the suction side piping on the basis of check valve failure on the discharge side. Per API 521, the reverse flow through two identical check valves in series (which is one of the things I am looking at) would (could) be based on 10% of the forward flow Cv through one of the two valves. I also have a discharge pressure control valve in the system which has a wide-open Cv that I can take credit for. I end up with an equivalent Cv given by Cvt = [(Cv1^2*Cv2^2*Cv3^2)/(Cv1^2*Cv2^2 + Cv1^2*Cv3^2 + Cv2^2*Cv3^2)]^(1/2); sadly, in my case, the resulting Cvt = 99, and thus the requirement for an enormous PSV results on the suction piping. In the ideal case, what I'd like to do is estimate an equivalent reverse flow Cv4 for the pump itself and derive an expression similar in form to this one (which I have done) in the hopes that the pump Cv4 is small enough to pull the PSV down into the "D" thru "G" orifice range.

If I do this, I end up with Cvt = [(Cv1^2*Cv2^2*Cv3^2*Cv4^2)/(Cv1^2*Cv2^2*Cv3^2 + Cv1^2*Cv2^2Cv4^2 + Cv1^2*Cv3^2*Cv4^2 + Cv2^2*Cv3^2*Cv4^2)]^(1/2); then all I need is Cv4.

I might be able to get an estimate of the worst case reverse flow through the stopped (not running) pump from the vendor, or else by rationalizing it in terms of an energy balance at various points on the pump curve and quantify the wasted energy as proportional to fluid that "slips" (i.e., is not delivered) for each given head. But...it feels a bit hokey to me. I just wondered if this situation has been encountered by someone before and what the accepted practice might be. It might end up requiring a LOPA or SIL study to avoid the PSV.

Thanks in advance.

 

[Compositepro]

The resistance of a centrifugal pump to reverse flow is not much that you can count on. I would basically treat it as a pipe elbow the size of the outlet nozzle. I do not think that is what you want to hear. Sorry.

 

[EdStainless]

The pump won't be stopped, it will 'windmilling' in reverse. This is why there is so little resistance. There are many systems where the controllers detect the voltage generated by this and prevent starting the motor. Depending on the speed in reverse it is a very good way to shear off a shaft.

= = = = = = = = = = = = = = = = = = = =
P.E. Metallurgy, Plymouth Tube

 

[Artisi]

Why is there a need for a relief valve in the inlet pipework, is this to protect the source from a high discharge head in the case of the failed NRV?
I agree with Compositepro, think you are being too academic, the "pump" will only appear as an additional friction loss in the system under reverse flow conditions, so simply treat it as a 90 bend in the pipeline friction line loss calculation - and what would the head loss thru'the pump as a percentage of the overall calculation anyway?


It is a capital mistake to theorise before one has data. Insensibly one begins to twist facts to suit theories, instead of theories to suit facts. (Sherlock Holmes - A Scandal in Bohemia.)

 

[JJPellin]

This has been studied. There was a paper presented at the International Pump Users Symposium that addressed this for a particular pump configuration. If you search the proceeding, I believe there were other relevant papers.

http://turbolab.tamu.edu/proc/pumpproc/P27/Hazards of Reverse Pump Rotation.pdf

Johnny Pellin

 

[georgeverghese]

Is the pressure ratio of the max normal downstream pressure to the upstream design pressure less than 1.5?

Use the criteria in the Shell DEP on underpressure / overpressure on the subject of check valve failure if you have access to this. You get considerable credit for using Class I check valves, if your plant HSE culture is familiar with how to look after these.

Also, cannott tell if your formula for summed up Cv is correct - I would do this by trial and error by guessing an overall flow, computing the interspace pressure between any 2 devices and working backwards till you reach the upstream side of the first check valve. The correct flow would yield a pressure upstream of the first check valve = to the relieving pressure.

 

[SNORGY]

JJPellin,

Thanks for that, I had actually come across that piece earlier this afternoon.

Artisi,

In this situation we have a client with an existing booster pump and 150# ANSI piping system feeding the inlet to a LACT skid. There are about a dozen manual block valves in the line and LACT unit between the booster pump and the downstream shipping pump. The downstream shipping pump is a multistage Reda Schlumberger barrel pump that takes the pressure from about 400 kPag to 8900 kPag, i.e., from ANSI 150# to ANSI 600#. The downstream system is a pipeline network with about 8000 kPag worth of line pack. Check valve failure on the discharge of the shipping pump, when it is not running, has been identified as a credible source of overpressure in the LACT system piping. Depending on the Code Of Construction, and a few other things, there are a few safeguarding strategies that might be acceptable. The LACT has been built to B31.3, so Code requires either OPP in the form of relief or HIPPS supported by a LOPA study. We could also just swap design codes and put it under the umbrella of CSA and get by with OPP in the form of pressure limiting devices such as PITs and ESDVs. The bigger concern, however, is that the site is unmanned so if a large relief valve is called for then so is a large containment volume in order to provide adequate operator response time. At the end of the day, if what I am seeing is true, we need a LOPA to kill the PSV idea.

Georgeverghese,

I just started with the classical formula for combining two components in series to get an equivalent Cv, then used *that* as an effective Cv to combine with another, and then used *that* to combine with a fourth. It appears to be correct; I have checked the results using a bunch of random Cv values and they agree whether I do the calculation stepwise or in combination.

You are all probably right, it's getting a bit academic. When this first came up in conversation during an internal design review yesterday, my comment was that we should probably seriously challenge our assumption / conclusion that this is a legitimate, non-multiple-jeopardy OPP scenario. The team consensus was yes, so the action item, as action items frequently do, became mine.

 

[MikeHalloran]

So, what happens if the suction line PSV is 'too small'?
Won't it reduce the reverse flow like a not so badly failed check valve would?

Nah, SNORGY, you'd have thought of that.
You must be trying to analyze some Doomsday scenario.
Not my area of expertise.





Mike Halloran
Pembroke Pines, FL, USA

 

[SNORGY]

MikeHalloran,

The thought is that the resistance to flow from the high pressure side to the low pressure side will give rise to a reverse flow rate that is a function of the pressure differential, so that if the same flow rate is not coincident with the relief valve capacity at its set point plus the allowable overpressure, the suction piping will pressurize to levels well in excess of flange rating until the equilibrium condition (with respect to flow) is reached.

 

[georgeverghese]

To sustain a reverse relief scenario, a primary precondition is that there should be some vapor space pressure on the shipping pump discharge side to keep the system pressurised at 8900kpag during the reverse flow. If it were all liquid, even a tiny reverse flow of liquid would have completely dissipated the 8900kPag. So do you have such a vapor space somewhere downstream at 8900kpag?

 

[Artisi]

I realise this probably doesn't get to your full answer, but the question I see is, what is the head loss thru'the pump?

For me I like to keep things as simple as possible so, why not do a series of calcs. using various numbers for the pump headloss, starting at zero with stepped increments to some preconceived number, then look at the results to see if or what the results are- maybe the "pump" loss is insignificsnt in the overall scheme and can be neglected or an assumption made to suit your model.

It is a capital mistake to theorise before one has data. Insensibly one begins to twist facts to suit theories, instead of theories to suit facts. (Sherlock Holmes - A Scandal in Bohemia.)

 

[SNORGY]

georgeverghese,

The sustained pressurization on the pipeline side arises from the fact that other pumps from other sites feed the pipeline. However, I agree that the calculation needs to be done of how much mass transferred out of the pipeline back through the pump will reduce the line pack to the point where the issue goes away. I think you are completely dead-on with that. That is also one of my action items.

Artisi,

I will try your approach. In effect, it would be the same as choosing the relief valve size I want and backing out the Cv4 value I would need in order for it to work.

So...this is how I typically spend my weekends...