Excessive Pump Suction Nozzle Loads 1

[FlashPurp]

Hello,

I am working on stress on pump suction piping coming from a tank. I am seeing high moments and nozzle loads on my pump during hot and cold operation. I believe the issue is with how my header is growing and contracting during the 2 cases, so I was wondering the best way to restrain without transferring loads to the pump nozzles? Or would it be best to reroute the header to add flex and reduce movement?

Information:
Fluid: Water
Low Temp: 30 F; Hot Temp: 150 F
Design Pressure: 50 PSIG
Material: A312 TP304
Header Size: 18" SCH 10S
Branch Size: 12" SCH 10S
Nozzle Size: 8"

This is the layout I currently have it:


Movement during cold operation:

Movement during hot operation:

 

[XL83NL]

There are comparable threads here on ET that, albeit with a few different details (like linesize, temperature, material) deal with the same issue. The solution(s) can all be found in those topics. You'll find them easily using google or the site search engine. I suggest you read those, and do the math. Usually, the standard solution works well for 90% of situations, but you haven't provided many details. Let us know if you need advice on a specific question.
PS: at forehand I'd talk with the piping engineer and let him reroute this system as it doesn't seem to have (sufficient) inherent flex, so any time spent by a competent pipe stressed is useless.

 

[Snickster]

The following thread on ET discusses the issue thoroughly:

Excessive Pump Nozzle Load - Suggestions to reduce it? - Pipelines, Piping and Fluid Mechanics engineering

Hello All, I am new to pipe engineering, and I am trying to do pipe stress analysis to learn the software(Autopipe). I would like to get your opinion on routing the followng pipe based on your experience to reduce the nozzle load on the centrifugal pumps. I would also appreciate if you could...

www.eng-tips.com

 

[LittleInch]

OK, here's my scribble to give you an idea as it's clear that there is insufficient flexibility in this set up and the longer line from the tank is playing havoc.

 

[RVAmeche]

I'm a little surprised 150F is giving you such grief. Are you comparing to actual manufacturer allowables or assuming you have a problem?

Are you only getting a high MY based on the Z growth/contraction of the header?

 

[FlashPurp]

@Snickster and @XL83NL thank you for referring me to the attached threads.

@LittleInch thank you for the attached sketch, I'll check with my designers if that is a layout we could possibly try since we are pretty close to grade and have the discharge crossing over the suction header. Could I also use bends in the same elevation as long as I keep the 5D to 10D requirement of straight run?

@RVAmeche I am using API 610 loads and evaluating with Annex F, I am getting high moment and axial loads into my pumps. I believe its is due to how the header is behaving.

 

[LittleInch]

Your biggest issue as far as I can see is that large diameter long length from the tank is very rigid and is pushing and pulling the header around. Your little bends won't provide enough flexibility IMO and that longer length is dominating the issue

 

[RVAmeche]

Yes your header is very rigid as LI mentioned. Note that elevation changes to increase flexibility may impact NPSH calcs depending on how close things are.

If the simple header modifications don't get you there, you're probably best installing flex hoses.

 

[LittleInch]

You need to restrain the header in the X plane and add some flexibiltiy to that main inlet pipe

But in all these analyses, be careful how the program models the pump connections which are often taken as absolute anchors, so even the smallest temperature change manifests itself as huge forces and moments which in reality only need 1mm of movement to go away.

 

[FlashPurp]

@RVAmeche I see, thank you for your thoughts.

@LittleInch thank you for your insight, I'll try to restrain in the X plane without transferring loads/movement in the Z axis. As for the 1mm is that applied as the gap at the anchor point for the pump nozzle? To be more realistic, should this always be the case when analyzing pump nozzles instead of just using the typical anchors?

 

[Snickster]

You might want to consider using a rubber expansion joint at the tank nozzle if it is just a water pumping system. From your model printout it is obvious that the most problem comes from the expansion of the line from the tank to the header. Or as LI suggested install an expansion loop in that line and a limit stop at the header connection point that pushes the expansion back into the loop but allows the header to expand axially. Another solution would be to install anchors at pump nozzles like I have detailed in the thread I gave a link to. In fact the installation that these anchors was used was similar with suction from a tank going directly to pumps without much flexibility. Note that the installation I gave an example of in that referenced thread was a retrofit of an existing system so those anchors were there in operation for a very long time and worked. We replace all the piping to the pumps but kept the same arrangement of the anchors at the pump nozzle.

I have used these before:

Products | Mercer Rubber

Mercer Rubber Company Product Overview The Mercer Rubber Company was started in 1865 as a small factory on Mercer Street in Hamilton Square, New Jersey. The ...

mercer-rubber.com

 

[RVAmeche]

IMO pump nozzles are a little tricky. Especially for an API 610 pump that normally provides displacements and API 610 loads - if you make the anchor less rigid you don't get the quoted displacement because the piping forces push back on the nozzle. I also understand the sentiment that the load "isn't real" because gaskets will compress ever so lighlty more or the pump casing isn't 100% rigid but still. That seems like handwaving away the analysis at one of the key locations the analysis is for.

I've run anchors on pump inlet branches for hot services before like Snickster suggested but the loads came out too high (also in that thread snickster linked). And this was using a rigid between the pipe/elbow to the support baseplate, with support baseplate rigid anchor and pipe connection CNODE'd to rigid so flexes. If there's a "better" way to model this I would love know.

I know we have lots of different thoughts on this site and I've never seen a pump manufacturer or pipe stress forum come to a real consensus outside of "reduce the loads". I've read Snicksters old Peng articles and stuff and (to me) its frustrating some of the thoughts are "well the computers aren't good at this". Okay well this has been a Code requirement for decades, how about we reach a consensus on doing it via the software we have to use.

 

[LittleInch]

Pump nozzles are bad and normally tank nozzles are even worse.

I'm not really trying to dismiss the high forces and movements you can get at these connections, only to point out that the analysis programs can treat these connections as immovable and generate high forces and loads, whereas the reality is that nothing has really changed in centuries of pipe stress and design and pumps didn't break all the time either.

One way is to e.g. anchor the header, leave the connections as a free end and see what sort of movement you get or make the conneciton a spring connction somehow..


The 1mm I quoted was just to say if you allow some very small movement in the flange connection the forces can often reduce by 90% because we're talking such small lengths of pipe. I'm no stress analyst, so don't know what the options are, but I do know that the result of the analysis needs to be looked at carefully to not create an issue which doesn't exist in real life. Hell when you bolt up a flange connection you move the whole pipe by 2-3mm in many cases as the bolts get tensioned up. So maybe a flange conneciton which allows a small movement in the anchor settings to more realistically model reality?

 

[Snickster]

For stainless steel with coefficient of expansion is 0.00001 in/in F, If anchor is placed 18 inches from pump nozzle, the maximum expansion into the pump nozzle at 150 F (at 70 F installation temperature) will be 0.015 inches (0.4 mm) but if the pipe really expanded this much into the pump there would be no more load on the nozzle, therefore the expansion will be less than this. Honestly I don't think this will effect the alignment of a pump but can't say for sure. Although I have only done this solution on about 3 installations without any complaints after startup and operation, but I have seen it done by others on more installations.

That being said I agree with RV and would only apply this (nuclear) solution if all else attempts fail to provide a flexible enough piping without anchors. Whenever I did a pump piping system I would spend hours or days trying different ways to make piping flexible and/or installing stops and guides in the right places until the computer program stress calculated passed the allowables.

Once we had a bank of high pressure turbine driven gas compressors on and offshore platform that we could not get the loads within allowables. In the end we had to reroute the entire suction piping up and down the platform to get the loads correct.

On another location we had a water injection system piping at an oil field in Texas where the arrangement was very similar to the OP where suction was routed directly from tank to header then branch off to 3 injection pumps just like OP but suctions of pumps were facing in opposite direction (same direction and line coming from tank). Here I put a expansion loop in main line from tank like LI suggested as this was all out in an open field with alot of room to do it.

 

[Snickster]

Here are the drawings for the exp loop on suction of pumps that I referred to above:

 

[RVAmeche]

Yeah I understand the difficulty. For me, I'm going to try to get the nozzle loads down to the acceptable value since this is the tool we have. For what its worth, in the official Caesar II training course, they have an example on pump piping and recommend the more flexible layout that Peng/others don't prefer.

 

[KernOily]

Don't use API 610 allowable flange loads unless the pumps are API 610 (the post doesn't say what they are). API 610 loads are really high and non-API pumps won't withstand those.

Your pump manufacturer should have published allowables. Contact the manufacturer or their rep and ask for those. Then it is sometimes possible to exceed the published allowables if you can get the manufacturer to agree to that. Sometimes they will say yeah, OK to exceed by 30%, but we won't publish that.

There are a lot of tricks to ensure your CAESAR model is a reasonable representation of physical reality. First trick is you should consider modeling the stiffness of the pipe supports. That will loosen it up a lot. If that doesn't make the system pass, next you can model the stiffnesses of the suction and discharge flanges and include those as a flex anchor in the model. You can create a reasonably conservative model of the pump flanges using any FEA; even NozzlePro will work. If no FEA, you can use some stiffness for the pump flange that meets the horse sense criterion. Example, set the stiffness at 1E5 and think about that for a minute. Knowing what you know about the pump, would a force of 10,000 lb deflect the flange by one inch? If so, use that. If not, use something else. One thing is for sure, the default anchor stiffness in CAESAR of 1E9 lb/in is not even in the universe of reasonable. Then model the stiffness of the other ends of the piping system (tank nozzle, anchors at tie points, etc.).

I'm not surprised 150° F is causing problems. If the system has low inherent flexibility, and esp with SS pipe, it does not take much temperature to cause a LOT of problems.

 

[KernOily]

IMO pump nozzles are a little tricky. Especially for an API 610 pump that normally provides displacements and API 610 loads - if you make the anchor less rigid you don't get the quoted displacement because the piping forces push back on the nozzle. I also understand the sentiment that the load "isn't real" because gaskets will compress ever so lighlty more or the pump casing isn't 100% rigid but still. That seems like handwaving away the analysis at one of the key locations the analysis is for.

I've run anchors on pump inlet branches for hot services before like Snickster suggested but the loads came out too high (also in that thread snickster linked). And this was using a rigid between the pipe/elbow to the support baseplate, with support baseplate rigid anchor and pipe connection CNODE'd to rigid so flexes. If there's a "better" way to model this I would love know.

I know we have lots of different thoughts on this site and I've never seen a pump manufacturer or pipe stress forum come to a real consensus outside of "reduce the loads". I've read Snicksters old Peng articles and stuff and (to me) its frustrating some of the thoughts are "well the computers aren't good at this". Okay well this has been a Code requirement for decades, how about we reach a consensus on doing it via the software we have to use.

The problem is we have allowables published by the pump manufacturers. If there is a failure and subsequent court action and the failure can be shown to have been caused by the flexibility model not having followed a published allowable, you will be hung out to dry for the buzzards.

In my experience a lot of issues with models out of compliance are caused by folks not understanding the basic mechanics behind how the system is moving (or not moving). One frequent manifestation of this is how the boundary conditions are set up and the stiffnesses at the model's boundary conditions. I have observed this frequently with senior piping designers who have moved into stress work. They can run the software just fine, but they don't have the engineering training in solid mechanics and statics, so they aren't able to diagnose what is happening with regard to things such as combined stresses/Mohr's Circle/Tresca failure criteria and how those are included in the Code and the calculations, whether done by hand or by CAESAR.

Like you, I have been up against this a thousand times. I live and die by hot SS piping systems. The key to success with these is including in the model all the hidden flexibility you can possibly identify (supports, tank/vessel/hopper nozzles, foundations, equipment flanges, etc.), including expansion joints if the owner permits, then if all else fails, you just have to gag down re-routing the pipe. One of the best things I ever did was make the investment in NozzlePro. That has saved my bacon on innumerable cases. If that doesn't work, have a consultant in your back pocket with COSMOS or RISA-3D and that will definitely get you any stiffness you can dream up. The best thing about using FEA to generate stiffnesses is it gives you a defensible backstop for any stiffness.

It is so important to be able to distance oneself from the model and to step back and think about what is fundamentally happening from an engineering mechanics perspective. I learned that the hard way and that was some hard lessons learned. "All the world's a spring", as StressGuy used to say.

 

[RVAmeche]

Definitely agree about boundary conditions and yeah, we've had projects where the vendors asked for stress results, saw that the loads didn't meet their stated allowables and immediately used that as a get out of jail free card. NozzlePro is also great but my company is cheap and won't pay to keep up a license all year despite our requests lol.

For whats it worth, in the 3day Caesar course they also do an iterative example of a pump discharge to tank nozzle. Initial run is basically straight, then we add nozzle flexibility per WRC 297 for the tank, then add FEA nozzle stiffnesses for the tank, and then add an expansion loop in the pipe. Expansion stresses below-
Initial Design: 38,000 psi
WRC 297 Nozzle: 34,000 psi
FEA Nozzle: 23,000 psi
Expansion Loop: 14,600 psi

 

[XL83NL]

For whats it worth, in the 3day Caesar course they also do an iterative example of a pump discharge to tank nozzle. Initial run is basically straight, then we add nozzle flexibility per WRC 297 for the tank, then add FEA nozzle stiffnesses for the tank, and then add an expansion loop in the pipe. Expansion stresses below-
Initial Design: 38,000 psi
WRC 297 Nozzle: 34,000 psi
FEA Nozzle: 23,000 psi
Expansion Loop: 14,600 psi

Interesting comparison. It just shows how much conservatism there is in standard methods, and how nowadays advanced tools are capable of reducing the complexity (provided the engineer is competent)