13 replies to this topic

[jprocess]

Dear All,
Standard procedure for flare KOD sizing is according to API 521.
The amount of liquid in inlet stream to flare kod is usually unknown.API 521 did not mention which amount of liquid should be considered for sizing.
This problem is especially true for cases which liquid stream leaving flare KOD is NNF.
So what approach should be followed by designer?
Thanks in advance.
Cheers.

[Art Montemayor]

jprocess:

First of all, API 521 is NOT a "standard". It is a RECOMMENDED PRACTICE. You can take it or leave it.

Secondly, please do us all a favor an be specific about what you assert. You fail to specify that what you refer to is found in API 521, paragraph 5.4.2 - DRUMS AND SEALS. This helps everyone to zero in on what you are talking about and come quickly to the subject at hand.

Thirdly, it really helps everyone a lot if you identify your acronyms. Everyone is not immersed in the same subject and we can't afford the brain cells required to memorize every conceivable acronym.

You ask "what approach should be followed by designer?". I believe it is pretty obvious to all that the equation is specific and that is what is used (although there is an errata on p.30 of the fourth edition). This equation is just like the famous Brown-Souders equation (which we have battered about for years on this forum) - there is no flow rate involved. The principle variables are the liquid and vapor densities. API says it all: "Sizing a knockout drum is generally a trial-and-error process".

I hope you are working with the right equation, as corrected by API.

[jprocess]

Dear Art,
Thanks a lot for your reply.
Before speaking about this topic I want to ask you about the new edition of your workbook about gas blanketing.If you remember you stated earlier that you will upload the new edition of your workbook in the future.Do you have any news about it for us?
But about this topic:
I think that 2007 edition of API 521 is not a recommended practice.Now it is a standard.Am I wrong?
You are right about section 5.4.2 of API 521-1997 edition about the terminal velocity equation.
But my question refer to page 66,equation 34!
We have not the value of liquid,so we can not calculate the value of liquid cross sectional area!
Page 66 of API 521 contains an example with known amount of liquid.It does not speak about this concern.
So I repeat my question that what approach the designer should follow?
Thanks in advance.
Cheers.

[pleckner]

Let's clear up one thing first. It is true that the latest edition of API 521 is now also an ISO Standard. The RP (Recommended Practice) designation for the API document has been removed. However, and this is something I just picked up at the last DIERS Users Group meeting, there is no difference between a "standard" and a "recommended practice". They both carry the same weight in this world, which is nothing more than a guide to use to achieve some form of "good engineering practice". Point of note, I also found out that when API 2000 was first issued, it included "Recommended Practice" in the title but a subsequent update dropped the "Recommended Practice" and just called it API Standard 2000. The value and intention of the document did not change.

For the K/O drum, the diameter is set by vapor velocity and K/O requirements as Art pointed out. The overall height can be set to include a reasonable liquid retention time...if there was liquid to retain. In the absence of any "real" liquid to accumulate, just assume a reasonable retention time and move on. "Resonable" basically means what you need for downstream processing to handle. You want to put enough of a "bump" in the line to allow downstream equipment to function properly. I've seen one guide say 10 to 30 minutes based on a 10 to 30 minute relief. Perhaps you can have a two-phase relief and you need to accomate that amount of liquid for 10-30 minutes. You may assume a 10% carryover for 10-30 minutes as a basis.

The bottom line, without having a real basis, make one up that seems resaonable without breaking the bank when designing the vessel.

[jprocess]

Dear Phil,
Thanks a lot for your valuable comments.
Let us review a part of your comments:
"For the K/O drum, the diameter is set by vapor velocity and K/O requirements as Art pointed out. The overall height can be set to include a reasonable liquid retention time...if there was liquid to retain. In the absence of any "real" liquid to accumulate, just assume a reasonable retention time and move on."
1.I agree with you and Art about diameter.
2.As I stated earlier,the liquid flowrate is unknown.The reasonable retention time value is only applicable when we have the value of flowrate.Without knowing it, the value of retention time is useless.(we multiply liquid volumetric flowrate by retention time to obtain the volume and then we can proceed according to page 66 of API 521)
3.What do you mean by 10% carryover?
Can we consider 10% of vapor flowrate as an estimation and obtain the liquid volume with 10-30 minutes retention time?
Your valuable comments are appreciated.

[joerd]

@jprocess
Re. your item 2, if you don't have any liquid, you may design for a minimum mechanical requirement for your level instrumentation. You will want to measure if there is any liquid in the vessel, and how much, even though you don't expect it. So, you have to place nozzles on the shell, and use a minimum size level gauge. Consult with your mechanical and instrumentation engineers what the constraints are, and you may be surprised what "dead volume" you end up with in the bottom of your KO drum.

[pleckner]

@j:

I think you got my meaning pretty good. You can assume any reasonable % of liquid you want in the vapor but I like what @joerd has to recommend instead. If you can't see any way for liquid carryover or a two phase flow, then go with @joerd's recommendation; I like his idea better than mine.

[JoeWong]

@j:

I think you got my meaning pretty good. You can assume any reasonable % of liquid you want in the vapor but I like what @joerd has to recommend instead. If you can't see any way for liquid carryover or a two phase flow, then go with @joerd's recommendation; I like his idea better than mine.

jprocess,

Hardly see any Flare KOD has NO liquid design case. Not sure what kind of plant your are working on. If you have any blowdown system, hot-wet stream meeting cool-dry stream at KOD will generate some liquid.

Secondly if you hot-wet stream, severe condensation may occur during winter.

Third, please check if you have any potential liquid relieve into flare system e.g. tube rupture...

JoeWong

[jprocess]

@j:

I think you got my meaning pretty good. You can assume any reasonable % of liquid you want in the vapor but I like what @joerd has to recommend instead. If you can't see any way for liquid carryover or a two phase flow, then go with @joerd's recommendation; I like his idea better than mine.

jprocess,

Hardly see any Flare KOD has NO liquid design case. Not sure what kind of plant your are working on. If you have any blowdown system, hot-wet stream meeting cool-dry stream at KOD will generate some liquid.

Secondly if you hot-wet stream, severe condensation may occur during winter.

Third, please check if you have any potential liquid relieve into flare system e.g. tube rupture...

JoeWong

Dear JoeWong,
I agree with you that "hardly see any flare KOD has NO liquid design case" but my question is that which value should be considered for liquid flowrate? I beleive it is not easy to stimate.I did not find any comments about it in related standards.
My special case have no liquid because the company which is responsible for basic engineering have considerd the overhead vapor from a de-butaniser tower as governing case for low pressure dry flare network and flare KOD sizing.
His assumption seems to be logical and refers to cooling water failure case for de-butaniser condenser(my guess).GPSA chapter 5 states:
"Loss of cooling water may occur on an area-wide or plantwide basis. Affected are fractionating columns and other equipment utilizing water cooling. Cooling water failure is often the governing case in sizing flare systems."
But I do not know that it is applicable for flare network piping and back pressure checking using isothermal flow method? As I remember for that approach we start from flare stack outlet and do backward calculation and include the flows from "ALL" sources and not one of them which is the governing!
By the way I found a value of 267.38 ft3 of liquid as an approximation in page 62 of "Flare Gas Systems Pocket Handbook"
Warm Regards.

[pleckner]

If a plant wide cooling water failure can cause multiple reliefs, then yes, you would then consider all vessels that can relieve simultaneously. Most people think of this in terms of fire but a plant wide cooling water failure, or even a power failure, might have the same affect.

For the header sizing during multiple relief, you want to use the required flow rate from all vessels that will relieve simultaneously. You DO NOT need to use the PSV stamped capacity in this case. Note that I said "header". For the individual tail pipe, you would still use the stamped capacity of that individual PSV.

Don't just use that volume of liquid you found on page 62 of "Flare Gas Systems Pocket Handbook". You must first undestand if it really pertains to your situation. Otherwise, it is only one guess of many that can be made. I would still go with the recommendation of @joerd above.

[jprocess]

Dear Phil,
I describe what I get from joerd approach to see if I understand it right?
His method is based on level points that are LLL,LL,NL,HL,HHL.
By setting these values and calculating the drum diameter(D) we have the value of HLL/D and then we can extract the value of (Partial liquid volume/Total vessel volume) from books like pressure vessel handbook.But knowing this value will not help us because we have not the vessel length so we have not the vessel volume and we can not calculate the liquid volume!
Secondly about your statement:
"For the header sizing during multiple relief, you want to use the required flow rate from all vessels that will relieve simultaneously. You DO NOT need to use the PSV stamped capacity in this case. Note that I said "header". For the individual tail pipe, you would still use the stamped capacity of that individual PSV."
You mean for back pressure calculation and tailpipe sizing using isothermal flow equations,we should include rated flow of PSVs?
Thanks in advance.

[JoeWong]

joerd advice will be a good approach if relieve fluid has no liquid condensed.

Phil has clearly indicated cooling water failure may results multiple relief.

I would advise that total power failure may lead to cooling water pump stop and cooling water supply stop...



JoeWong

[pleckner]

@jp:

Re-read @joerd's post above. You will need a bottom nozzle. This will probably be a minimum of 6" above the bottom tangent line (or the weld point) of the vessel. Then choose a minimum sized level gage. As @joerd said, someone from your instrument department can choose one for you. The length of this level gage will the maximum level of "fake" liquid you would consider ever having in the K/O drum at any time. From here, you can set the vessel overall height.

For backpressure calc in the header ONLY, you need to only consider the total required flow of all PSVs that will relieve simultaneously. The required flow is just that, the maximum relieving rate from the controlling scenario, it is not the stamped capacity of the PSVs. Then when you pick up the tail pipe that ties into the header, this is where you use the stamped capacity of that particular PSV to finish the pressure drop calculation. And yes, for ease of calculation, you can use the Isothermal gas equation.

@joe:

I mentioned plant wide power failure in my previous post.

[jprocess]

Here is the SHELL company DEP's comment about API 521 method:
"In API 521 the vessel sizing method is based on the settling of droplets, assuming a horizontal and uniform vapour flow.
However, in reality the vapour flow is far from uniform, especially if a small feed nozzle with no feed inlet device is used.
Also, in reality, the vapour flow contains a vertical component, in particular in the vicinity of the vapour outlet.Due to these two effects, the API 521 method results in vessels that are too small."
Any hint?
Thanks in advance.