Small volume of propane released suddenly into flare 2

[SNORGY]

I am looking at a propane condenser (air cooler) with a relief valve discharging to a flare system. I interpret API-521 to suggest that, with free drainage to the accumulator, I do not need to size for fire based on bundle bare tube area; in fact, I don't need to put the relief valve in at all. However, the process packager has installed one and piped it to flare. The orifice size is "P". If one is to assume the full relief valve capacity in terms of mass flux rate at the flare tip, the GLR from the stack exceeds regulatory allowance by a factor of 3.0, and, correspondingly, the stack is 100 feet too short.

The condenser actual volume is about 4% of the flare system actual volume. Understanding that the propane at condenser pressure has a mass (vapor-based) approximately equal to the mass of air and purge gas in the flare system, and that the relieving event duration is less than 2 seconds, and that the stack is about 350 feet away, do I need to assume that full mass flux rate of pure propane exiting the flare tip, or is there some accepted way of quantifying the effects of mixing and displacement that might contribute towards the reduction of heating value and GLR? If such a calculation methodology exists, can someone point me towards it?

I have this picture in my mind of a valve going fully closed to fully open for 2 seconds at one end of 350 feet of 18"-D pipe and waiting for this instantaneous plume of gargantuan proportions frying everything within 100 yards of it...and I am struggling with accepting it.

 

[EmmanuelTop]

Is Propane relief to flare a credible scenario?

Dejan IVANOVIC
Process Engineer, MSChE

 

[SNORGY]

It is, now that the packager has installed said relief valve.

 

[EmmanuelTop]

If the PSV on the accumulator vessel is of sufficient size to handle the relief from any of the identified scenarios, can't you simply insert a spade upstream of the PSV on the air condenser?

Apologize if I asked a dumb question, perhaps I didn't figure out what is the real concern. A sketch would help tremendously.


Dejan IVANOVIC
Process Engineer, MSChE

 

[SNORGY]

Thanks EmmanuelTop.

I'll scan and upload a hand-drawn sketch after my morning coffee here.

I have an idea about how to try to predict the fluid / mixed gas behaviour in the downstream piping system. I don't have any sophisticated software at my disposal here like AFT or HYSYS, just a TI-89 with a pencil and paper and some books. It won't be elegant...

 

[EmmanuelTop]

I will gladly simulate the properties of mixed stream for you, if you provide me with individual stream properties and conditions/flows. It takes a few minutes only and Aspen is by my side.

When you upload the sketch, please clarify if your concern is the relief from the air condenser only (and why the relief from the downstream receiver would not be a concern, if it discharges the same fluid), or if the concern is relief of Propane to the flare system in general?

Dejan IVANOVIC
Process Engineer, MSChE

 

[SNORGY]

Hi EmmanuelTop.

Thank you for your offer to help with this, I certainly don't intend that you should do that for me, especially on a Sunday; I was just looking more for suggestions from folks as to whether or not they had seen / done something like this before, and if so, was there an accepted methodology for it. API-521 does allow for dynamic modeling of relief events, and it does suggest alternatives to fire case relief valves on air coolers. No matter what this calculation ends up showing, I have options available to me to arrive at a safe design.

Here is my sketch...and a star for your post simply for *offering* to help - but please don't feel like you need to do this, especially on a Sunday.

 

[SNORGY]

EmmanuelTop,

I am only concerned about the outdoor relief valve, PSV-1, as there is a building and geographic separation between the condenser and the accumulator.

Inlet temperature to condenser is estimated at 53 C; each fan is 40 hp; condenser design duty is 4.85 MMBTUH; summer case air 35 C. I have assumed 3 rows of 60 1"-D tubes each and a face area of 12'-0" x 46'-0". I have also assumed that the last pass at least is sloped.

Note that if the rules of API-521 are followed for heat input and the condenser is assumed 100% full (100% wetted area), not even a "P" is even remotely large enough; so the packager has made some kind of rationalization towards sizing that valve - which is not in dispute - I am just trying to evaluate he consequences of what I assume is a correctly sized relief valve on the connected system external to packager's scope.

My concern is that if I am to assume the full PSV capacity of pure propane vapour at the flare tip, the GLR is such that the stack is almost 100 feet too short. However, I rationalize (right or wrong) that the entire contents of the condenser are put into the downstream system before any propane even makes it to the stack; further, if the gas already occupying the system volume is displaced, in whole or in part, as this happens, then to assume the total mass flux of propane at the stack tip is equal to that at the PSV outlet seems very pessimistic to me.

 

[EmmanuelTop]

Thank you for the detailed sketch. This is what I think:

1) Based on API 521 paragraph 5.15.7.3 and comparing it with your drawing, credit for self-draining can definitely be taken in this case. This supports the initial assumption that the Propane air condenser PSV can be removed or permanently isolated.

2) Sizing of the PSV on the accumulator vessel should be reviewed carefully. It seems strange that, for example, loss of power (fin-fan motor failure) does not result in higher relief loads, unless there are other protective functions (e.g. SIS) which prevent this from happening (e.g. system shutdown upon power trip).

Once the relief rate is (re)confirmed, the rest of the calculations is pretty much straightforward.
For points (2) and (5), (6), (7) and (8) from the sketch there is a sophisticated tool which can be used for accurate calculations - Flaretot. I have used it myself and I was quite impressed with its capabilities and level of calculation details. The link for download is pasted below - you'll get 30 days trial (either directly after installation or after contacting the software supplier with the software installation code, I forgot). It will provide you with very accurate stream analysis, hydraulic and radiation calculations.

http://www.flaretot.com/

There are many "tricks" which API 520 allows you to use in order to obtain more reasonable/accurate GLR, depending on site location/conditions. For example, solar radiation may not be needed to be accounted for if the plant (and relief scenarios) meet certain criteria (i.e. the expected frequency of release is negligible). The same applies for maximum allowable GLR in the ground area around the flare stack - depending how much "manned" is that area in real life, and whether any human intervention is required under given relief scenario. It would surprise many people how much "fat" is built into Flare design with regards to radiation criteria.

Once you document all these facts and figures, you can clearly observe whether you have a real problem, or not.


Dejan IVANOVIC
Process Engineer, MSChE

 

[SNORGY]

Thanks again for this, EmmanuelTop.

I will post later today or tomorrow what transpired after completion of my calculation.

Have a good Sunday!

 

[MortenA]

In case the relief scenario ir right i would consider doing a dynamic simulation of the whole flare system taking all the pipes and the flare KO drum into account. Its not going to be cheap but there is a lot of "hold-up" in the flare system that the usual design methods normally dont capture. Of course i cant promise you that it will smoothed the peak sufficiently but there is always hope

 

[SNORGY]

Thanks, MortenA.

I agree completely. This all started with me asking for that very thing (dynamic simulation), but our process engineering resources were tied up with other things. I worked most of the day today at home taking my best flyer at doing the calculations manually. I got frustrated by 9:30 pm and went to bed when the results I was getting indicated high GLR no matter what approach I tried. I woke up at 2:30 am with a "Eureka Moment" and I think I have found a rationalization that is plausible and addresses the GLR concerns. Your reference to "hold-up", as you have termed it, is consistent with my revelation. In essence, I can make a lot of problems go away, theoretically, by taking more credit than I have been so far with the HPFKOD volume. I will go to work early tomorrow (later today) and I think I can put this issue to rest - something I cannot seem to do for myself.

 

[EmmanuelTop]

Depending on the original plant design, there may be several fire zones which would effectively limit the number of equipment exposed to fire, in all subsequent flare load evaluation calculations. There is no real basis to consider the entire facility under fire simultaneously, unless the whole plant occupies very small - and I mean very small - footprint.

You have mentioned that air condenser and Propane accumulator are in different plant (geographic) areas. If there is a pipe rack underneath the air condenser, API 521 (paragraph 5.15.7.1) gives you direct recommendation to exclude the condenser from any fire affected area.

Dejan IVANOVIC
Process Engineer, MSChE

 

[SNORGY]

EmmanuelTop,

OK I closed the loop on this today. What I did was in complete accordance with the plan I had illustrated in my sketch. To establish the instantaneous relief rates, I normalized the entire piping system to be reflected by a single diameter pipe, to which I applied Oliphant and used Weymouth as a sanity check. I assumed fully blended mixtures of propane, natural gas and diluted air to recalculate the different calorific values of the flared contents. Since the total mass of propane was fixed, the calorific value of the mixture - as well as its density, flowing pressure and, by association, volumetric flow rate - could be reduced by increasing the amount of available volume in the system, which I achieved by lowering the liquid level in the knockout drum until the net effect was to arrive at an acceptable GLR. On that basis, I was able to demonstrate that the system was acceptable without further modification.

I then suggested to CSO the isolation block valve between the condenser and the accumulator, and re-evaluate the fire case risk with a view towards taking the PSV out altogether.

Thanks for your help, EmmanuelTop and MortenA.

 

[EmmanuelTop]

Then you have arrived to the optimum solution - the wolf's belly is full and none of the sheep is missing.

Dejan IVANOVIC
Process Engineer, MSChE

 

[GHartmann]

While API may relevant here, of more relevance is ASME. ASME DOES require a relief valve in this scenario.

Then utilizing normal relief valve "case" analysis your first relief valve should be sized for "blocked flow". The need for this relief may depend if there is another relief valve upstream at the source of your propane vapor. You MUST have at least one relief valve in this system.

The second relief valve you have indicated is for the fire case. This is a simple calculation of the surface area of the accumulator tank (including dished ends) and will specify your capacity.

You now have your 2 maximum flows for your flare system. You need no more information. The maximum flow sets the flare sizing

 

[EmmanuelTop]

PSV on the air condenser is not mandatory for blocked outlet. Credit can be and should be taken for the locked-open isolation valve downstream of the fin-fans. If credit cannot be taken in such cases, we would end up with PSV's rated for full flow, installed between each two manual isolation valves.

Dejan IVANOVIC
Process Engineer, MSChE

 

[SNORGY]

Thanks, GHartmann.

Present but not shown in my sketch is the blocked flow PSV on the compressor and oil separator upstream of the condenser. But - to your credit - you made me look...

The system design was by a third party. My mandate was simply to lokok at what would happen if their PSV listed and "...had a "P"...", so to speak.