Estimating Separator Carryover 4

[ChEMatt]

I've searched around but it's escaped me. What correlations exist for the estimation of carryover from a two or three phase gas/liquid separator?

I'm rating some spare 3ph vertical separators we have in our yard. The inlet nozzle is definitely too far up the side of the vessel, based on the first pass sizing using a commercial process simulator (and some common sense when looking at it in the field). What I'm looking to do is demonstrate the carryover potential based on the inlet flow, composition, properties, etc.

I have no information on internals, so for all I know there's a downcomer or a baffle in there that diverts the flow. For now, I'm assuming nothing and we're going with worst case scenario.

I could lower the flow until the nozzle height is "appropriate" and the rest would be attributed to surge capacity I suppose, but in this instance we have many of these vessels. I could put another one downstream. Or two. Whatever we need.


Thanks for your help!

-matt

 

[don1980]

Based on superficial velocity of the gas that's flowing upward through in the vessel you can use Stoke's Law to determine the droplet size which will settle out and the size that will be carried off with the gas. In real world separators, however, there are other very important factors that determine how much is actually carried over.

1) Inlet conditioning: Is there a vane or a mechanical device on the inlet nozzle (inside the separator) that's actively coalescing droplets from the inlet stream?
2) Splash: When designing a vessel to remove droplets, it obviously critical to ensure that the design doesn't actively create more droplets. This is an often overlooked detail. Without knowing it, the inlet design can easily create of more droplets. That happens if the inlet velocity is high and it's pointed downward at the liquid level in the drum, causing splash. This can also happen if the inlet stream impinges on the opposite wall at a high velocity. In the latter case a high percentage of those droplets will have an upward trajectory - the opposite of what you want! A tangential inlet is ideal for preventing the formation of spray, while at the same time causing small droplets to coalesce into large ones which will settle out.
3) Outlet conditioning: Is there a vane or other type of droplet coalescing device near the outlet? If so, and if it's properly designed for the vapor velocity inside the vessel, then it can greatly increase the effectiveness of the separator. These demister devices work by causing small droplets to coalesce into bigger ones. The small droplet impinge on the demister (vane or wire mesh) and merger together into large droplets.

So, without knowing what's inside the vessel, a calculation isn't much better than a guess. If you do know what's inside the vessel, then you can determine whether or not the inlet design is likely to cause spray (formation of additional droplets). If the answer is "no" then your theoretical calculation of settling is a reasonably valid one. Next consider what percentage of those droplets will be coalesced by the outlet device, if there is one. If it's a commercially purchased demister, then refer to the vendor's manual to determine it's effectiveness.

 

[ChEMatt]

Yeah, I got to thinking that no matter what the disengagement space, if a droplet is going up, it's going up.

I'm hoping that there's some sort of diverter or baffle. These are very simple vessels, cheaply made, non-code (built in TX).

I appreciate the suggestion of something on the outlet to coalesce any droplets. The vapor stream will head to gas sales; compression isn't required due to the presence of a low pressure gathering system in the area. On the other hand, I could throw in a properly designed coalescing filter to get the last bit out.

Thank you for the reply!

edit: I'm probably better served by really looking at what the droplet size would be, as accurately as possible.

 

[georgeverghese]

Design guidelines tell me you need a min of 0.5*vessel D( or 1m whichever is greater) as gap between the top of the feed device and the bottom of the gas exit demister if you have a primitive feed device ( half open pipe or simple impact plate). If you have a better feed device ( a well designed slotted distributor or similar), you could manage with 0.3m or feed pipe pipe d as this gap. Also enable a straight piping length of 10d upstream of this separator, without bends or other restrictions.

In a 3 phase application, you have the bigger challenge of enabling good L1-L2 phase separation.

 

[EmmanuelTop]

You are saying these are existing separators. I'd try to pull a sample somewhere downstream of the vapor outlet nozzle, ideally from a low point. If you have another KO vessel downstream, you can directly measure buildup of the liquid phase.

I played a lot with separators (mostly KO drums and wiremesh demister-types) in various plants, and I found that the actual amount of liquid carryover can go anywhere between <0.1% of inlet liquid volume to over 2% of inlet liquid volume. The factors were: inlet device type, actual gas velocity, the amount of dispersion (that depends quite much from liquid/vapor phase ratio and upstream piping configuration), outlet separation device, and liquid/gas phase properties. There are no general correlations. If something is really really important, building a CFD model can give some very good estimates.


Dejan IVANOVIC
Process Engineer, MSChE

 

[ChEMatt]

Dejan,

That would be nice, to take a sample, but I'm looking at evaluating these for a *future* application given the design flows/pressures/etc. So if I put this vessel into X situation, how well will it perform? That's my problem. All the design equations are for "base" type of design where you should get little to no carryover.

Hysys includes a correlation from a paper, something called "Proseparator". I'm trying to find that paper.

 

[apetri]

for vapor-liquid and vapor-liquid-liquid separations I have evaluated many correlations (working with Excel or Open Office and Prode Properties for thermodynamics) and I can confirm Dejan point of view, CFD can be an option but complex,
the models available in literature can include several parameters, you can redefine parameters but lose the benefits of predictive models...
particle-size distribution can be evaluated in different ways,

https://en.wikipedia.org/wiki/Particle-size_distribution

simple correlations are given in textbooks such as Gas-Liquid by Stewart , Arnold etc.
carry ove can be evaluated in different ways, for example assuming that all droplets smaller than a user-specified critical droplet size are carried over (a option available in Aspen)
or (as I do) with correlations based on statistical distribution (more accurate).

 

[ChEMatt]

I'm trying to obtain the paper behind the Hysys correlation; I have not yet been able to obtain that.

Can you describe in more detail your statistical distribution method? I'm familiar with the effect on droplet size on separator sizing, but what I don't have/know is how to estimate the total volume of droplets as opposed to what would be a part of the non-droplet volume of liquid coming into the separator, if that makes sense.

 

[apetri]

I do not understand your question, if you think that a software gives the results you wish to obtain then use that software what is the problem ?
as said by others (see previous posts) I do not consider generic procedures accurate in a large range of cases except for specific applications where one can fit statistical correlations and obtain accurate results...
With tools as Excel, Matlab, Mathcad etc. and a thermodynamic library (I have Prode but there are others) you can define you own procedure, I mean you can fit parameters, there are several papers discussing these topics.

 

[ChEMatt]

apetri,

The software does NOT give me the results I am looking for. Software tells you what the proper size separator is. Software does not tell you what happens when you try to use an undersized separator in a given service.

The original problem that I stated in the first post was that I have a separator, existing. If I try to run it in a certain situation, what will happen? What kind of separation will I achieve? What % carryover will I see? How much oil will I get in my water, and water in my oil if there's not enough residence time?

If something exists that will allow me to rate a real, physical, separator in other services to determine performance, carryover, etc. then that's what I'm intersted in. I have Excel and a commercial process simulator. I don't have Mathcad or Matlab nor do I know how to use those products.

You mention "several papers" discussing these topics. Would you kindly direct me to some of these papers? That would be quite helpful!

 

[apetri]

probably the the most common method is based on Souders-Brown equation (force balance applied on a droplet) and one can adopt regression analysis to predict optimal parameters of models (based on Souders-Brown equation),
google with these keywords,
however there are many generic correlations, in addition to Stewart and Arnold book mentioned in my previous post you may consider Gas/Liquids Separators—Quantifying Separation Performance—Part 1-3 author Mark Bothamley and similar works...

 

[ChEMatt]

Thank you for those suggestions. I'll follow up!