15 replies to this topic

[Nirav]

Hello,

What should be the best approach for calculating relief load of a distillation column in case of reflux failure?

It is little difficult to predict the heat and mass balance during "reflux failure". So, in normal or conservative manner, we calculate the total vapors coming into the column at the "relieving" conditions. We deduct the "normal" vapors going out from reflux drum from above calculated rate (If we want to take credit of condenser, we take "normal" vapors in column top outlet).

So, for example,

Relief load = Vf (feed) + Vr (reboiler) - Vo (vapors out from reflux drum)

However, I came across situations wherein a process licensor optimized the vapor flow during reflux failure based on simulation of the column approaching "reflux rate=0 kg/hr".

Is it possible to optimize the vapor rate during relieving conditions based on simulations?
What specifications should we provide on simulations and what approach should be followed to find the relief rate through simulations?

Does anyone have any suggestion / idea ?

Warm regards,

[rxnarang]

I would be very careful about using simulation to predict the relief rate. The standard industry methodology of envelope heat and mass balance ( Unbalanced heat load calculation) should be used. The simulator can be used to provide properties at relieving conditions. Ignore the rebolier pinch which may occur at relieving conditions.

In my book, the concept of " optimizing" relieving rate is fraught with danger. A conservative approach, but based on type of failure ( e.g. partial or total electrical failure) may be used.

Dynamic simulators claim to give a realistic (lower?) relieving load, but I do not know of any company which mandates this approach. Maybe somebaody else has experience in this.

Regards

[Nirav]

Rajiv,

Thank you for your reply.

Even, I think it is not advisilbe to optimize when there is no specific calculation for it. And simulation can not be relied upon in "abnormal" situations.

That is why we use the approach what i described in my original post.

I am interested to know what others are following in case of "reflux failure" to determine the relieving rate. What do you mean by "reboiler pinch"?

Regards,

[pleckner]

I would agree that trying to optimize this type of failure is not the best approach. But I would disagree when not taking into account reboiler pinch. (Nirav, this is when the boiling point of the liquid goes up because of the higher pressure in the tower at relieving conditions. If you were to be using steam as the heating medium, the LMTD between the steam and the boiling liquid can become very small and thus heat transfer is limited).

I see no reason to put in a larger PSV just for the sake of doing so when there are ligitimate and safe ways available to reduce its size. I am probably one of the most conservative people around when it comes to PSV calculations and analysis but I won't cost my customer more money then necessary. And, putting in too big of a PSV just invites chattering problems later on.

For Nirav's specific problem, if this is a licensed process and the licensor uses their normal means of providing pressure relief for their process, this should be OK; they know their system better than anyone. Saying this, I'm not against questioning them, as a matter of fact the owner is obligated to do so if he has a problem with the sizing basis.

Also, using a simulator in this case does not necessarily mean the system is optimized. I think Nirav might just be making an assumption the licensor is optimizing the relieving rate but in reality is just using a simulator to obtain a relieving rate. Also, I can't see why using a simulator for an abnormal condition is not valid. Today's simulators are very sophisticated and can handle a host of situations. If you (being the process engineer) are not comfortable with using a simulator for this situarion, then don't use one. I will go with "better safe than sorry" but if you are well versed in setting up the simulation and you choose the correct model, your answers will be as valid as using any other method. If the model you choose is wrong, then even using the simulator for a simple flash calculation to get relieving properties will give you incorrect results. (By the way, "model" in this case means the VLE and thermal correlations).

Now, to address Nirav's approach as he presents it in his original post, this is fine with me with one exception. I would take credit for the condenser in operation (but only if this is not the cause of the loss of reflux to begin with) but I would recalculate the amount of condensed vapors based on the relieving conditions, assuming fouled conditions and turndown cooling medium flow rate. For the vapor properties, these should be the same as what one is using for the PSV sizing calculation, not the "normal" vapor. You may find you can't condense as much of this vapor as you could the "normal" vapor.

Summary: Perhaps I will just be in disagreement with "rxnarang" on some issues.

[Nirav]

Phil,

Thank you for your explanations. We used the analysis for reboiler performance at relieving conditions. But I was just not aware about the terminology "reboiler pinch".

Now, coming specific to my problem, I was not trying to question licensor for the relief rate. I was just trying to learn as how they arrived at lower flow rate. Otherwise, when relief rate is not provided by licensor, we use the conventional approach (better safe than sorry!!).

When asked 'informally' to licensor, I got answer that they used simulations to calculate the vapor flow during reflux failure. I have original simulations (HYSYS) provided by licensor for normal case. Therefore, I am very much confident about the thermodynamic property package. But I find it difficult to 'converge' column without reflux. And that is what I am searching for.

For analyzing reboiler performance, as per my understanding, there are two things to be kept in mind.

[1] Lower heat transfer because of reboiler pinch :
Q = UxAxLMTD will give lower Q (compared to normal) due to lower LMTD at relieving conditions.

[2] Lower mass heat of vaporization (latent heat) : At higher pressure, latent heat reduces which will boost the vapor generation as per... Q = m (vapor mass flow) x Latent heat

So, in first impression, lower heat transfer rate could reduce vapor mass flow . But it might be compensated by lower heat of vaporization which is lower at relieving conditions. Sometimes, it may give "higher" flow rate of vapor than normal even with lower heat transfer rate. Therefore, it comes to good engineering judgment or analysis, if engineer is trying to go in details to find out required relieving flow rate.

I'm also in agreement with Phil for considerations for changes in condensing flow at column top, in case if we want to take credit of it.

Ultimately, my curiosity is still unanswered. i.e. Specification (or trick or analysis) for column to get it converged "without reflux" or with very very low rate (almost negligible) of reflux.

Warm regards,

[pleckner]

I deleted my previous response because I had a revelation!

How is this for a radical thought; loss of reflux is NOT a credible relieving scenario!

After more thought and talking with a colleague who has more distillation experience than he cares to remember, loss of reflux is a pretty common problem in distillation operation and does not result in over pressurization. Look at my attached Excel sketch, crude but effective. The reflux is within the confines of the boundary of the system. Loss of reflux by itself should not affect the ability of the column to remove heat nor does it affect the column’s ability to input heat. Therefore, there is really no reason to expect column pressurization from loss of reflux in-of-by-itself.

Loss of reflux just affects column overhead purity. The overhead will be richer in heavies, will increase in temperature and all that actually makes it easier to condense. The problem will be if the condenser fills up with condensate because you can’t rid of it quickly enough in the distillate. Then this becomes a loss of cooling scenario. If other things begin to happen, such as pressure control valves close because more condensing will tend to drop column pressure, then this too is a different scenario from loss of reflux.

Bottom line, loss of reflux is not really a credible over pressure scenario.

Any one with other (or conflicting) thoughts?

[rxnarang]

We simulated this scenario ( loss of relux only) in a simulator, and were able to converge the column by tweaking the reflux rate to be almost negligible. The trays above feed tray were running dry, so essentially the vapor load at the feed tray is the required "relieving rate".

As Phil pointed out, this is not a credible scenario, because, the system is still balanced ( both material and energy). In Phil's sketch this is apparent by drawing an envelope around the system, and one can see that the reflux is within the envelope. All that happens is the mass transfer in the rectifying section gets affected.

Now, losing the condenser due to flooding is a very real situation. In this case the column will get unbalanced in energy. The excess energy will then have to be relieved. We are still working on this scenario in the simulator, but I have a basic question.

The steady state simulator will balance the heat and mass in the envelope. What we have here is an unbalanced situation. The excess energy from the reboiler is used to evaporate the column contents. Is a steady state simulator man enough to do this job?

When we do this manually, we introduce a pseudo-stream; which is equal to the amount of vaporisation caused by the excess energy. Presumabaly the simulator should be able to do this automatically, but I doubt this.

Will report back with some more observations, when we have something new.

Rregards

[pleckner]

As I pointed out in my last post (and we all agree) loss of cooling is very much a credible scenario and I will add that it is this scenario that typically controls the PSV size on a column.

rxnarang: Your simulation of the loss of reflux case should be the same as for loss of cooling except now the column is simulated at relief pressure. The relieving rate is all the vapor generated.

There have been a number of articles written on the use of dynamic simulation to obtain the "optimum" relieving rate. I don't have them available right now so I can't give references. We didn't have good dynamic simulators years ago so we had to trust the good ole steady state simulator. This should be adequate. Heck, even Nirav's initial idea is adequate (except don't remove the vapor from the reflux drum)!

[Nirav]

Phil & Rajiv,

Thank you for your replies.

I'm in agreement with both of that column can not be overpressurized just because of loss of reflux. But over a period of time, "after effects" may cause overpressure.

I referred back to API-521 (March 1997), Clause 3.6 : Cooling or Reflux failure. I would like to reproduce a couple of its sentences.

3.6.1 The required relieving rate is determined by a heat and material balance on the system at the relieving pressure. In a distillation system, the rate may require calculation with or without reflux."
3.6.6 In many cases, failure of the reflux that results, for example, from pump shutdown or valve closure will cause flooding of condenser, which is equivalent to total loss of coolant with the capacity established as in 3.6.2(total condensing) and 3.6.3(partial condensing).

So, in brief, loss of reflux may result in loss of overhead condenser due to liquid overfill which will cause overpressure. That is the reason why we seldom take any credit of overhead condenser and why we have been reporting relieving rate as SAME for "reflux" & "Cooling water" failure for a column. API-521 tells same.

Phil said in his last post that we should not take credit of vapors out reflux drum for calculating relief rate. I have some concern over this.
[1] If it is “total” condenser, there is no vapor out of drum theoretically and there’s no question about taking credit.
[2] If it is “partial” condenser, we can take credit of “normal” vapor rate out of reflux drum.

Even API-521 (3.6.2 & 3.6.3 on Pg.9) suggests the same. The important aspect is to re-calculate vapor rate at “relieving conditions” instead of normal operating.

I'm trying to study some more aspect of column simulation on HYSYS over above. If I find anything, i will write it. I'm also looking for Rajiv's findings.

Thanking you again,

[pleckner]

Interesting that API RP521, paragraphs 3.6.3 says nothing about taking credit for "normal" vapor rate from the reflux drum. It only states, "The relief requirement is the difference between the incoming and outgoing vapor rate at relieving conditions." Nor does it suggest how to determine what this "outgoing" vapor rate is; Nirav, (1) you need to re-read this paragraph more closely, (2) you need to read the posts and responses more carefully.

The scenario under discussion is a loss of reflux that leads to a "loss of cooling" as a result of liquid backing up into the condenser. I stick with my original statement, we should not be taking any credit for vapor from the reflux drum for this particular sceario, other scenarios perhaps but not this one. Whether it is a full or partial condenser, if it fills with liquid, there won't be anything condensing.

And by the way, API RPs are recommended practices, not code. Anyone who has been reading my posts in the past will tell you I've disagreed with API RPs before as long as I am more conservative than they are and if I believe they are technically correct or not. And one more thing on APR RP 521, it is still in the Fourth Edition, 1997 (with an errata published in 1999). It high time it is updated and perhaps they will clarify these particular sections as well (or perhaps not).

[Nirav]

Yes, I got little diverted from "reflux failure" to "condenser failure" while refering to certain aspects. Thank you for pointing out.

I know, in previous other discussions, you have said that API is RP (recommended practice) and not "standard". So, it's upto engineer to evaluate the situation properly and apply it.

So, I would like to clarify a bit on my opinion about taking credit or not taking credit of normal vapor out of system in case of "partial condenser".

[1] If we think of failure of condenser as a consequence of "reflux failure", we should not take credit of normal vapor out, since the reason of failure is flooding of it. So, it won't condense any of the vapors. I would correct my equation by removing "normal vapors out from reflux drum"

[2] If we think of failure of condenser due to its own reason (e.g. cooling water fail), we can take credit of normal vapors out of reflux drum. Isn't it logical? I would like opinion on it.

Thank you for various input.

[pleckner]

Nirav, addressing your point number 2, "If we think of failure of condenser due to its own reason (e.g. cooling water fail), we can take credit of normal vapors out of reflux drum. Isn't it logical?"

Again, API RP521 does not address taking credit for "normal" vapors anywhere. But I think I might see where you are coming from. If you are assuming that there is a vent line off of the reflux drum going "somewhere", and this vent line is wide opened to this "place", then I guess one can consider taking credit for something; what this something is would be purely a hydraulic balance (pipe, valves, etc.) at the relieving conditions. But in the interest of time (money) and conservatism, you better be sure your hydraulic calculaitons are spot on and you better be sure you can really reduce the size of your relief valve enough to justify the added cost in time it will take to really make sure your hydraulic calculations are correct.

For me, it isn't worth the time or the possiblity I might not get the hydraulics correct and my conservatism goes right out the window. Better safe than sorry!

[Nirav]

I just wanted to put an example for what I was referring

Consider one amine regenerator column. It is stripping off H2S from "rich" amine to get back "lean" amine.
It will have H2S gas always going out of system from 'reflux drum' during normal operation. This would be thorough a pressure control valve over reflux drum to maintain pressure of column. The hydraulics are well in place since we have to specify control valve. There is no extra time needed to calculate anything extra. In such case, we can take credit of "normal" (not max possible) vapors out of the system, can't we? Because we do not consider 'failure' of the pressure control valve (because it would create "double" jeopardy which are not related to each other).

Just wanted to add about API's quote about 'partial condensing'
3.6.3 : The relief requirement is the difference between the incoming and outgoing vapor rate at relieving contions.

Can it be interpreted what I am saying?

[pleckner]

Nirav:

Yes I know that quote because I added it in one of my previous repsonses. And in repeating what I said before, API RP521 gives us no clue as to how to determine what this "outgoing" vapor rate is other than it is at relieving conditions. And how about the fact that as the column pressure increases, this very same pressure control valve will open to relieve pressure. This means that even more flow can leave the system, but how much more? You need to calculate the hydraulics properly including the contol valve at the new conditions. Just because a pressure control valve was specified at a range of normal operating conditions doesn't mean the hydraulics at relief is fully defined by inference. To specify a control valve, I need to give my insrument engineer or the vendor three conditions, minimum/normal/maximum at coincident pressures and (in this case) gas properites; and these are around operating conditions, not relieving conditions. The vendor will come up with an acceptable Cv corresponding to my operating conditions. But what about the Cv at relieving conditions? I can get that number but I have to spend time to get it.

Look, if you want to try and interpret what I think is a dangerously vague statement made by a committee of people who are human (and can and have made mistakes) and de-rate the relieving rate by taking credit for the "normal" vapor rate, go ahead, it's your process and you have to be able to sleep at night. All I can say is if you worked for me, it wouldn't happen.

[Nirav]

Phil,

Thank you for all of your responses. That's what my intension was. To learn or know the things from people much more experienced than me. It helps a lot.

Thanks again.

[deltaChe]

Dear Sir

Could you please upload the sketch (loss of reflux) you mentioned cause I didn't see any ?
Beside that, I didn;t get it the you said "loss of reflux" is not a credible overpressue scenairo.
As I know about loss of reflux, it may casue the condenser quickly flood and the condensity duty become zero.
So, the column must be overpressure if there is no liquid relief.


Could you please explain a little more ?


Thank you very much.