Heat of vaporization for relief valve design 1

[MikeClay]

I am having problems sizing relief valves for HC mixtures(C1 to C8). I've looked at it three ways and can get no agreement from the senior staff.

Method One: Choose the heat of vaporization for the mixture as reported by HYSYS. This HoV is essentially an average required to vaporize the entire mass and typically runs in the range of 200 to 500 BTU/lb.

Method Two: Perform fractional flashes and calculate the HoV based on the heat required to vaporize the material. This method includes the sensible heat. Initial vapor commonly has an HoV greater than 2000 BTU/lb. In no case have I seen an HoV less than the average reported by HYSYS.

Method Three: Look at the predominant component and choose that HoV. These values typically have a value of 50 to 150 BTU/lb.

I am inclined to use Method Two since API RP 521 3.15.3.1 and Appendix A.3 support this method. But method 1 is attractive for its ease of use.

Any suggestions?

 

[SooCS]

What is the set pressure of relief valve?

I am supprise the HoV for both Method one and two is so high. You might want to check your input data.

If the molecular weight and the set pressure is not too high, I would use the fig in API. If the set pressure is out of the range, I would set it at 40 btu/lb, which was used to be the recommendated value when there was no data.

 

[pleckner]

I don't do a series of flashes but just one. I flash the mixture to about 1% vapor and use this.

Some clients I've dealt with just take the latent heat of the most volatile component in the mixture. I don't particularly like this but it is probably the most conservative for a mixture.

You may get a lot of opinions about this so you will have to make your judgement call but my opening statement has been my way of doing it for over 10 years, and it has been approved by many clients.

As far as you not wanting to go the extra step to do it correcly and prefer the easy way out, no offense, don't be lazy.

Oh, the recommended latent heat by API when there is no data is now 50 But/lb max. They recommend an evaluation because you may even be able to justify an even lower latent heat.

 

[JoeNava]

What I ´ve been doing is to flash the stream at accumulated pressure and 1% vapor. And getting the enthalpy difference between phases. The first 1% is generally accepted because is going to give you the bigger specific volume and smallest vaporization heat (lighter components), i.g., the relief load.

Anyway, in a case I was asked to calculated at several vapor fractions (25%,50%, 75%), obviously enthalpy difference spread, liquid became heavier and got bigger HV, but temperature was higher, and I got a little increment in the relief load right after 25%, but it did not exceed the relief load at 1%.

 

[Assumptions]

Depends on the installation. If it is a fire case then:
Is it a tower or a vessel where there is a uniform composition?
If it is a tower then I would take the composition at the top for your analysis. If it is a vessel then I would do an analysis over time as the batch is vaporized.
But it is also dependent on the scenario.

 

[EGT01]

As far as methods go, I've also used the 1% approach, taking the initial liquid composition up to its bubble point at accumulated pressure then flash off 1% of the mixture. Maybe the merits of that are yet to be proven but from what I've seen, that approach seems to be typical.

I suppose if you really wanted to rigorously size your relief valve you would do a time dependent simulation to account for change in composition as material is vented but I can't say I've seen that approach generally taken except for runaway reaction. However, I happened to find the following link which had an interesting case study that I wish I had more time to analyze.

http://www.iomosaic.com/iomosaic/pdf/bench1.pdf

Then from a time dependent perspective, I think you certainly want to approach the evaluation based on flashing a small portion to estimate HoV, this would be done incrementally. Actually, I think that is what happens when you look at a condensation (or vaporization) curve used for sizing heat exchangers. A portion is vaporized then the remaining liquid is taken and a portion vaporized again usually in even increments until it travels the temperature range you specify.

Using BJAC to look at an arbitrary initial mix of 10wt% C3, 20wt% n-C4, 30wt% n-C5, 40wt% n-C6 and taking it from bubble point to dew point, there is very little difference in the latent heat reported. Reversing the numbers in the composition gives a little more difference with some drop in latent heat after about the first 30% vaporized but then changes little after that. Of course, as others have pointed out, HoV is not the only thing that affects the size. That may be what happens in the link example referenced above, after a quick check, it looks like the vapor mol wt decreases with amount vaporized.

Okay, whether or not you perform a fractional vaporization approach probably should depend on the system which you are evaluating. But on another note, after you perform a flash calculation, how is everyone determining the HoV? Does your software give that to you?

I have to agree with SooCS, the HoV numbers given by MikeClay seem to be high but we obviously don't know the details to which they are referred.

My understanding what you are interested in is the HoV of the vapor that is generated, Right? The classic definition of latent heat
"the amount of heat required to convert unit mass of a liquid into vapor without a change in temperature"
really doesn't apply for a mixture since temperature won't be constant through the vaporization process. But we are interested in the amount of heat required to generate 1 unit mass of vapor. That way we can take
(fire_heat_load)/HoV = relief rate.

For a liquid mixture fed to a flash block that generates some vapor and remaining liquid then (neglecting sensible heat to the vapor)
Vapor_HoV = ((Enthalpy_in) - (Enthalpy_out) - (&[ignore]Delta[/ignore];Enthalpy_Liquid))/(Vapor_formed)

Another approach I've seen is to take the weight fraction composition of the vapor formed from the flash block along with the pure component latent heats at relief temperature and determine as follows
Vapor_HoV = &[ignore]Sigma[/ignore];i(WFi * LHi)

Seems I've heard of another approach of taking the difference between enthalpy/unit mass of the remaining liquid at its bubble point and enthalpy/unit mass of vapor formed at its dew point, but I'm not sure about the dew point and was never keen on that approach.

Comments?

 

[CHD01]

What all of you are addressing here is determining the vapor rate required to depressure the API tank within recommended guidelines in API521.

The originator in his first sentence, uses the term "relief valves" - I just want to caution the originator that this section in API addresses "depressuring" to protect the tank against rupture due to the unwetted tank wall being exposed to fire.

This protective system described in API521, 3.19 functions separately from the tank emergency vent (not really a true relief valve); and a protective relief device to limit max pressure is still required; IN ADDITION to the depressuring system you are discussing.

Determining the required relief rate for the tank's emergency vent; and the rate for depressurization are two separate calculations.

I just wanted to be sure that these issues do not become mixed in our comments and lead to some confusion.



The more you learn, the less you are certain of.

 

[joerd]

HYSYS and Pro/II both allow you to use the depressuring utility/unit operation to do fire case relief scenarios. This is similar to the fractional flash approach, and usually seems to give a lower relief rate than calculated by the "classic" methods.

Cheers,
Joerd

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[CHD01]

JOERD: Just because HYSYS and ProII do it does not mean its right. For example, I once found a well known Corporate program contractors were using that included BOTH recoverable and non-recoverable losses in calculating the inlet pressure drop to a relief valve (clearly a violation of the ASME Code). They were including the velocity losses for a reducer. Don't want to do this; only friction losses are included for inlet/outlet pressure drop for a relief device. In another case I found a bug in ASPEN where too much input data resulted in the wrong answer.

Getting back on subject; if we are using a pressure control valve for depressuring; I would not rely on this for my relief valve. I believe this to be a viloation of ASME Code. Normally we try not to rely on control valves and interlocks for relief, although we do rely on them to reduce the severity of a release.

Now this is my opinion and I recognize the right to disagree; I will try to find something in our codes that might support my position.



The more you learn, the less you are certain of.

 

[EGT01]

CHD01,

You give good advice about scrutinizing the basis of and the results from computer programs. That was one point I was hoping to explore with how HoV was being determined.

But, I have not interepreted this thread to mean that relief valves were being used as depressuring devices. And maybe this is a point that MikeClay could clarify.

This is what I believe is at issue, look at API RP-521, 4th edition, March 1997, Section 3.3 Effects of Temperature, Pressure and Composition
"During pressure relieving, the changes in vapor rates and molecular weights at various time intervals should be investigated to determine the peak relieving rate and the composition of the vapor."

Now, look at the link I provided earlier for the case study where a 2H3 and 2J3 relief valves were evaluated for adequate capacity to prevent overpressure. It is not clear from the case study but I would assume that both valves may be adequate at the onset of relief but if the relief event continues for a period of time, the 2H3 becomes inadequate.

Whether you use a fractional approach over time or just an initial fraction vaporized may not make much difference when dealing with components that are similar. However, in the case study, it would appear that though HoV may not change much over time, if the molecular weight of the vapor decreases over time, you may want to take a closer look at the time dependency effect.

 

[CHD01]

EGT01: Interesting discuusion. I believe I understand what you are saying. I can only provide the approach that we generally use at our site. At my site we do look at incremental changes in the relieving rate to determine the peak rate of release where applicable (usually for reactive conditions, 2-phase flow, or some extreme flash cases); we then use the peak release rate to size the required relief device. Once a relief device is selected, an analysis is then done for the inlet and discharge lines to determine its affect on the relief device capacity once we have selected the relief device.

I think we are pretty much in agreement here; its basically a matter of the engineer understanding the relief system and what is required to understand the nature of the release and the transient flow conditions it undergo's. I guess I would, however, prefer to make the sizing basis of the relief device as simple as possible so long as we do not significantly oversize a device or endanger anyone from a release.

Last, I was just saying that I think pressure control valves (for example) designed to vent and depressure a tank or vessel should not be relied upon as the sole relief device for the tank or vessel - that a separate true relief device is required. I'm interested if you can support that or have difficulty with this suggestion? I can undertand arguing this position.

Thanks for comments.







The more you learn, the less you are certain of.

 

[pleckner]

Yes, in my opinion, ETG01 has it correct and my previous repsonse was a little too simplified. The 1% flash method to get the latent heat was the way I did the calculations for years and they were approved by many clients (they didn't know any better either). However, I've seen a number of studies that show that some mixtures do indeed go through a maximum relieving rate as the contents are boiled. So to do it right, one really needs to go through the cycle.

 

[EGT01]

CHD01,
Sorry if it sounded that way but I have no difficulty with your suggestion.

 

[engzeidan]

There is a misconception that light components have lower latent heat. NOT TRUE. Light components has lower boiling temperature but that does not mean that they have lower latent heat. Look at Cox chart and you can confirm that.

One issue in the way the latent heat was calculated is taking the difference between vapor and liquid phases enthalpies at 1% flash. This is NOT correct (only valid for pure component flow). For a mixture, take the heat flow difference for the stream at 0% flash and at 1% flash. Then devide by the mass flow.

If you have water or non-condensibles in your stream, the latent heat for the initial flahses will be huge. Since the presence of water and non-condensibles is hard to predict accurately in normal refinery operation, it is safer to exclude these components from the latent heat calculations.