42 replies to this topic

[Alistairm]

Hi,

I am designing a Temperature Swing Adsorption unit for natural gas dehydration using 4A mol sieves.

I'm using the GPSA method to arrive at the dimensions and the wall thickness using SA516 Grade 70 steel. I'm not sure if this is the right material selection. Does anyone have any pointers where I can find the right material? And how to cost it?

FYI:
Pressure: 68 bar
Temperature: ~20-300 oC
Diameter: 2m
Height: 6m
Calculated thickness for SA516 Grade 70 Steel: 2.33 inches

Other info:
3 bed system (1 leading, 1 trim, 1 regenerating)
2 Trains
Cycle time: 22 hours



Thanks

Edited by tonyflow1, 24 August 2010 - 12:11 PM.

[Zauberberg]

For wide variety of Natural gas compositions, Killed Carbon Steel is normally used as Mol Sieve adsorber material of construction - so unless you are drying a gas stream containing significant amounts of Hydrogen or some other aggressive compound(s), you are on the right track. Even for a gas with high H2 content you could run the adsorption cycle with no concerns regarding steel resistance, but using dry, Hydrogen-rich gas for regeneration at elevated temperatures could be a concern. What kind of gas you intend to dry?

[Art Montemayor]

Tony:

I don’t think the GPSA gives you any directions on calculating the wall thickness for a pressure vessel. You must be referring to something else.

Before ever getting involved with the mechanical design of a pressure vessel, you need the established process conditions to be stated and documented. If you are a chemical engineering student, then you should know that you can’t start to discuss or contemplate any material of construction without a thorough knowledge of the chemical and physical properties of the fluids your vessel is going to contain. This basic and fundamental step must be covered first and foremost. We can’t comment on whether the universal SA 516 Grade 70 Steel is adequate for your application. You haven’t even told us the composition of your natural gas! Is it sour (any H₂S?) or acid (any CO₂?)?

I’ve designed Adsorbers, so I know exactly what you are up against. You have a cyclic process and your ceiling temperature is your regeneration temperature. You haven’t stated what the regeneration fluid is. In any event, at 300 ^oC you are quickly approaching the limits on SA 516 – even at a presumed low regeneration pressure – which you haven’t stated either.

I presume this is a student assignment of some sort and you have to generate a size and cost estimate for your instructor. In order to furnish a serious design you have to state the composition of the fluid and all the operating ramifications around the vessel. For example, I don’t believe you are going to regenerate the adsorbent at the 300 ^oC AND 68 bargs.

Give us ALL of the basic data and all of the story and then we can offer some serious comments and possible help.

[Alistairm]

Hi Art,

Thanks for the quick reply. The details are:

sweet gas (downstream of an amine unit) with no H2S and 0.01% CO2
Molecular weight of 19.3 - 82% CH4, 5% C2H6, 3% C3H8 etc
Slipstreaming the dry gas back for regeneration

I've attached a spreadsheet that I'm working on, that should have all the details - hopefully it's not too hard to follow. I've bolded the case of 2 trains (6 beds total) which I think is the optimum configuration.

How come we can't regenerate at a high pressure of around 67 barg - assume a pressure drop of 1 bar?
Also GPSA 12th edition (2004) goes through a set of calcs that determines the thickness of steel (in imperial units):

t = [12.D.P/(37600-1.2)]
D is bed diameter
P is design pressure
37600 is tensile stress of 18800 psi x 2

GPSA also goes through a worked example that, I think, uses a regeneration pressure equal to the adsorption pressure. I went to see my professor just then and he said that the regeneration should be set at about 1 bar less than the adsorption pressure.

We are dehydrating the gas so that it can be processed for high ethane extraction, and the demethanised product sent for LNG liquefaction.

Inlet Conditions:
Flow Rate 816000 kg/hour
Water Content 423.47 kg/hour
Pressure 6800 kPa
Viscosity 0.01345 cP
Density 62.87 kg/m^3
Temp 22.3 oC

Zauberberg, there is no H2.

Cheers

Attached Files

•  NEW1.xlsx   23.08KB   162 downloads

Edited by tonyflow1, 24 August 2010 - 08:44 PM.

[Zauberberg]

Yes, SA516-70 would be the material of choice. It is used in our plant for Mol Sieve vessels, and it has been used for the same purpose in the previous plant as well.

Regarding the regeneration process: yes, it is normally done by using a slipstream of dry, treated gas (about 10% of total flow), but the pressure at which regeneration will be done depends on the following:

- Calculate the amount of heat required for desorption process, including heat losses and heating the vessel walls as well;
- See what quantity of gas is required to supply the required heat;
- Accounting for Heating time = 0.5 X Cycle time, calculate flowrate of the Regeneration gas;
- See if you are within the recommended ga velocity limits (per single bed); if not, you might need to lower the regeneration pressure, or to change the Adsorber vessel dimensions, and start all over again.

Also, the wet regeneration gas is commonly recycled to the feed gas (Amine unit feed gas) after cooling. In that case, you will need a recycle compressor as well.

Good luck,

Edited by Zauberberg, 25 August 2010 - 12:10 AM.

[Alistairm]

Zauberberg,

Great, I've done all that and the velocity is above the minimum required velocity (130m/hour > 90m/hour (GPSA)). This means that I will be regenerating at a high pressure ~67bar, and ~290oC using SA516-70

Is there any way that you could recommend that I find the temperature of the dry gas coming out of the adsorber, the max and min.

And the amount of hydrocarbons that are going to be adsorbed too?

And lastly, when I'm drawing my PFD for this, should I put both trains on one sheet (2x3beds) or just draw one train per sheet?

Regards

tonyflow1

Edited by tonyflow1, 25 August 2010 - 12:56 AM.

[Zauberberg]

Hydrocarbons: normally assume quantity of co-adsorbed Hydrocarbons as ~50% of adsorbed water (kg/hr);

Drawings: Each Mol Sieve Unit in a single set of drawings (one train per sheet);

Temperature profile: it depends on the piping layout, heater outlet temperature, vessel insulation etc., but normally the maximum adsorber outlet temperature during regeneration is about 10-15 degC lower than the heater maximum outlet temperature, and the minimum is very close to the cold dry gas outlet temperature (from another adsorber). You can use a +5 degC higher value.

[Alistairm]

So, that's for the outlet of the regeneration column?

What about the temperature profile of the outlet of the adsorber column - the dry gas to be sent to the demethaniser? Surely it needs to take into account the heat of desorption.

Regards

tonflow1

[Zauberberg]

The temperature change is negligible, due to high gas flows through the adsorber. Even 1 degC temperature increase across the adsorber is too much, based on my experience. This process can be considered as purely isothermal, from practical point of view.

And don't forget the Mercury removal bed, and the Particulate Filter downstream of adsorbers, if you are routing this gas stream to further cryogenic processing.

[Zauberberg]

Attached is the Material list for Adsorber vessel, for your reference. The vessel is internally insulated.

Design P: 76 barg
Design T: -12/320 degC

Attached Files

•  Adsorber Material List.JPG   45.48KB   116 downloads

[Alistairm]

Zauberberg,

Thank you so much, you have been an incredible help. I've now decided to go with a two train system with 2 adsorbing at one time, 1 cooling, and one regenerating, 8 hours each.

When you say that approx 50% of hydrocarbons (compared to water) being co-adsorbed, is there any way that I can make that rough calculation - any reference sources? which hydrocarbons are preferentially co-adsorbed?

I am going to draw up a PFD, and might post some more questions soon if I run into trouble with that. Or possible, do you have any PFDs that can guide me?

I can't speak higher of this forum. After my design project is done, I feel compelled to help out the forum with whichever way I can.

Regards

tonyflow1

[Zauberberg]

Hydrocarbon co-adsorption depends on the Mol Sieve pore size. In your case (4A - probably the best choice), other molecules which are adsorbed are: CH4, C2H6, Ethanol, H2S, CO2, SO2, i.e. all compounds with <4A (Angstrom) effective diameter. You can find that information in almost any Mol Sieve data page on the internet, even including Wikipedia.

As for the PFD, I can't upload documents that belong to the company but in case you have any questions feel free to post it in the forum.

Porocel have some interesting spreadsheets online (bed pressure drop, adsorbent load etc.): http://www.porocel.c.../toolsindex.asp

[Art Montemayor]

Tony:

I’m going to add my comments to your recent data submittal. I can’t make engineering comments through personal email.

You state, “We are dehydrating the gas so that it can be processed for high ethane extraction, and the demethanised product sent for LNG liquefaction” but this seems to be illogical. Your statement is not correct. You must mean “....the de-Ethanised product is sent for LNG liquefaction”) I have to presume that you are drying a methane-rich gas stream in order to prepare it for a cryogenic process downstream where you will extract the heavier hydrocarbons. I also have to presume that since you are eventually going to have the methane liquefied, you are going to meet LNG process specifications. That, to me, means that you are going to dry your gas to a level of less than 1ppm(v) of water content. That also explains why you are using Molecular Sieves.

You also state you are regenerating the adsorbent beds at process pressure. I don’t believe you are going to achieve your desired product dewpoint doing that. The way that I have been able to successfully regenerate Mol Sieves to produce an LNG dewpoint is to regenerate at essentially as low a pressure as I can – atmospheric. That is why it is so important to detail out your COMPLETE cycle – including your regeneration part. As Zauberberg has pointed out, you should be using a compressor to re-compress your regen gas – if you intend to return it to the feed. Picking your regen gas stream and your regen conditions is as vital as your adsorption conditions. If you can’t regen successfully, you won’t meet your dewpoint goal. A drawing of your proposed bed setup would have been very helpful.

Your regen cycle is going to be tough to meet if you continue with what you propose: regen at process pressure (approx. 1,000 psig) using dry product gas recycled back to feed. You will need a compressor for the regen capacity and for the pressure conditions – something that is going to be tough to find. You almost require a reciprocating type of compressor to withstand the high pressure at suction and discharge. Otherwise, you will have to find a very rugged rotary screw or similar design.

My experience tells me you will need up to 700 ^oF as a regen design temperature to ensure you obtain the 1 ppm(v) dewpoint – and that is at essentially atmospheric pressure. Your valves are going to have to be of special design. Orbit valves are the type of choice in this kind of application.

There is a range of preferred Superficial Velocities employed in adsorber design. Your spreadsheet does not account for how you arrived at your bed diameter. The design superficial velocity determines the diameter.

You should design your beds keeping the following in mind: rugged and firm bed supports as well as bed tie-downs. You do not want any bed movement during blowdowns or bed swings. Your drying flow direction should be in a DOWNWARD FLOW, while the regeneration gas should be in an UPWARD direction.

Make sure that your design dynamic sorptive capacity is based on empirical adsorbent manufacturer’s recommendations for the application and that you allow for aging, loss of capacity, attrition, and some poisoning of the beds with time. For beds of this size, always allow for efficient loading and adsorbent removal and your insulation should be internal rather than external. This means that you must allow for a larger diameter to enable you to apply insulation and internal supports as well as positive, non-bypassing of the adsorbent bed.

I hope these comments are of some help.

[Zauberberg]

Very good points.

Interestingly, we used a centrifugal machine as the Regeneration Gas compressor (Sundyne) in LNG plant in Bioko Island, 400HP driver, single stage, flow 44,100 Nm3hr; regeneration pressure was very close to adsorption pressure of approx. 74 barg (essentially the same, minus system pressure drop), and a regeneration temperature of 247 degC (Grace 3A Mol Sieve, slightly lower regeneration temperature than for the 4A). The Regen gas was re-compressed/recycled to the Amine Unit feed gas inlet line.

[Art Montemayor]

As an additional item, I am attaching a paper on designing an adsorption unit. It covers the subject as well as - or better than - most. It mentions and covers a lot of the things my experience verifies.

I hope you find it of some help in your design.

Attached Files

•  How To Guide for Adsorber Design.pdf   536.46KB   261 downloads

[Alistairm]

Hi Art and Zauberberg,

Thanks again for your help and sorry for not replying earlier. Art, your comments and attachment really helped, I was just really pressed for time on Friday because of the first submission, thus I didn't have time to thank you. In the end, the report was really rushed but I believe I managed a very high quality design. I've attached the PFD if you are interested in having a look at what drew - I went for a 2 train, 4 bed mol sieve design. I would be very interested if you have any criticisms of the PFD.

I'm now doing a detailed mass end energy balance & mech design for the unit so I may have some more questions if you are willing to help.

Cheers,

tonyflow1

Attached Files

•  PFD 2 of 6 Dehydration.pdf   210.91KB   192 downloads

[Zauberberg]

Didn't have much time to look into your drawing in more details, but here are my initial comments anyway:

1. All PFDs should use the same topology and terminology, i.e. make sure that you show reference arrows in a correct and consistent way. PFDs should also be tagged, something like Project Number/Project Code/Drawing Type/Drawing Number, including revision number and other information.

2. The exchanger configuration looks uncommon, as you should ensure in 100% cases that Adsorber water load does not exceed the design figure. By having feed gas cooled with some external fluids (apart from cooling water), you expose yourself to a situation where the final gas outlet temperature is not actually controlled. Normally one goes for CW or Propane refrigeration dedicated solely to cooling the feed gas. No heat recovery schemes here.

3. Check your material balance, particularly with respect to water content of cold gas, and liquid hydrocarbon dropout in S-201. I would expect bigger number while in your case there is only 0.05 %mole liquid phase.

4. You have routed this liquid stream to the Waste Water Treatment Unit. What if it contains significant quantities of Hydrocarbons? That is an environmental concern, but also an economic one since you would be loosing potential revenue.

5. I have never seen the same gas being used for heating one bed and cooling the other bed, at the same time.

6. The same applies for Hot Oil heater. Usually it's a fired heater - based on my experience, which doesn't mean that your proposed solution could not work. But I didn't encounter such one so far.

7. Why don't you use Cooling Water exchanger for cooling the Regeneration gas, instead of air cooler? According to the information available in your PFD, cooling water reduces gas temperature to 35 degC while if you use air cooler, it cools the gas only to 45 degC which increases water load on the adsorber, through the recycle loop?

8. The regeneration gas compressor should have its anti-surge loop.

9. Why the flow pattern in Mercury Removal vessel is upflow?

10. You need the particle filter downstream of Mercury Removal bed as well, especially if the gas is processed further in the liquefaction unit.

These are some of the quick observations, I hope I will find some more time soon and have a look into your PFD in more detailed way.

Best regards,

[Art Montemayor]

Tony:

Your PFD is done in good, meticulous detail. You and your cohorts are to be congratulated on its good student quality. However, there are some areas of concern with flow and process logic.

• It is normal design practice to have two, identical capacity, production trains installed in parallel. This gives you the operating flexibility to maintain at least a 50% of rated production capacity while one train may be down. Based on this, I see no reason to have the “recycle” (Regen?) stream from train #2 entering the feed of train #1 as shown.
• I can see a logical justification for having a common, pre-cooling process for both trains. However, what is stream #208? It should be identical to stream #207 in everything. Yet this doesn’t appear as such.
• The use of a 4-tower unit means that you have found that it is more expedient to use 2 vessels in parallel rather than one for adsorption. This is OK, except that you should be fore-warned: unless controlled by instrumentation, it is essentially impossible to ensure that the stream #209 will be equally divided into streams #210 and #211. This simply is naïve over-simplification. Each bed in each vessel must be expected to behave differently vis-à-vis its pressure drop, thereby causing an imbalance in flow splitting. This will happen regardless of ensuring external identical, symmetrical piping and fittings. This is one of the trade offs you have to accept when using 2 parallel towers.
• Although you are adsorbing in parallel with two towers, you show a regeneration system that employs dry product gas in series through the two spent towers. In fact, you show that only one (the second one) tower receives hot regen gas. I am unable to determine how one could ever get hot regen into the first tower. (Note: Never fail to clearly show a separation space between two crossing lines that are not joined in the process. It is very difficult to determine the logic of the flow otherwise.). Without the ability to equally heat and cool both regen towers during the same regen cycle, this system - as shown – will not work.
• A single tower system (or two shorter ones in series) is a much simpler arrangement to operate and control. The series flow is for both adsorption and regeneration.

[Alistairm]

Thank you for your review.

Zauberberg:

2. The 1st HX is a water cooled, the 2nd is another process stream integrated into this unit, the 3rd is a refrigerant from the refrigeration PFD. Would this not be easy to control?

4. Can the knock out water be sent to the amine makeup water?

5. I'm sure that I've seen that in lots of literature - I'll have to double check

Art:

I did make some mistakes that you have pointed out in your review.

As for splitting the feed into two streams - surely this multimillion dollar unit would have some controls taht could manage it, no?

The first tower (after adsorbing) receives the cool regeneration gas because it is still cooling after the regeneration process (this helps warm the regeneration gas and cool down the bed simultaneously).

Regards

tonyflow1

[Zauberberg]

Tony, a quick one:

2. Having refrigerant in the final exchanger will do the work for you - maintaining inlet temperature as you want it to be. That part is OK. I made my comments without knowing that it is a refrigerant loop, since it was not shown in the PFD.

4. Usually, this water stream is recycled to the Rich Amine Flash drum, where it undergoes Water/Hydrocarbon separation. If your stream does not contain any liquid hydrocarbons (please confirm once more), you can use it as a part of make-up water in Amine unit.

I also support Art's view regarding feed/recycle streams split. If you have both trains online, each of them should handle its own feed and recycle streams - no reason to mix it, and then split it again. Unless you have only one set of pre-cool exchangers serving both trains.

[Art Montemayor]

Tony:

We are not criticizing your work product. We try to make comments that will add value to your product. If you make a mistake, as a future engineer you should confront it and make the appropriate corrections or adjustments.

You have not addressed all my comments and questions. These are all directed to draw your attention to misconceptions or errors in judgment that you have made (or it appears that you have made). That is why I try to clearly point out: Without the ability to equally heat and cool both regen towers during the same regen cycle, this system - as shown – will not work. Your drawn process lines cross or seem interconnected. There is no way shown as to how the first tower is going to get any regen heat.

[Alistairm]

Hi Art,

Please don't think I'm taking any of your replies as criticism. I appreciate all your comments and am taking it on board, hopefully it doesn't come across as anything else. After all, you are voluntarily providing this valuable information.

As for the process lines, I'm pretty sure I have them drawn as I intended. The first two towers are adsorbing in parallel. The regeneration gas is slip streamed from the product gas and directed to the 3rd tower for cooling. It is then sent to a heater before entering the 4th tower that desorbs the water. At the end of this cycle the cooling tower becomes an adsorbing tower, and the hot tower becomes the cooling tower.

I liked your comment that the trains should not be mixed with the recycle stream.

I had some errors in labeling the stream, for example #207 & #208 that you pointed out

For splitting the feed into two towers, I've going to have control systems that handles this. You may be right about it being too difficult, but I'm too far into the design to change this.


I'm now trying to determine the thickness of the steel required. If I was to use SA516-70, GPSA states the tensile strength is 18,800 psi (130kPa) at 650 degrees Fahrenheit, but when I search other resources, it is much higher than this (>70,000psi). If I use the GPSA approach I get a thickness of about 2 inches, but if I use 70,000 psi I get about 1/2 an inch. Which one sounds reasonable?

Regards

tonyflow1

[Zauberberg]

The correct resource for allowable stress values is ASME Boiler and Pressure Vessel Code, Section II: Materials.

[Art Montemayor]

Tony:

I have always delegated the mechanical design of major pressure vessels to registered, professional mechanical engineers. I've mechanically designed and fabricated minor vessels and I've always used the steel's ultimate strength (70,000 psi nominally, in the case of A-516 Grade 70) and divided it by 4 to obtain the allowable tensile strength at working conditions.

17,000 to 18,000 psi is the range of most steel designs when you use the ASME code as Zauberberg has correctly pointed out. Don't forget to add the normal, realistic factors to your calculations:

• a corrosion allowance; in your case, considering the diameter of the vessel, 1/16" is not unreasonable;
• an allowance to employ the next, higher, available standard plate size that comes out of the steel mill; this is what all vessel fabricators do and the result is the vessel thickness that, when back-calculated, yields the Maximum Allowable Working Pressure (MAWP) of the vessel.

I am attaching some mechanical design tools to help you do this. I hope this helps.

Attached Files

•  PipeAndShell21.zip   61.2KB   106 downloads
•  4Heads.zip   207.98KB   94 downloads

[Alistairm]

Thanks, I'll have a close look at it tomorrow! The directive is to use my national standard (AS120 for Australia) so I'll need to investigate which steel corresponds to SA516-70 before I move on.

Regards

tonyflow1