Jump to content



Featured Articles

Check out the latest featured articles.

File Library

Check out the latest downloads available in the File Library.

New Article

Product Viscosity vs. Shear

Featured File

Vertical Tank Selection

New Blog Entry

Low Flow in Pipes- posted in Ankur's blog

Tuning Stage Efficiencies (Murphree) For A Cdu In Aspen Plus


This topic has been archived. This means that you cannot reply to this topic.
5 replies to this topic
Share this topic:
| More

#1 TX-2

TX-2

    Junior Member

  • Members
  • 26 posts

Posted 21 April 2016 - 01:33 AM

I am modelling a CDU using Aspen Plus. The current objective is to tune the model to the actual plant behavior (test run data) so that the CDU model can be further integrated with preheat trains, VDU, stabilizers etc for optimization studies. 

 

I have test run data for the CDU, including D86 volume curves for the crude and its products which I am trying to match using an Aspen Plus CDU model (PetroFrac). I am able to converge the column with all its pumparounds, strippers etc but I am having trouble matching the D86 curves for the products that I get from the model. 

 

For example, the Heavy Naphtha cut from the test run has an IBP: 104 C and FBP: 144 C while I get a much wider range from 87-177 C. Similarly, I have issues with Kero, LGO, HGO cuts as well, where the differences are even greater. Even if I ignore the IBP and FBP and look at 5-95% curves, the differences are significant. Moreover, I do not see a trend, the Light Kero cut is surprisingly closer to the test run. I am unable to manipulate draw flowrates or return temperatures as those have been provided by the client and cannot be changed, that leaves me with only being able to manipulate the stage efficiencies. I will appreciate if you could provide some guidelines on how to manipulate the efficiencies so that my cuts have a better/closer D86 fit or if there is anything fundamental that I may be overlooking. I am using section-wise efficiency method and have divided it into 10 sections.

 

Section Starting stage Ending stage Efficiency
1 1 1 0.9
2 2 30 1
3 31 33 0.6
4 34 37 0.7
5 38 40 0.52
6 41 44 0.82
7 45 47 0.78
8 48 53 0.6
9 54 57 0.53
10 58 58 1

 

I am attaching an Excel spreadsheet with the BPs of the cuts that I get from the model vs the test run data (highlighted in yellow). Thank you very much for your help in advance. 

 

 

 

Attached Files



#2 PingPong

PingPong

    Gold Member

  • Members
  • 1,446 posts

Posted 21 April 2016 - 12:09 PM

IBP and FBP are not very accurate from the Lab, and very inaccurate from any simulator.

5 % en 95 % D86 are not accurate from any simulator.

 

But simulated 50 % D86 should be fairly accurate, so the very big difference between actual and simulated indicates that your simulation is very wrong. Why that is cannot be determined from this data only, but would require an experienced design engineer looking at all the operating (test run) data as well as your model.

 

In any case you first of all need to have an accurate TBP curve of the whole crude, preferably a SimDis, or at least a detailed crude assay.

A D86 of the crude is useless, no matter what the refinery people may say.



#3 TX-2

TX-2

    Junior Member

  • Members
  • 26 posts

Posted 24 April 2016 - 05:40 AM

IBP and FBP are not very accurate from the Lab, and very inaccurate from any simulator.

5 % en 95 % D86 are not accurate from any simulator.

 

But simulated 50 % D86 should be fairly accurate, so the very big difference between actual and simulated indicates that your simulation is very wrong. Why that is cannot be determined from this data only, but would require an experienced design engineer looking at all the operating (test run) data as well as your model.

 

In any case you first of all need to have an accurate TBP curve of the whole crude, preferably a SimDis, or at least a detailed crude assay.

A D86 of the crude is useless, no matter what the refinery people may say.

 

Thank you very much. I incorrectly mentioned that I am using the D86 of the crude, however the TBP of the crude is available and I have been using it, the D86 are provided for the products. I was able to make changes in Petro Characterization Options within Aspen Plus and get a significantly better fit. 



#4 shvet

shvet

    Gold Member

  • Members
  • 135 posts

Posted 01 May 2016 - 10:22 PM

 

A D86 of the crude is useless, no matter what the refinery people may say.

 

What about API TDB method 3A1 for ASTM D86 intercorvention to TBP? How do you think, can it help?



#5 Pilesar

Pilesar

    Gold Member

  • Members
  • 1,347 posts

Posted 02 May 2016 - 12:52 PM

Tex,

You will not achieve your goals by the method you are using. I suggest you start over. A few points:

1) You want to match your product flows and compositions. So start with the D86 characterizations of all your product streams with their measured flow rates. Combine the streams together to characterize the feed stream. This WILL NOT match the characterization you have for your current feed, so ignore the measured feed characterization! Because the product streams are over a much narrower distillation range, they are measured much more accurately than the TBP of the crude feed. How recent is the TBP of your crude feed anyway? Crude composition changes with time as it comes from under ground. Use the more recent product stream characterizations or you will never get a match with your simulation! How can you expect the simulation to give you results to match your product measurements when your input feed stream can not make the material balance? It just will never happen.

2) Use an overall efficiency instead of individual tray efficiencies in your model. Choose 16 to 25 theoretical stages. You will only have 1 stage between each pumparound inlet to pumparound return. It is not 'better' to use individual tray efficiencies, but will provide worse answers if you can get answers at all!

3) You should be able to match your product streams pretty well. Adjust the light ends in your pseudo feed to get close to the column overhead temperature if you want. If you cannot match the overhead temperature, live with the discrepancy as your main goal is to match the side products.

4) The number of cut points in your assay matters a lot. Make sure you have overlapping cuts between products. Feed characterization is much of the battle in modeling a crude column. If in doubt, add cut points after your model converges and see if the answer is significantly changed.

5) You should be able to get good guidance on modeling your crude column from your simulation vendor. They may even have a worked example to offer. There is published literature available, but I don't know specifics of anything currently in print.

6) Do not expect the bottoms products stream to give you enough detail to model your vacuum column. You will need a D1186 or better to characterize the vacuum column feed. You will need more cut points for that simulation, also, as your product separations will need more definition in the important ranges.



#6 TX-2

TX-2

    Junior Member

  • Members
  • 26 posts

Posted 27 May 2016 - 03:38 PM

Tex,

You will not achieve your goals by the method you are using. I suggest you start over. A few points:

1) You want to match your product flows and compositions. So start with the D86 characterizations of all your product streams with their measured flow rates. Combine the streams together to characterize the feed stream. This WILL NOT match the characterization you have for your current feed, so ignore the measured feed characterization! Because the product streams are over a much narrower distillation range, they are measured much more accurately than the TBP of the crude feed. How recent is the TBP of your crude feed anyway? Crude composition changes with time as it comes from under ground. Use the more recent product stream characterizations or you will never get a match with your simulation! How can you expect the simulation to give you results to match your product measurements when your input feed stream can not make the material balance? It just will never happen.

2) Use an overall efficiency instead of individual tray efficiencies in your model. Choose 16 to 25 theoretical stages. You will only have 1 stage between each pumparound inlet to pumparound return. It is not 'better' to use individual tray efficiencies, but will provide worse answers if you can get answers at all!

3) You should be able to match your product streams pretty well. Adjust the light ends in your pseudo feed to get close to the column overhead temperature if you want. If you cannot match the overhead temperature, live with the discrepancy as your main goal is to match the side products.

4) The number of cut points in your assay matters a lot. Make sure you have overlapping cuts between products. Feed characterization is much of the battle in modeling a crude column. If in doubt, add cut points after your model converges and see if the answer is significantly changed.

5) You should be able to get good guidance on modeling your crude column from your simulation vendor. They may even have a worked example to offer. There is published literature available, but I don't know specifics of anything currently in print.

6) Do not expect the bottoms products stream to give you enough detail to model your vacuum column. You will need a D1186 or better to characterize the vacuum column feed. You will need more cut points for that simulation, also, as your product separations will need more definition in the important ranges.

 

Pilesar,
 
Thank you very much for your detailed response. It has been over a month since I posted this question, during this time I learnt a lot about the challenges of crude characterization and the importance it has on the overall simulation results. I fully agree with you, Garbage In = Garbage Out is well and truly applicable here. I actually thought about combining all the products and getting a feed stream out of it, unfortunately I did not have the CDU overhead and RCO characterization. After a lot of trial and error and manipulating the crude itself, I was able to get reasonably good product D86 curve fits. Crude TBPs were several years old (SimSci Spiral) but after playing around with the characterization methods within Aspen Plus and light ends, I was able to get good CDU product D86 curves and column temperature profiles. I have divided the CDU (57+1 stages) into 10 sections, being able to tune the individual section efficiencies eventually produced good results.  The data fit and regression tool within Aspen Plus proved to be very handy. I fully appreciate what you mentioned under point 6, in absence of the actual RCO characterization the bottoms from the simulation is the feed to the VDU and I am having a very hard time with the VDU product D1160 curves, in comparison it is making the CDU look like a piece of cake. 





Similar Topics