9 replies to this topic

[Zauberberg]

Prior to submitting petroleum fluid (crude oil or any hydrocarbon fraction/product) to characterization process in simulation software, validation of available laboratory data is required in order to determine if existing set of analyses is appropriate for model development. One of the methods used for that purpose, is plotting "Campbell's diagram" - graphical representation of log(Ki) versus Tc values, where:

Ki - individual components K-values
Tc - individual component critical temperature

For every fluid submitted to characterization, 100% linear curve should be obtained. According to the base rule in this procedure, for the light components, any significant deviations from the linear relationship indicate possible non-equilibrium separation, suspect analyses or numerical errors in the data reporting. Because it is a log scale any variation of log K of more than 0.1 is significant.

In one of the projects I am working on, detailed characterization of petroleum fluid is required in order to set very accurate basis for Gas Concentration Unit design - involving compressor train, condensate stabilizer, gas dehydration and sweetening units, amongst others.
I have encountered very strange results, especially in light-ends region (encompassing C1, N2, CO2 and H2S). While the Campbell's line is 100% linear for C2+ fraction, methane/N2/CO2/H2S do not follow this same pattern. It looks like analyses are invalid, but I would like to be absolutely sure about this.

I did many laboratory data validations before, but fluids submitted for characterization were 100% hydrocarbons (basically with negligible content of other components). I need help from ChE forum members regarding following issues:

1. Is "Campbell's rule" applicable only for pure hydrocarbon systems, or other components present in the sample should also obey this criteria?
2. Is the linearity of Campbell's curve such that there cannot be any changes in curve's slope - from negative to positive value?

I am attaching Excel worksheet for easier tracking. Any assistance would be appreciated.

Thanks in advance and best regards,

[attachment=741:Campbell.xls]

[joerd]

Good post. I think the plot is only applicable to hydrocarbons, but C1 should definitely be on the line. It looks like you need to use Kelvin, not Celsius for your temperature units. Also, in your post you state the x-axis is Tc, but it should be Tc² (Tc in K).
However, doing this doesn't provide the straight line that you are looking for, so your component breakdown (or the determination of the Tc values) looks a little suspect. I'd be interested to see what you find out.

[Zauberberg]

Thanks Sjoerd, I think I got it today.

I realized that temperature scale has to be on absolute basis (Kelvin or Rankine), otherwise the negative values of Tc in degrees Celsius would spoil the curve.
Literature resources are showing respective curves for 100% hydrocarbon systems only, but there is universal (physical) meaning of Campbell's procedure which makes me think all components present in the sample(s) should follow the same pattern. Now, the second - big - question is how simulation software (HYSYS in this particular case) calculates solubilities of non-hydrocarbon species (N2, CO2, H2S) in hydrocarbon liquid phase. Based on my experience with HYSYS, solubility of non-hydrocarbon compounds (and especially gases) in hydrocarbon liquid is always lower than measured in practice. As you can see, only slight deviation in calculated solubility causes shifting of log(Ki) away from linear pattern because Yi/Xi calculated by software is higher than, I guess, it should be.

Here is Campbell's plot, revised:
[attachment=744:Campbel_REV.A.xls]

Thanks for support,

P.S. I see that two distinct curves for one single system can be made: one is based on HYSYS flash separation of given feed (I see it as "ideal" Campbell plot) and the other one is based on separator liquid and gas sample compositions - which is actually the "true" Campbell's curve. Additional tool in characterization process is "mass balance plot", where compositions of feed, separator liquid and gas are entered and simple mass balance calculation is performed in order to validate laboratory analyses of all three samples. More information is available in excellent Campbell's book, "Gas Conditioning and Processing" and Shell's procedure for crude oil characterization using C7+ unit operation in Hysys.

[joerd]

I totally agree.
Campbell uses the curve mainly in the second sense, to build a consistent set of parameters (K-values) out of a set of experimental data. I would think that the first use of the Campbell plot (the "ideal" version) helps you to see if the parameters of the different hypo-components are correctly chosen so that you get reasonable VLE predictions.
I share your opinion about solubility of non-hydrocarbons in a hydrocarbon phase. I am not sure if it is just a practical / measurement / non-equilibrium phenomenon that explains the discrepancy, but I also find that field data usually shows a higher solubility.

[rxnarang]

Could you gentlemen please give me reference to this method? Is it in the Campbell's Natural Gas processing book?

Any pointers will help. Thanks

Rajiv

[Zauberberg]

Hello Rajiv,

As I wrote in my second post, the references are:

1. John M. Campbell, "Gas conditioning and processing", vol. 1: The basic principles
2. "Guide for fluid characterization using C7+ method in HYSYS", Shell GS.06.53139.

Best regards,

[rxnarang]

Got it. Thanks a lot.

Regards
Rajiv

[mishra.anand72@gmail.com]

Leave it.

[Steve McGahey]

Thanks Sjoerd, I think I got it today.

I realized that temperature scale has to be on absolute basis (Kelvin or Rankine), otherwise the negative values of Tc in degrees Celsius would spoil the curve.
Literature resources are showing respective curves for 100% hydrocarbon systems only, but there is universal (physical) meaning of Campbell's procedure which makes me think all components present in the sample(s) should follow the same pattern. Now, the second - big - question is how simulation software (HYSYS in this particular case) calculates solubilities of non-hydrocarbon species (N2, CO2, H2S) in hydrocarbon liquid phase. Based on my experience with HYSYS, solubility of non-hydrocarbon compounds (and especially gases) in hydrocarbon liquid is always lower than measured in practice. As you can see, only slight deviation in calculated solubility causes shifting of log(Ki) away from linear pattern because Yi/Xi calculated by software is higher than, I guess, it should be.

Here is Campbell's plot, revised:
[attachment=744:Campbel_REV.A.xls]

Thanks for support,

P.S. I see that two distinct curves for one single system can be made: one is based on HYSYS flash separation of given feed (I see it as "ideal" Campbell plot) and the other one is based on separator liquid and gas sample compositions - which is actually the "true" Campbell's curve. Additional tool in characterization process is "mass balance plot", where compositions of feed, separator liquid and gas are entered and simple mass balance calculation is performed in order to validate laboratory analyses of all three samples. More information is available in excellent Campbell's book, "Gas Conditioning and Processing" and Shell's procedure for crude oil characterization using C7+ unit operation in Hysys.

G'day Zauberberg,

Looks like I got your PM too late! That' OK though - Campbell plots aren't my thing. HYSYS, however, is a bit more my thing.

You've got some problems with the solubilities of the non hydrocarbon in the hydrocarbon (HC) mix.

I've been doing some flow assurance work with a linked HYSYS-OLGA model recently, and I was suprised to see the odd phase envelope that HYSYS was predicting for my hydrocarbon + CO2 + H2S + N2 system. Turns out a pile of the binary interaction parameters (BIPs) (for Peng Robinson) were rather rubbish.

They were rubbish, specifically for the interaction of CO2 and N2 with the HC's in the component system.

I ended up grabbing some different BIPs from PVTSim, which has (in the opinion of my thermodynamically inclined colleague here) a much more reliable physical property databse.

Using those BIPs really improved things for me. If you're using HYSYS 2006, you can also include the Peneloux correlation, which improves density (I think it's the densities) predictions.

Hope this helps,
- Steve.

[Zauberberg]

Thanks Steve, I think I solved this problem.

The idea was to focus on operating envelope of Gas Plant (the range of temperatures and pressures is given as design basis) and examine fluid phase behavior within these parameters. Since there was a set of laboratory analyses provided, describing condensate equilibrium at different pressures, I managed to define plant feed completely. And sensitivity analysis has shown the deviations are negligible.

It is strange, though, that PVTSim - based on my experience - is considered as quite unreliable for getting HYSYS inputs. And the way how pseudocomponents are managed in subsurface model doesn't give me any confidence in this software.

Greetings,