14 replies to this topic

[SeanHawkes]

Hello All,

I am a Junior Process Engineer working for a Design Firm that deals with companies working with the Oil Sands resource in Alberta, Canada.

I am trying to use a paper written by Herbert Loria, Pedro Pereira-Almao and Marco Satyro in order to determine the density and viscosity of bitumen with an excel spreadsheet. Currently I am using the data provided in the paper to check if the spreadsheet is working. I am trying to find a way to calculate these properties for use in process data sheets (Can't get into too many details because I can't discuss clients.)

The paper calls for the critical properties (Tc, Pc, MW and acentric factor) to be calculated using the lee-kesler correlations. Using these properties and the derivations in the paper, as well as the SOLVER function in excel you are able to estimate the overall density of the bitumen at different temperatures and pressures (that was what made the method for determining these properties so appealing initially.)

I used HYSYS to calculate the critical properties of the bitumen based on the assay data provided in the paper and they were off by 10% percent or so (I used other correlations than Lee-Kessler initially, but after switching the correlations the critial properties still diverged by about 10%). Also, the final results for the density of the bitumen completely diverged from the results posted in the paper.

I have attached the paper along with the essentials from the excel spreadsheet (I'm sorry if the spreadsheet is confusing, reading the paper will probably help.)

Another possible solution to my problem would be to find another way to estimate the bulk properties of bitumen (and bitumen blends) other than the method proposed in the paper, All possible solutions are welcome.

On a side note, is there anywhere that assay data for bitumen from the different Alberta regions can be found on the internet?

Regards,

Sean

Attached Files

•  Viscosity and density determination (Cheresources).xls   46.5KB   345 downloads
•  Prediction+of+Density+and+Viscosity+of+Bitumen+Using+the+Peng-Robinson.pdf   141.42KB   392 downloads

[JMW]

I always get caught out by the terminology.

What I call bitumen isn't necessarily what others call bitumen; and in the Candian oilfields, it seems "bitumen" is what they call the crude oil - before or after dilution with distilates from the refinery for pipeline transport?
So I'm assuming that basically that is what you are looking at here.

I can't help you further than that, but It may help some others.

[gegio1960]

Hello SeanHawkes.
I think you should test your spreadsheet vs lab data of your crude oil.
Other general hints around heavy HC stuff:
- it is better to use Maxwell-Bonnel (also called Esso Tab) instead of Peng Robinson
- ProII is better than the other process simulators (for this type of components)
- never rely on simulators for property predictions (especially viscosity). You shall have a good set of lab data to create and check your simulations.

[SeanHawkes]

Thanks for the replies,

The paper mentioned in my first post uses lab data to compare the percent error of the calculation method for density and viscosity, and they had about a 1% error if I'm not mistaken. I'm thinking of abandoning using the papers method (troubleshooting problems with large excel spreadsheets can be a pain) and using HYSYS to determine the properties I need.

I used some bitumen and naphtha (used as a diluent in this application) assay data to perform a HYSYS simulation, the purpose of which is to look at the density, viscosity, heat capacity and thermal conductivity of a stream enterring and leaving a sample cooler (I'm also using the simulation to look at the duty of the sample cooler). @gegio1960,The simulation seems too be pretty accurate, but you mentioned that for viscosity, simulation data isn't very trustworthy. Is that because the nature of viscosity makes it difficult for simulators to predict? Should I be suspicious of the other values (density, heat capacity, thermal conductivity) that I am using my simulation to predict?

The reason I wanted to use the paper in the first place was to have another way to determine the properties determined by the HYSYS simulation (particularly the density and the viscosity) in order to compare the values and have to avoid requesting the clients to test the viscosity of the fluid in the lab.

Regards,

Sean

[JMW]

If the clients have a lab, (as they must surely do?) then they should have a wealth of viscosity and density data covering compositions and pipeline temperatures.
Of course, if what you need is the viscosity and density at a range of pipeline temperatures, that may not be so easy - in normal applications - but if your client is a pipeline operator or feeding a pipeline then they should have a lot of real world data available.

Of course, if you are trying to work out properties in a processing plant and not simply pipeline, then you will need properties at a range of different conditions which becomes a bit more difficult. Take extra care with viscosity.
But surely it is better to work with actual raw data than generic data and your clients are often the best source of this?
Why try to work the simulations when you can have the real thing?

[Technical Bard]

In my experience, simulators (HYSYS, Pro/II, others) all fail to accurately estimate the density, viscosity and thermal conductivity of bitumen unless you have extensive laboratory data - and most Clients have not done enough lab work to get that data. Further, diluted bitumen blends are often not estimated properly by the simulators because their blending methods are estimating properties that were generated from other estimated properties. VMGSim claims to have a better blending method in their latest version, but I have not tested it.

I have done significant work in the field you are in, and I would strongly recommend asking your customer to do more lab work to actually measure the properties of these oils and blends. Bitumen (Alberta extra heavy crudes) are highly aromatic, have no light ends (IBP is kerosene range, usually), and the available diluents have differing shrinkage and viscosity effects. If you are planning on distilling the material, a measured molecular weight curve is critical as the traditional methods will overestimate the molecular weight.

[Technical Bard]

Further to your last question - there is no generally available set of bitumen assay data from the various fields in Alberta. No one has seen need to publish it. There is data available from crudemonitor.ca and some of the producers who are trying to market some diluted material - but these assays are generally superficial and do not contain enough data for design purposes.

[gegio1960]

SeanHawkes,
I substantially agree with Technical Bard comments.
What I can add, based on my experience of design/revamp in Vacuum Distillation Units (VDU) and in Linear Programming:
- if you have good lab data, you can put them in the "User Property" of the simulator and then obtain pretty good results from the streams obtained by your model. For instance, you can (carefully) use them for blending purposes.
- the problem is that the User Properties can't replace the simulator built-in properties. For instance, if you input your lab data for viscosity they are not used for heat exchanger calculation.
- generally speaking -but it's not so easy to go in details- you should generate your "model" by testing some lab/field data. Once you've obtained a good enough reproduction, you can extend the model to analyse the modifications of some process parameters and obtain new designs by applying some "safety factors".
Just for info, Mr Satyro was one of the developers of Hysim, the "father" of Hysys (and Unisim etc.)

[gegio1960]

SeanHawkes,
do you really need a simulation to design a sample cooler? This kind of items are generally copied from "typicals".
You wrote: "The simulation seems too be pretty accurate, but you mentioned that for viscosity, simulation data isn't very trustworthy. Is that because the nature of viscosity makes it difficult for simulators to predict? Should I be suspicious of the other values (density, heat capacity, thermal conductivity) that I am using my simulation to predict?". (Also in this case) generally speaking, if you start from good characterization data, you shouldn't have big discrepancies for density and heat capacity whilst the same doesn't apply to viscosity and thermal conductivity.

[SeanHawkes]

Hello all, I've found the best way to predict the viscosity of the bitumen blends ended up being ASTM D341, which requires historical data.

[GS81Process]

Sean,

I would recommend that you perform the lab testing as mentioned by others. It seems important to confirm the bitumen's distillatation curve, otherwise any estimates of density and viscosity using the method outlined in the paper would be inaccurate. And if you're using the lab anyway, why not measure the viscosity at various temperatures and pressures?

I have seen a correlation to estimate the liquid viscosity in Puttagunta et. al (1993). I would be cautious though because the viscosity can change drastically over small temperature increments at lower temperatures.

[JMW]

If you have found ASTM D341 to be the best method then try the spreadsheet in Cheresources http://www.cheresour...wnloads&mid=987
Note that this spreadsheet requires you to enter the viscosity at two different temperatures in order to find the viscosity at another temperature.
In actual process in the Candian Oilfields, they use two inline viscometers separated by an electirc heater to generate a continuous real time measurement of the viscosity of the "bitumen" at two different temperatures so as to derive a real time measurement of the viscosity at a reference temperature (the temperature and viscosity limits being set by the pipeline operators).

However, what I discovered was that the operators used this technique to create a library of data comprised of the two viscosity measurements for a range of conditions (different oil and distillate blends) which they then used as reference curves for the multi curve method.
Technically the multi-curve method enables a single process viscometer to measure the viscosity at the process temperature and then uses a ratio method to compare the measured viscosity to the viscosity in the reference curves at the same temperature and then applies this ratio to the viscosities from the reference curves at the target temperature to infer the viscosity of the process fluid at the reference temperature.
It is worth noting that a dual viscometer system fully installed (including civils, cabin etc) can be expensive at CAN$800,000 or more but the real time measurement of the accuracy achieveable can deliver a payback period of around 12months (very conservative - one user is said to have achieved the ROI in 4 months).

The multi-curve method is also available as an excel spreadsheet which I haven't published but if you'd like a copy, please let me know.

[SeanHawkes]

JMW,
I know its months later, but we ended up suggesting (based on the guidance of one of our technical advisors) the very configuration you have just described. Our clients wanted to hold firm to the idea of cooling to the reference temperature, even though the viscosities were extremely high (350-400cSt) which causes some flow regime issues. We had some doubts about the the viscometer measuring a representative sample due to the highly laminar flow combined with the very short vibrational penetration depth of the viscometer probe. That is another reason the heating configuration is so useful, heating the bitumen reduces the viscosity and can avoid this problem.

Regards,

Sean

[JMW]

Hi Sean,

sorry for the delay responding,

I haven't logged in for a while.

 

When measuring the viscosity online a great deal depends on the selection of the viscometer and the installation.

Marimex I know have been working on this but I couldn't tell you how successful they are. Their sensors are suitable for dynamic viscosity at very very much higher viscosities than here but measure the dynamic viscosity only. However, in pipeline applications it is necessary to measure the density as this is a critieria also applied to the accetable pipeline limits.

 

Many of the successes in Canada have been with the Emerson 7827 digital viscometers (that measure density from the resonant frequency and hence also the kinematic viscosity necessary for the ASTM D341 calculations but while accurate enough for the viscosit, not accurate enough for the density control.

 

A more recent viscometer, the LEMIS DC 52 ViscoAnalytic, is comparable to the Emerson 7827 but has fiscal density accuracy capability.

 

Note that the LEMIS and Emerson sensors displae the fluid and can measure density an dynamic viscosity and hence generate the kinematic viscosity. The Marimex uses torsional vibration which does not displace the fuid and a different method of determining the viscosity from the vibrational sensor (amplitide sensing I believe rather than bandwidth) and hence needs separate density measurement.

 

In real terms the Emerson and Lemis sensors are routinely used for viscosities greater than 400cSt. For example in marine bunker fuel applications where the viscosities of the commonest fuels are around 380cSt and higher with some fuels.

It is necessary to use the appropriate installation methods, which is typically the flow chamber, appropriate fixed flow rates e.g. around 25//min and PTFE or similar non-stick coatings for the sensors and to insulate well. A static mixer is also often recommended for the upstream section to ensure homogeneity of the fluid entering the chamber.

 

(examples of both the Emerson and Lemis sensors are found on www.viscoanalyser.com along with photos of some installations in Canada - the "bitumen" and Russia - heavy fuel oils)

 

The key to the success of this method also depends on the sensor accuracy and the accuracy of both viscosity and temperature measurements.

 

So it is quite feasible to heat or cool (or both, cooling for one sensor and heating for the other) but in the Canadian pipeline applications it has been usual for the first viscometer to measure at the sample original temperature i.e close to the pipeline temperature but subject to any ambient temperature effects on the sample line and use an electric heater between the viscometers to generate a sufficient temperature difference.

 

Note that in some applications such as where the oil is very waxy, it is necessary that both measurements are made above the wax point which is then sometimes above the reference temperature. So it is not necessary that either viscometer operates at the reference temperature though with more viscous product it is advisable to operate in the region of the reference temperature with one of the viscometers if possible.

The advantage of the method is that using indirect (calculation) methods, the temperatures do not have to be precisely regulated as is the case with the capillary viscometers and when the instalation is well designed and the viscometers properly specified, the on stream factor is very high and service, calibration and cleaning are then only needed in exceptional circumstances.

This is one of the reasons that until the advent of the Emerson (then Solartron) 7827 with accuracy, response time and methodology (callculation methods) along with a good understanding of process application installation practise, pipeline blending applications tended to rely on sampling and off line methods which necessarily also then involved "safe" blending i.e. use of excess distilate to ensure the density and viscosity were within limits.

 

These installations, including the civils, environmental enclosures etc. tend to be expensive but the pay back periods are very quick.

 

If you need further assistance, please follow the contact page on www.viscoanalyser.com

[JMW]

PS I have posted here on Cheresources two of the spreadsheets I created and linked to above.
However, I do also have a spreadsheet which uses the multi-curve method.
This is typically the calculation used on the Emerson 7827  7950 viscosity computer and in the LEMIS DC 52 ViscoAnaltytic multi-curve methods.
If you would like a copy please use the contact link as above.
 
There is a blend calculator now available online at Jiskoot http://www.jiskoot.c...ator/index.html but note that it does not allow for shrinkage.
 
Oh, and I have discovered that while everyone has been paying attention to the crude oil properties, they have been making assumptions about the distillates.
The refineries ship the distillates by tanker to the oilfields where it is blended with the oil so that it can be sent by pipeline to the refinery.
However, since the distillate is simply going round in circles and because it is only used for cutting back the oil, the quality can vary according to circumstances. Rather more than anyone suspected I gather.
So this may be a factor to consider when doing simulations and calculations.

Edited by JMW, 01 February 2013 - 07:11 AM.