Base oil Viscosity Mindbreaker 2

[Orakio]

Hello Everyone,

I have a situation which I have been pondering about for days, but I just can't come up with a decent solution or explanation for it.

Here goes:

Part of my company's activities is to offer advice to clients in regards to blending petroleum products to gain new products with desired qualities. The assignment I received last week has put me against an endless wall of problems though.
The two products used for blending are 2 naphtenic base oils, one with a kinematic viscosity of about 100 cSt (mm²/s)at 40°C and one with a kinematic viscosity of 4000(!)cSt at 40°C. the desired viscosity in this matter was 800 cSt.
At first this looked like an easy task, as most blends are easily calculated with the double logarythm function log(log(v)), which is linear for a certain product, or a certain temperature (viscosity blends are not linear). This has worked for most blends with minor exceptions to the rule. however, as the current blend was calculated to be 805 cSt at 40°C, with the calculated mass% quantities the actual result yielded ONLY 500 cSt, which is way too low and unlike anything I've ever seen.

My personal reflection: Perhaps there might be a problem with the fact that the viscosity indici (VI), which is the degree to which temperature changes change the viscosity of a product, were too far apart from eachother in value. THere is no way however, that this number can be used in any calculation. at least to my recollection.

So my question: Has anyone here come in contact with viscosity blending, and if you have, have you found a decent way to calculate the viscosity of a blend, using multiple parameters?

My hopes of this being solved are slim, but it's worth a shot.

Thank you for reading, and possibly giving it a thought.

 

[jmw]

I can suggest that you ask the question of Jiskoot at

http://www.jiskoot.com

who manufacture inline blending and sampling systems including for crudes, fuels and lubricants.
They may be able to offer some insights.

One of the most notable discoveries is the sheer magnitude of errors between the calculated (expected) viscosity and the measured viscosity (they use digital viscometers for feedback control of the blend).

The problem is to know if these errors are due to:
[ul]
[li]poor temperature measurement[/li]
[li]poor viscosity measurement[/li]
[li]poor accuracy of the blending calculation[/li]
[li]poor sampling technique for both components and blend [/li][/ul]
It is surprising how poor this often is and how poor some laboratory measurements are; I have witnessed samples collected in whatever container came to hand, watched the sample separate as it cooled and then watched the refinery engineers simply draw a sample from the top of the container.
Then take a single pass of the sample through a single capillary viscometer tube and quote the result with all the decimal places the calculator could produce).
or a combination of factors.

I have also seen some pretty poor temperature viscosity corrections applied using proprietary calculations, poorly understood.

I usually suggest using the fuel oil blend calculators which should work fine if used to find the viscosity at the same temperature as the data is input. For lubricants you may want to ask DNV PS for a copy of their fuel blend calculator because it does allow you to determine the VI which many such fuel oil calculators don't... they assume a typical VI for fuel oils.

However, you are correct that VI doesn't figure in the calculation of the blending ratio except if the blend is performed at some temperature other than the reference temperature; predicting the blend viscosity at some other temperature is a problem and for lubricant blending.

Viscosity trim is important as the calculated ratio needs to be very precise. Viscosity can change quite significantly even for small changes in ratio i.e. a problem with achieving the target ratio exactly.

Perhaps you can provide a bit more detail about the blending process and how it is controlled and the process conditions.

Alternatively you could use the ASTM D341 graphical solution for blending. Lubricants, synthetic or not, also follow the ATM D341 temperature viscosity relationship.

Incidentally, using the Shell blend calculator I get 63.20% of the 4000cst to obtain 800cst in the blend. This is confirmed by the DNV PS calculator. I couldn't use the Exxon calculator because it wants one viscosity at 40degC and the other at 50degC. (these calculators are all free from the oil companies Marine Fuels groups... DNV PS is a test service).

I'd be interested to know what your ratios where.

JMW

http://www.ViscoAnalyser.com

 

[25362]

A typical BP chart gives the same result as found by jmw, while the calculator on the following link

http://www.baseoils.shellglobalsolutions.com/calculations/blending_pop.asp

gives ~56.4% for the 4000 cS grade.

Using blending indices (which blend linearly) gives 62.1% by volume for the same grade.

I adhere to JMW's request for the ratios Orakio has found.

 

[Orakio]

Hello again chaps,

Since you're both interested in this case I've gathered the more detailed information that I have in my posession.

The subject and preferred blend result is a base oil with a viscosity of 800 cSt +/- 25 cSt.

The components used for this are base oils with a viscosity of:

1) 368.4 cSt @ 40°C & 19.05 cSt @ 100 °C
2) 4533 cSt @ 40°C & 40 cSt @ 100°C

I apologise for handing over or stating inaccurate or wrong information before. It was out of the top of my head, and it appeared to have been wrong.
AT least now, we work woth the same figures.

With my own calculations I became a ratio of 34.21 % V of product 2, and 65.79 % V of product 1.

This yielded 808 cSt theoreticly and a bloody 660 cSt after testing. I must say that I didn't make notice of the sampling procedure here, so it could be badly performed. in fact, after reblending we got a viscosity of 778.4 cSt with 55,5%V of product 1 and 44.5%V of product 2. this is within spec, but dubious since the results of lab-scale blends with samples I did monitor being sampled, provided something entirely different

the first lab-scale test was with the aforementioned %'s (55.5 & 44.5) and yielded 852.8 cSt at 40°C
The second one contained 60%V of product 1 and 40% of product 2 and yielded 728.4 cSt at 40°C

I have taken a look at your link, 25362, and noticed that they indeed require the viscosity at 2 different temperatures before a blend is even calculated. This is different from my own calculation where I use only the viscosity at one temperature. I'll have another look at the viscosity indici (VI) of both products and at the various temperatures, and see if there is any logical connection between them.

Thanks for the responses, keep them coming.

KInd regards

 

[25362]

Orakio, in reference to that link, if you try out different viscosities at 100 deg C (ie, different VIs) you'll find that the result at 40 deg C is the same, namely the VI doesn't affect the result at 40 deg C.

Besides, it appears you are blending volume-wise not mass-wise... How did you measure volumes, how was the blending done by the lab testers ? How did you estimate the blend viscosities ?

Would you kindly enlarge on these subjects ?

 

[jmw]

I can't think that VI has anything to do with the basic blending calculation but I'd better explain my thinking:

Let us start with the blend at 40degC.
The blend calculation will give the correct(?) ratio of 4533 to 368.4 for a blend to have a viscosity of 800cst at 40degC.(65.2%/34.8%)
However, this cannot predict the viscosity of the blend at any other temperature. This is because the VI of both oils may be significantly different.
For lubricants we know that two lubricants can have the same viscosity at one temperature but different viscosities at a different temperature.
Therefore, to calculate the viscosity of the blend at any other temperature we either need to know the VI of the blend or the viscosities of the components at the target temperature for the blend.

This means we need a viscosity index calculator such as the one 25362 suggests; except that I can't get ratio agreement between this Shell calculator and the Shell fuel blend calculator (the latter agrees with the DNV PS calculator).

This one gives a ratio of 69%/31% to give 802 cst at 40degC and at 100degC, 24cst, compared to 24.27cst at 100cst for the shell blend calculator and the DNV PS gives 24.6cst at a 65.5%/34.5% ratio.
These were calculated by plugging in the viscosities at 100degC (there is a sad lack of resolution in all these calculatros whcih makes mathematcially precise comparative determinations difficult but may reflect the limitations of the theory)

Why does the online calculator vary so much in the ratio from the fuel blend calculators? I don't know.
The fuel blend calculators are giving me the mass ratios and it might be that the online calculator is using volume ratios but it doesn't say so. I'd like to find out some detail about their underlying calculations or an alternative calculator to use....

This one might be better:

http://www.chevron.com/products/oronite/site_demo/calculators_tables/

though you need an account to access this calculator which i don't have. I didn't find any others online.

This basically suggests the blend at 65.2:34.8 having a viscosity of 800-802 at 40degC and 24-24.7 at 100degC.
I can now put this data into the ASTM D341 spreadsheet (download from

http://www.viscoanalyser.com)

and I can now get the viscosity at any other temperature for the blend.

Interestingly, when I plug in 660cst as a target viscosity i get atemperature of 42.21 degC.... and that is suggestive since this is measured in the very steep part of the curve.
This prompts me to refer you back to temperature measurement errors. A shift from 800 cst to 660cst as a result of only 2.21Cdeg difference is worth investigating.

This is all fine but how does it relate to your experience?

JMW

http://www.ViscoAnalyser.com

 

[Orakio]

In reply to 25362, Indeed, I was too quick to assume this calculator used multiple viscosities on various temperatures.

As to the blending procedure, we don't use volume%. I just used volume% since your earlier reply quoted one

Blending was done with mass%. Samples were heated to approx. 60°C and weighed carefully. Our measurements are performed with calibrated thermometers and repeated measurements on a calibrated and validated Herzog auto-viscosimeter. I'm quite confident that measurements are thus correct, as are the lab-scale blends that were made.

However, the result I gave to you, 660cSt, was performed under the circumstances where the blend was made by our forwarding guy. A blend incorrectly made would have a different Viscosity index, and that might explain the difference in temperature you noticed.

Alas, I think I'll need to perform more tests i order to get a clear view on the current blend. Perhaps I'll be able to obtain the results as soon as next week, but I won't pin myself to a date.

As to my own way to calculate the viscosities, I in fact used ASTM D341, jmw, in order to obtain an empirical equation. Somewhere in the method a mathematical equation is given which contains log(log(Cst+0.7) = A-B.log(T)
It also says that everything on the right-hand side of the equation is a constant, as long as the temperature is a constant.
The reason I went to research this was since the first blending problems showed themselves with residual fuel oil.
now, back to the constant. it would dictate that the double logarythm could be used as a constant. (hell knows why the +0.7 is in there, I used it none the less).

With this knowledge I simply blended these constants in regards to mass%. so you'd have

(log(log(cSt1+0.7))*m1+(log(log(cSt2+0.7))*(1-m1) = x
and
power(power(x)))-0.7 = cStblend

unlikely to be a success? seemed to me as well it would be like that, but to my great surprise it worked for residual fuel oil, and later for low viscosity base oil as well!

Never the less, in the case that the high viscosity is used in a blend, it messes up this theoretical approach entirely. As I said, I'll try and obtain more of the sample and try various mass% ratio's.

My belief that the viscosity index is in fact a factor which plays a role here has only ben strengthened after a quick calculation of the viscosity indici of the various products using ASTM D2270:

The product with 368.4 Cst returned 35
Other oils with lover viscosities, and residual fuel oil all gave VI's in the range between -10 and 45.
Now, the product with 4533 cSt smashed all records, with an incredibly low Viscosity index of -260!!!
How it relates to blending properties remains a mistery as of yet.

I as well would like to know the mechanics behind the calculators like the shell one. Alas, I doubt they'd be willing to share it with the commoners.

Hopefully this makes my situation a bit more understandable for you both (and any other lurker of this thread)

 

[25362]

I've the feeling that blending at 60 deg C when one of the oils has a viscosity of ~ 570 cS and the other ~ 100 cS may not result in a truely homogeneous mixture. Making the blend at ~100 deg C might give more consistent results. Have you checked blends made at two widely different temperatures ?

 

[Orakio]

I haven't done that no. But I'm only saying that the samples were taken and mixed at 60 degrees. Once the desired weight was added, is was assured that the mixture was homogenous. This is relatively easy with base oil, since they are all clear and bright.
I have no doubt that the viscoity was measured according to the written procedures either. Our chemists act with great care and responsability.

The real problem is finding the true mechanics between blending kinematic viscosities. My model is clearly not working. neither is the shell one. there must be a factor we are not taking into account...

 

[jmw]

Using the ASTM D341 spreadsheet mentioned, I calculated the viscosities of the base oils at 60degC as 103.13cst and 544.51cst respectively.

Using these values in the calcution and the same ratio as found for the 40degC calculation I get a blend viscosity of 174cst (at 65%, not 65.2%... resolution problem again, it wouldn't let me enter 65.2%).
From the previous calculations I get the blend viscosity of 800cst at 40 and 24cst at 100degC. Putting this into the ASTM D341 spreadsheet and calculating the viscosity at 60degC I get 173.04cst....

In other words, all the calculations appear consistent if you don't use the shell online calculator (where the viscosity values agree even if the ratio does not).

In my previou post I suggested that 660cst results from a 2Cdeg error... but this doesn't have to be the error associated with the blend sample, you can share it among the viscosities of the base oils and the blend... or you can look at the HERZOG accuracy, which I believe is around 0.01degC on temperature but 0.25% of reading on viscosity.
However, the other possible source of error is in the mass of each component blended.

Representative sampling is another area of concern.

25362 has suggested that the mix may not be homogenous. This is a common problem and you haven't yet given us details of the mixing system or the sampler.

Are you batch blending? i.e. adding the calculated masses of each oil into the tank then mixing till homogenous?

Though it is far less sensitive as an indicator of quality despite the better accuracy of measurement, you might also report the densities of the base oils and the blend.

If you measure density at 15degC of the blend, and you know the density at 15degC of each component you can calculate the actual ratio quite accuractely and cross check your mass measurement (using mass meters or volume meters corrected for the density of the oils at 60degC?)

Are you doing a TMB (Top Midle and Bottom) sample?
Are you taking multiple samples?
Is your Herzog a single capillary or multiple capillaries?
(running the same sample through several capillaries is necessary to obtain a good sample average (discard highest and lowest and average the rest). Repeat with different samples.





JMW

http://www.ViscoAnalyser.com

 

[25362]

Thermal history may affect oils with additives.
However, when blending Newtonian fluids as base oils are supposed to be, inhomogeinity in (macro) mixing has been shown to be a function of a series of dimensionless groups, among them: log([η]₁[÷][η]₂) and [Δ][ρ][÷][ρ], Reynolds number, Archimedes number, Schmidt number, etc., for constant mixer geometry and stirrer rotation.

Where [η] is viscosity and [ρ] is density.

That's the reason for my considering the importance of temperature to achieve homogeinity.

 

[jmw]

Thanks for that post 25362... it's great to learn new things.
Is there a web site where I can learn some more about this?

JMW

http://www.ViscoAnalyser.com

 

[25362]

Sorry, jmw, no website, only books on fluid dynamics dealing with mixing. Ullmann's Encyclopedia of Industrial Chemistry, Vol B4, has a short expos[é] on the subject that might interest you.