Fuel blend calculation errors for predicting viscosity?

[jmw]

The production of HFO blending residues with distillates results in a significant number of viscosity errors reported; Lintech reports that over 7% of fuels sampled are off spec for viscosity.

The fuel blend calculation determines the required ratio of components to produce a blend with the desired properties.

Most properties are additive e.g. density, %mass sulphur, sodium vanadium etc.

Viscosity is not, it has a logarythmic relationship between the ratio and the viscosities of the components.

If the ratio of the components and the viscosity of the components are accurately known, how well should the calculated viscosity compare with the measured viscosity?

Or, is the variation in viscosity a function of its sensitivity to small errors in ratio (that are not significant when applied to additive properties), or small errors in the viscosity of the components or is it because the model for predicting the viscosity is imperfect?


JMW

http://www.ViscoAnalyser.com

 

[25362]

Viscosity blends in the petroleum field are usually predicted using Maxwell's blending indices, based on vol% and the components' kinematic viscosities. Errors can be introduced at all steps of the blending process. Blending indices are also used for estimating blends of pour points and flash points which are not additive either.

Anyway, D.S.J. Jones' Elements of Petroleum Processing (Wiley) brings an example of a blend of six components using the above-mentioned blending indices and compares the estimated 2.65 cS with actual plant's 2.7 cS !!
I'd say this is pretty accurate.

 

[jmw]

Thanks 25362,
I was rather suspecting this might be the answer since I have come to depend on some of these relationships for their accuracy.
ASTM D341, for example, has evolved over pretty nearly 100 years (is that right?) and is a fine model of the viscosity temperature behaviour.
With anything as core as blend properties I had to hope that the same was true.


JMW

http://www.ViscoAnalyser.com