Potpourri

Gasoline Blending Plus

Stuff that could interest a petroleum refinery engineer

An Interaction Paradox
Combined Tankage & Component Splitting
More on Splitting CAT Cracked
MMT Affects Octane Interaction
Why In-line Blending Works
Planning Experiments
Hire a Mentor?

Suppose a gasoline component blends linearly (zero interaction) with every other component in the gasoline pool. Even so, its blending value will vary with gasoline composition. This seeming paradox results from the dilution of interactions between other components. For example, suppose A has no interaction with either B or C, but B and C have a large positive interaction. Addition of A to a mixture of B and C will dilute that positive interaction and consequently the blending value of A will be below its measurement. It will have a negative bonus.

Combined Tankage & Component Splitting

Interaction blending equations are essential tools in evaluating the economics of combined tankage or of component splitting. Both keeping components separate and separating the light and heavy fractions of a component provide greater flexibility in blending. The flexibility must be balanced against the cost of splitting and the costs associated with separate tankage. A linear programming analysis of the advantage of the flexibility requires knowledge of the interaction between the components that are combined or between the light and heavy fractions of a component.

The comparison should be based on separate components with and without pooling. The positive Research octane coefficient between light and heavy reformate diminishes any advantage of splitting reformate. Typical Research and Motor octane differences from full-boiling components are as follows: (Reference 11)

	CAT CR		80 REF		100 REF
	R	M	R	M	R	M
25% LT	4.9	2.9	-6.2	-3.9	-16.3	-7.7
75% HY	-1.6	-1.2	-1.1	0.5	2.3	1.8

50% LT	1.3	0.9	-10.4	-7.7	-13.3	-7.7
50% HY	-1.2	-1.5	4.2	6.2	7.0	6.2

75% LT	-0.3	0.0	-7.6	-4.0	-6.3	-3.1
25% HY	0.9	-0.9	13.6	9.8	9.5	7.2

More on Splitting CAT Cracked

Some refiners split cat cracked to obtain the high octane light fraction for blending premium. A 25% light fraction is about four (R+M)/2 above the full boiling cat cracked.

Here are some estimated octanes of different fractions, if separated in a refinery, relative to the full-boiling cat cracked:

	Research		Motor
0 - 50%		1.3		0.9
50 - 100%		-1.2		-1.5

0 - 33%		3.1		1.9
33 - 67%		-2.8		-1.6
67 - 100%		-0.2		-1.2

0 - 25%		4.9		2.9
25 - 50%		-2.3		-1.1
50 - 75%		-3.3		-2.1
75 - 100%		0.9		-0.9

MMT Affects Octane Interactions

Oh yes, for those who have the option to use MMT for octane improvement, it is included in BLENOC. Typical MMT octane effects are included but specific effects can be entered.

Interaction coefficients are affected by MMT. Similar to TEL, the octane effect is approximately proportioned to the square root of concentration.

Why In-line Blending Works

Research octanes are obtained at low temperature and Motor octanes are obtained at high temperature. Gasolines have higher Research octane than Motor octane because they are more sensitive to temperature than are the reference fuels.

Because of the temperature sensitivity of gasoline, the increased temperature due to buildup of combustion chamber deposits causes their octane ratings to decrease with time. This effect can be compensated for by using direct match ratings with prototype fuel of about equal Sensitivity thereby providing accurate octane during in-line blending.

The short term variation of knock intensity with time is a source of octane error, but this is compensated by the large number of direct match comparisons made during in-line blending.

Surprising as it may seem, there is a small octane effect of the carburetor in a knock test engine. This effect is eliminated by using the same carburetor for both prototype and blended gasoline during in-line blending.

Planning Experiments

Here is a basic comparison of efficient experiment design with traditional experiment practice. Consider two independent variables (X1 and X2), that can be adjusted to find their effects on a dependent variable, (Y).

With traditional experiment practice, experiments might be carried out with six different tests as follows:

	1	2	3	4	5	6
X1	LOW	LOW	HIGH	HIGH	LOW	LOW
X2	LOW	LOW	LOW	LOW	HIGH	HIGH

Two estimates of the effect of X1 would be obtained comparing results for 3 and 4 with 1 and 2.

Two estimates of the effect of X2 would be obtained comparing results for 5 and 6 with 1 and 2.

Here is a more efficient experiment design with only four tests:

	1	2	3	4
X1	LOW	HIGH	LOW	HIGH
X2	LOW	LOW	HIGH	HIGH

You still get two estimates of the effect of X1 (Test 2 vs Test 1 and Test 4 vs Test 3) and you still get two estimates of the effect of X2 (Test 3 vs Test 1 and Test 4 vs Test 2)

There is of course much more to efficient experiment design conduct and analysis. Technical people who don't have expertise in this area might benefit from inexpensive training. Contact us if you think this might be appropriate in your case.

Hire a Mentor?

There is a large population of retired refinery people who possess valuable knowledge and would enjoy sharing it. With changes and consolidation in the industry, valuable know-how is being lost. Short term use of a retiree's services can be very worth while.

For example, suppose linear programming is not being used effectively, and a motivated employee is available but lacks the necessary know-how. Also suppose a highly skilled retiree is living in the area. The retiree and the younger employee could carry out the task together for say, a month to bring the employee up to speed.