The surprising and significant errors in reading a centrifuge tube

Posted by Ian Burgess on Aug 8, 2017 6:01:56 PM

The centrifuge test to determine the sediment and water (S&W) content in crude oil is one of the most frequently run tests in the oil and gas industry. S&W measurements are done every time a shipment of oil changes hands so that the buyer only pays for the oil volume. Even small errors can add up to significant amounts of money so accuracy is important. For example, a 0.1% measurement error on S&W can cause a producer (or receiver) to lose about $10,000 per year for a single well1. That is nearly $2,000,000 per year for a larger 100,000 bpd midstream operator!

Maintaining high accuracy makes centrifuge tests a pain for field operators to run. Protocols such as ASTM D4007 and D96 outline the many steps an operator must take to ensure the water and sediment are fully separated from the oil. These steps include: a) taking a representative sample, b) preparing and adding solvent, c) adding demulsifier, d) pre-heating the centrifuge tube before spinning, e) spinning the tube for a long time at high speed.

But what if large errors come from the most trivial step, when the operator looks at the tube and records the level of the oil-water interface?

This suggestion might sound silly at first - or worse, insulting to the highly trained operators working in the field - but there are actually several reasons why reading a centrifuge tube accurately is not a trivial task.

The classification of 'interface'

Most centrifuge tubes show more than just two layers after spinning. There is often a significant layer apparent at the interface between the water and oil phases (Figure 1). This layer may contain any number of components, ranging from a salable wax to an unsellable product such as asphaltenes. How this interface is accounted for is often left to the discretion of the field operators. They may do quick visual tests on the interfacial layer to assess what is in it. However it is common practice to report only one final value for BS&W, a more detailed record of how this number was obtained is often not accessible for anyone else to verify later.


Figure 1

Figure 1 - Two centrifuge tubes containing significant interface layers between the oil and water layers after spinning, with examples of very dark (left) and light (right) interfaces.

Leveling errors are surprisingly large

Figure 2 shows two pictures of the same centrifuge tube after spinning. On the left, the camera is exactly level with the meniscus. On the right, the camera is at a very small angle (4º) above the level of the oil-water interface. Notice that this causes a significant change in the apparent reading.

Figure 2-972739-edited-034428-edited

Figure 2 - Two images of the same centrifuge tube, imaged with a perfectly level camera to the meniscus (left) and with one 4 degrees above level (right). Even very small misalignments lead to large discrepancies in the apparent readings.

Since the standard 100 mL centrifuge tubes are fairly wide, the tick marks on the outside used for measurement are actually significantly closer to the eye than the peak of the meniscus, which lies in the center of the tube. As a result, the apparent S&W reading appears higher than it should be if the eye is slightly above level, as shown in the diagram below on the left. It is fairly easy to ensure that the eye is perfectly level when both liquids are transparent by lining up the interface at the front and back walls. However, this alignment method cannot be used with most crude oils, as they are completely opaque. With opaque liquids, you can correct for alignment errors by lining up the center of the meniscus with the tick marks on the side of the tube. This readout correction method is shown in Figure 3. The images in the center and on the right show the same tube as in the previous example above, and at the same viewing angles (0º and 4º), but rotated so that the tick marks are on one side. A consistent reading (0.3%) is obtained by drawing a straight line from the center of the meniscus to the tick marks at the edge of the tube.


Figure 3

Figure 3 - Schematic (left) showing how slightly non-level readings cause significant readout errors. This readout error can be compensated by aligning the meniscus with the tick marks on the side of the tube instead of the center (center, right). However, this readout method is challenging to manually, and is much more apparent on electronically saved images.

To see how much subjective error would affect readings in the field, we ran a small-scale study of 8 users trained on the ASTM D4007 method interpreting three centrifuge tubes. All tubes had very simple meniscus with no interfacial layer like those shown in Figure 1. All were prepared by adding a small amount of distilled water to clean samples of crude oil and condensate before running the centrifuge method. The results, shown in Figure 4, confirm significant errors arise in the manual interpretation. The error range of the manual readouts for each tube was between 0.1% and 0.15%, corresponding to losses of up to $1.8M or $2.7M respectively at a throughput volume of 100,000 barrels per day (Figure 4, left). This error corresponds to a potential significant cost to producers and midstream companies. Furthermore, we measured this readout error using very clean samples of oil that had no discernible interface layer. Had our study included samples with significant interfacial layers, the errors would have likely been much larger.


Figure 4

Figure 4 - Left: average readout value and standard readout error (error bars) for three centrifuge tubes, from a sample of 8 trained users. Right: Maximum annual losses resulting from 0.1% measurement error.

1 Assumes a well output of ~500bpd and an oil price of $50 per barrel.

How do manual processes and operator error affect your company’s bottom line?

Get a free consultation



Topics: BS&W

Interesting Topics in Crude Quality and Measurement

A collection of our industry learnings and insights.

Subscribe to Email Updates

Posts by Topic