When two batches of oil are processed one after the other, the test results for the second are not exactly the same as for the first, and it is common for the machine to be suspected on that basis. In fact, as long as the incoming state of the oil or the sampling conditions differ, the results differ. Taking the differences apart shows which of them are normal variation and which ones call for a check of the machine.
The difference comes from five stages
Between the oil entering the machine and the test report coming out, five stages can introduce a difference, and if any one of them changes the result changes with it.
The first is the incoming oil. If the two batches come from different storage tanks, their starting water content, gas content and particle load are already different, and the machine is working from two different starting points.
The second is the sampling. A different sampling point, a sample taken while running or after standing, a different oil temperature at the moment of sampling, all of them change the reading. Oil taken while running carries the temperature it was just heated to and the state it had just come out of the separation section in; after it stands, the bubbles near the surface disperse and the dissolved gas reaches a new balance, and the two states do not measure the same quantity.
The third is the oil temperature. Once the oil temperature changes, the viscosity changes, the atomisation state changes with it, and the degassing performance differs from one batch to the next. The distribution of the water content between dissolved water and emulsified water changes with temperature in the same way.
The fourth is the process settings. If the set temperature, the vacuum level and the running time are not the same on both occasions, the treatment is not the same either.
The fifth is the wait after sampling. Sending the sample to the laboratory without enough settling time carries bubbles into it, and the breakdown voltage comes out lower.
Which differences are normal and which call for a machine check
| How the difference appears | More likely to come from | Is a machine check needed |
|---|---|---|
| The two batches arrived in different states | The incoming batch | No |
| The sampling conditions were not the same | The sampling routine | No, make the routine the same first |
| The set temperature and the vacuum level differ | The operating settings | No, make the settings the same first |
| The incoming oil and the conditions are the same, yet the results keep drifting down | The state of the machine | Yes, work through the machine |
| Repeating the same batch of oil gives the same result | A stable process | No |
The dividing line is clear: only when the incoming oil, the sampling and the settings have been fixed and the batch results still drift in one direction does the machine come into question. If instead one of those three conditions changes each time, then however large the difference is, it says nothing about the machine.
Turning batch data into a record that can be used
For a trend to appear, the record has to contain the following: the batch number and the source of the incoming oil, the sampling date and the sampling point, the oil temperature at sampling, the set temperature and the actual vacuum level during the run, and the time from feeding the oil to taking the sample.
Once those are recorded, the difference between two batches has an explanation. If one batch was processed for longer and the breakdown voltage came out higher, the difference has a clear owner; if the processing time was the same and the results differ, the search moves to the sampling and to the incoming oil.
One point is worth emphasising: the result of processing the same oil again should be stable. That is a benchmark that checks the machine itself and depends on no outside condition, which makes it the most sensitive one available.
Most of the difference between batches comes from the incoming oil, the sampling and the settings; only when all three are fixed is it the machine's turn to explain the difference.
