PerkinElmer Headspace TOGA

[patt8rson]

Hi guys,

hopefully I can get some feedback on a GC issue I have been having for some time. I operate a PerkinElmer Claus 580 GC with Turbomatrix Headspace sampler for TOGA (Transformer Oil Gas Analysis). Following ASTM D3612 I have calibrated the GC using calibration gas purged vials at 3 different levels. I have found at gas that the GC is repeatable. However when testing gas-in-oil samples I have found that the unit reads typically 20-50% lower (depending on gas) than expected. The method used in the Turbomatrix Headspace sampler is the same for both gas and gas-in-oil samples except the thermostat time as the gas samples obviously do not need to reach an equilibrium. Typical results are shown below, calculated (theoretical)

H2 125 (182)
C2H2 132 (185)
C2H4 149 (180)
C2H6 151 (185)
CH4 136 (185)
CO 118 (183)

Has anyone found difficulties calibrating on gas and testing gas-in-oil? If so what were your solutions. I am using the correct equations, where the GC reports the detected headspace concentration Cg and I am then using the equation below to calculate the gas-in-oil CL0.

CL0 = Cg x (Kgas + Volumetric ratio)

Any suggestions or insights would be great!

[Peter Apps]

How accurately do you know the volumetric ratio ? Presumably you can accurately measure your sample volume, but how accurately do you know the internal volume of the headspace vials ?

Where do you get the values for K, how sensitive are they to temperature, and how accurate is the temperature control ?

Peter

[patt8rson]

Hi

In reply I have followed the recommendations in the ASTM standard. So I originally checked the total volume of the sample vials using deionised water at room temperature. Then I calculated the density of the oil which I regularly use. So when preparing my sample vials, I weigh the sample mass and divide by the density to get sample volume and headspace volume. The Ki values in use are appropriate for oil at 70 degrees centigrade, which is the constant well controlled thermostat temperature in the turbomatrix headspace sampler. The sample is heated and shaken for 30 minutes.

[Peter Apps]

Another thing to consider: the gas phase volume differs between your standards and your samples, this might affect the transfer of headspace to the column. How big are your vials and how much sample do you use ?

Peter

[patt8rson]

The gas phase volume is different from calibration and test samples, however this should be taken account for in the calculation of Cg to CL0 with the volumetric ratio, as mentioned previously. During test the sample is heated and shaken for 30 minutes to achieve equlibriation before being presurised with Argon, this argon flow is then turned off and the headspace sample is transfer to the 1ml injection loop, therefore this should be completely filled. This injection loop is then flushed onto the GC columns.

My issue is that there appears to be an issue with the gas extraction in the turbomatrix headspace sampler and I posted online in the hopes that someone has a similar system and could suggest improvements.

[Peter Apps]

The calculation of equilibrium concentrations takes into account the effect of liquid and gas phase volumes on the distribution of the gasses, but cannot account for possible impacts of differences in gas phase volume on gas flows, pressurization, loop fill times and residual pressures in the loop. For example, if your vial is half full with liquid there is only half as much excess gas volume available to purge and fill the loop. This could easily cause the kind of bias that you are seeing.

Headspace analysis is not very robust to conditions, and official methods are not usually optimised for particular instruments so you might well need to tweak conditions to get good accuracy. Needless to say the tweaking needs to be based on an understanding of what is going wrong, and the first step in that directions is a clear description of your analytical conditions.

Peter

[rb6banjo]

Would you supply some some "typical" values for Kgas and the volume ratio?

[patt8rson]

Typical partition coefficient values (Ki) are taken from ASTM D3612, suitable for Voltesso 35 oil at 70 degrees, which is similar to the oil and identical to the sample thermostat temperature I currently use.

H2 - 0.074
O2 – 0.17
N2 – 0.11
CH4 – 0.44
CO – 0.12
CO2 – 1.02
C2H2 – 0.93
C2H4 – 1.47
C2H6 – 2.09

I typically use a sample vial of total volume 22ml, and to this add the recommended amount of 15mls of sample oil for analysis.

[Peter Apps]

So your gas phase standards contain 22 ml of free gas volume, and your samples less than a third of that (7 ml).

Just as a trouble shooting exercise can you try a run with 7 ml of sample and see if it brings the sample peaks closer to the expected size. Or reduce the free volume in the standard vials by adding 15 ml (actual volume, not bulk volume) of glass beads or something.

With that much sample you might also have a problem with equilibration.

Peter

[rb6banjo]

Thanks for the constants. I should have asked also that when you say the headspace/GC is repeatable, what does that mean? What RSD% do you measured on a repeated analysis of the same sample?

I'm pretty much in agreement with Peter Apps on this. The biggest source of error, no matter how you define the parameters, is in the measurement of the gaseous concentrations (assuming you're getting as close to the required phase ratio as reasonable care dictates). If you're having trouble with this analysis it has to be because of imprecise measurement of the gas-phase concentration. Proper equilibration of the sample is the first place I'd check.

[patt8rson]

I have previously tested the same oil sample, using sample volumes of 5, 10 and 15ml. Ideally it would be expected that once the result from the GC (Cg) was taken and used in the relevant equation previously mentioned, the final gas-in-oil value should be the same. This was not the case with the 5ml oil sample showing the highest final gas-in-oil value. I believe this could be due to the effect of pressure in the headspace, meaning the Ki values which I use are not appropriate for the system set up and the gas is it not reaching the equilibrium expected.

I do believe that your suggestion of using glass beads so that the gas volume during calibration is similar to that seen during a test is a sensible test and one I had not previously thought to do, so thank you for that.

Relative Standard deviation of the results during same oil tests is typical 5% or better, not including atmospheric gases.

[Peter Apps]

Since the free volume between randomly packed spheres is about 30% adding glass beads just to the neck of the vial will give a sensible approximation to the gas volume above a 15 ml sample.

Have you tried the effect of equilibriation time with large sample volumes ?

Peter

[patt8rson]

Thanks for the tip regarding the glass beads, I was possibly going to weigh them out to obtain an accurate volume.

Unfortunately I have had to recitify a recent mechanical fault which has prevented me running certain tests, for example increasing the thermostat time to investigate whether the equilibirum has been reached.

However I am following recommendations from recognised standards and wanted to maintain that traceability and yet our set-up has shown these 'low' values. This has been quite frustratiing, but I wonder whether this is a common issue, and whether it has been more noticeable as I have attempted to calibrate using gas standards and test gas-in-oil samples.

[patt8rson]

Following on from your suggestions above, I have extended the thermostat time, i.e. the time the sample is heated and shaken to achieve equilibrium, and have found that this had no effect. Therefore I can assume that a thermostat time of 30mins, as recommended in the standards is sufficient to achieve equilibrium between the gas-in-oil and gas-in-headspace.

To ascertain the partition coefficients appropriate for the equipment set up I used the recommended technique of test numerous vials at different volumetric ratios. Therefore I tested samples at 6, 8, 10, 12 and 14ml oil volumes in a 22ml sample vial. I plotted the reciprocal of the area under the peak curve on the y-axis and the volumetric ratio on the x-axis. I then plotted a linear trend line and calculated the partition coefficient using the gradient and intercept. This method is recommended by the standards, however I don’t have a massive amount of faith in it and believe it still masks the underlying issue.

As stated before has anyone seen similar problems when calibrating with gas and then testing gas-in-oil samples?

[Peter Apps]

Using different fill volumes to generate data to calculate partition coefficients is vulnerable to the same potential problem as your basic set-up of comparing gas standards from vials containing 22 ml of gas with samples containing only 7 ml of gas. The different gas volumes generate different gas flows and pressures during transfer of sample to column, and that is why you get anomalous results.

Try putting only 1 ml of oil into a vial, and see if you get closer to the expected results than you do with 15 ml of oil.

Peter