Field study of LNG probe-vaporizers
A field study about quantification of precision, a comparison of sampler performance
Field study abstract
For comparison of two probe-vaporizer systems for LNG sampling first the main causes of bad sampling are discussed. Pre- and partial vaporization is a heat related issue that will cause part of the liquid sample to boil creating a two phase flow. Due to the different vapor pressures of the components a two phase flow can create concentration differences. The expected influence on the sample is the loss of homogeneity and compositional integrity, both important sample properties.
The experimental details are discussed, the used equipment and the gathering of data. Both tested systems are commercially available, as is the accumulator and online GC setup. Only the data from the online GC is gathered for the datasets.
Statistical analysis of the data is done with a Lilliefors, Wilcoxon, and Levene’s test. These are robust methods to determine if the data is normal, if measured means are equal, and if the variances are equal, respectively. System 1 produces a more stable trend with a higher precision. System 2 measures significantly higher concentrations on lighter components and lower concentrations on heavier components, indicating partial vaporization might be an issue.
Lastly, the experimental setup is discussed from a research perspective. Increased randomization or a matched pair comparison would be ideal if possible in a field test. To overcome the assumption of a stable process the comparison needs to be done at a test facility with specialized equipment. Such a testing facility is currently under construction for use in further studies.
Title and Author Information
Christiaan Mul BSc.
Project specialist analysis
LNG-sampling, ABC+ sampling and analysis, QCM-moisture measurements, and product development. It is good to combine the theoretical knowledge from my education as a chemist with the practical experience gained during my time at ASaP. Having a broad perspective will insure the developed products and services have excelling quality.
Started – Aug 2012
R&D, start up, commissioning, repair, and maintenance of analyzer systems.
University of Amsterdam, Vrije universiteit Amsterdam. Bachelor of science, Chemistry – Analytical sciences.
Hans-Peter Visser Ing.
Technical Advisor Analyzers and Systems
I’m part of the management, sales and R&D team. I’m focusing on what I like to do: the combination of technique and customer interaction.
My goal is to be the best analytical partner for our customers with ASaP (10+) . I started as mechanical engineer meanwhile I’ve studied chemical engineering/ process technology. I’ve worked for several Dutch system integrators in the past before I started ASaP.
Started 1989, founded ASaP in 2000.
Engineering, production, service, R&D of analyzers and systems.
Hogeschool van Amsterdam
Chemical engineering/ process technology.
Field study introduction
For the trade of Liquefied Natural Gas (LNG) the price is based upon the energy content in the LNG1. The accuracy of the energy content determination is therefore of the utmost importance. Small differences in the measured concentrations are multiplied by the volume of the transfer and could be of significant value. In this article we will discuss the common pitfalls in LNG sampling for analysis as introduction in the subject. The field study of LNG probe-vaporizers is based upon the experimental comparison of two vaporizing probes in a true field situation.
Sampling LNG has more than a few challenges, among the most important are pre-vaporization and partial vaporization. Both of these challenges are closely connected to the cryogenic nature of the sample. Described in ISO 8943:20072, the sampled LNG needs to be in a liquid sub-cooled state, that is at a lower temperature than the boiling point at that pressure. The risk is that due to the very low boiling point of the LNG any heat transfer into the system can readily cause vaporization.
LNG is a mixture of components therefore we do not have a precise boiling point. Figure 1 depicts a schematic representation of a bubble point curve for a two components mixture. The bubble point curve of a mixture shows the temperature at which the liquid starts to boil. Consequently the components vapor pressures determine the concentration ratios in the gas relative to the liquid. Based upon this principle is fractional distillation3. In the field this could happen in case a sample line is poorly isolated and the liquid is heated to the bubble point curve and vaporized. This temperature is also represented on the dew point curve, which means the vapor could condense in a different composition depending on the vapor pressures. For LNG we would expect higher methane readings in the gas than in the liquid, for the vapor pressure of methane>ethane>propane4–6.
The heat causing the LNG to pre-vaporize could come from the vaporizer upstream that heats up the tubing, poor isolation, or excess ambient heat. The heat influx is not necessarily enough to fully vaporize the LNG, which would result in any variation of two phase flow7–9 and an inhomogeneous sample. In my experience heat influx is mainly caused by bad isolation on a sample line, easily recognized by ice formation10. This phenomena influences both the composition and the homogeneity of the sample. One of the solutions provided to address this problem is the probe-vaporizer setup. These kind of vaporizers are directly connected to the probe which is installed in the subcooled process liquid. An example is shown in Figure 2, noticeably there is no sample line for the liquid reducing the opportunity for pre-vaporization.
Partial vaporization takes place inside the vaporizer and has the same effect as pre-vaporization. Incomplete boiling generates a gas with a different composition than the liquid. This could be caused by a bad vaporizer design, insufficient heat capacity, dead volume, or incorrect flow settings. During the vaporization of methane the volume expands by a factor 232 (1.013 bar at boiling point), if then immediately heated to 288K it expands by a factor 621. Pressure regulators reduce the flow depending on the output pressure. The increase in volume creates pressure that could stop the flow in the vaporizing pressure regulator, this situation maintains partial vaporization and heat flux upstream.
The analytical result of pre- and partial vaporization is a loss in sample homogeneity and representability, expressed as precision and accuracy respectively. The influence on the composition would lead to an increased measured concentration of components with a higher vapor pressure, and a decreased measured concentration of components with a lower vapor pressure. Lastly the calculated results, such as the Gross Heating Value (GHV), would change with the composition. GHV would decrease in a system where pre- or partial vaporization is present.
Field study experimental details
In this field study of LNG probe-vaporizers two commercially available LNG probe-vaporizing systems are compared based on the direct analysis data. For an effective comparison the vaporizers are the only variables in the sampling system. Both systems use the same accumulators, sample lines after vaporization, and the same Gas Chromatograph (GC). In Figure 3 the used installation is shown, both probe-vaporizers were installed on the same impact probe on the LNG-transfer line.
The experiment tests the stability of the systems during operation. All valves were fixed in position to prevent any interference. Unfortunately not all data could be used due to commissioning works. Four datasets were selected, 2 for each sampler, and combined assuming the sample to be a constant factor. The data is plotted for visual inspection, also normality, means, and deviation were calculated.
The data was collected during a real LNG ship to ship transfer in the port of Dubai. Data was only used on condition no works were being executed, parcel was loading, and cooling down was finished. Data for measured components Carbondioxide, iButane, nButane, NeoPentane, iPentane, nPentene, and Hexane were negligible. This data was computed but did not render significant results due to the low concentrations and were therefore omitted. The GHV was computed by the GC, the influence of these components on the GHV is included in the results.
Field study results and discussion
Figure 4 to Figure 7 show the plotted results of the stability runs for both systems for methane, ethane, propane and the GHV. Optical inspection reveals the signal is more ‘calm’ for the system 1 compared to system 2. The greater difference between consecutive measurements can be expressed as the line length. In system 2 the signal line length is 4.2, 4.1, 1.6, 0.7, and 3.8 times longer for methane, ethane, propane, nitrogen, and the GHV respectively. Another visual indicator is the reported concentration ranges, (highest – lowest) which are greater for system 2.
Figure 4 Combined methane signal for system 1 (left) and system 2 (right)
Figure 5 Combined ethane signal for system 1 (left) and system 2 (right)
Figure 6 Combined propane signal for system 1 (left) and system 2 (right)
Figure 7 Combined Gross Heating Value for system 1 (left) and system 2 (right)
Normality testing was done with a Lilliefors test since the true mean (µ) and deviation (σ) were unknown11. The results are displayed in Table 1. The data found using system 2 was found to be not normally distributed on the majority of measured parameters. It was decided to continue with the data assuming non-normality for all datasets. A normal distribution is expected to be found in the datasets by virtue of the following arguments: the sample should be homogeneous; fully mixed and of constant quality. The systematic error in the GC is normally distributed and in theory it should be possible to find this distribution in the data. With both systems non-normally distributed results are found, although system 1 performs better on methane and ethane.
The consequence of non-normal distributions is that the standard deviation (s) and mean () cannot be compared with the use of parametric statistics such as: F-, Chi-, or T-tests. Though they can be compared with the use of non-parametric statistics with less statistical power. In this experiment the means were compared with the use of a Wilcoxon sign rank test12, and variance with the use of Levene’s robust test for equality of variances13. These data is shown in Table 2.
|System 1||System 2|
|Dataset 1 (n=52)||Dataset 2 (n=47)||Dataset 1 (n=47)||Dataset 2 (n=52)|
|Methane||Normal||Normal||Not normal||Not normal|
|Ethane||Normal||Normal||Not normal||Not normal|
|Propane||Not normal||Normal||Not normal||Not normal|
|Nitrogen||Not normal||Not normal||Not normal||Normal|
|Gross Heating Value||Normal||Not normal||Not normal||Not normal|
Table 1 Results of Lilliefors test for normality
|System 1 (n=99)||System 2 (n=99)||Means compared||Variance compared|
|Mean of measurements||Mean absolute deviation||Mean of measurements||Mean absolute deviation||Wilcoxon α=0.05||Levene’s α=0.05|
|Gross Heating Value (MJ/nM3)||39.520||0.018||39.509||0.029||Different||Different|
Table 2 Means with deviation and Wilcoxon and Levene’s test results for system 1 and 2
In Table 2 the average measured concentrations on both systems is shown. When compared in a Wilcoxon test these means are found to be significantly different from each other. It thus makes sense to note the reported value of ethane, propane, and the GHV is higher on system 1, whereas methane and nitrogen have a higher value on system 2.
The precision of the mean, expressed as the mean absolute deviation, is also shown in Table 2. The Levene’s test reveals that for methane, ethane, nitrogen, and the GHV the deviation is significantly different. The concentration measurement for propane is not found to be significantly different. It can be said that measurements done on system 1 are equally or more precise than those on system 2.
When these results are compared with our expectations about the influence of partial vaporization on the analysis results we find the following. On system 2 higher concentrations of lighter components are found, nitrogen and methane, accompanied with a lower GHV. Also a higher deviation compared to system 1 and non-normal distributions can be found in the data. These might be indications that partial vaporization is occurring in system 2.
Other reasons for differences in the mean can be found in the assumption the sample quality is equally distributed over the complete loading. It is likely all liquid hydrocarbons are fully mixed with each other, though the immense volume of an LNG-cargo ship might have some adverse and unforeseen effects. Process flow and temperature, ambient temperature, front or back position in the process bypass, or whether it rains or not might have an influence on the performance of the systems. To be able to compare the data as matched pairs the measurements should be done at the same time while continuously changing between the systems. At the time of the experiment this was impossible, in the future such an experimental setup could improve the data.
The main issue with this comparison is that we can only indicate for one of the systems whether or not partial evaporation might be happening. There is no reference material to which we can measure the functionality of the better system. To test these setups independent from each other a testing facility is needed, as described in the NEN-EN 12838 standard14. Such testing facility is currently being built in the Netherlands for research purposes. At this test site the probe-vaporizers will all be measured against a common denominator.
Field study references
(1) Groupe International des Importateurs de Gaz Naturel Liquéfié. LNG Custody Transfer Handbook (second edition); 2001.
(2) Normcommissie 310 066 “Debiet- en hoeveelheidsmeting.” NEN-ISO 8943; 2007; Vol. maart.
(3) Atkins, P.; Jones, L. Chemical Principles; W. H. Freeman, 2008.
(4) Air Liquide. Methane Vapor Pressure http://encyclopedia.airliquide.com/images_encyclopedie/VaporPressureGraph/Methane_Vapor_Pressure.GIF (accessed Apr 20, 2016).
(5) Air Liquide. Ethane Vapor Pressure http://encyclopedia.airliquide.com/images_encyclopedie/VaporPressureGraph/Ethane_Vapor_Pressure.GIF (accessed Apr 20, 2016).
(6) Air Liquide. Propane Vapor Pressure http://encyclopedia.airliquide.com/images_encyclopedie/VaporPressureGraph/Propane_Vapor_Pressure.GIF (accessed Apr 20, 2016).
(7) Openmodelica. Two phase flow @ build.openmodelica.org https://build.openmodelica.org/Documentation/Modelica.Fluid.Dissipation.PressureLoss.StraightPipe.dp_twoPhaseOverall_DP.html (accessed Apr 21, 2016).
(8) Technifab. Cryogenic Liquid Flow @ technifab.com https://technifab.com/cryogenic-resource-library/cryogenic-thermodynamics/cryogenic-liquid-flow/ (accessed Apr 21, 2016).
(9) Filina, N. N.; Weisend, J. G. Cryogenic Two-Phase Flow: Applications to Large Scale Systems; Cambridge university press, 1996.
(10) Visser, H.-P. LNG sample take-off, vaporization and sampling, 2014, 23.
(11) Lilliefors, H. W. J. Am. Stat. Assoc. 1967, 62 (318), 399–402.
(12) Hollander, M.; Wolfe, D. A.; Chicken, E. Nonparametric statistical methods; John Wiley & Sons, 2013.
(13) Levene, H. Contrib. to Probab. Stat. Essays Honor Harold Hotell. 1960, 2, 278–292.
(14) Technical Committee CEN/TC 282. NEN-EN 12838:2000 Installations and equipment for liquefied natural gas – Suitability testing of LNG sampling systems; 2000.
Field study list of figures and tables
Field study table 1 Results of Lilliefors test for normality. 8
Field study table 2 Means with deviation and Wilcoxon and Levene’s test results for system 1 and 2. 8
Field study figure 1 Schematic showing the boiling point and bubble curve for a two component mixture. 3
Field study figure 2 Typical installation of probe-vaporizer system, ASaP Phazer, in process bypass. 4
Field study figure 3 On-site photograph of the installed systems used for the test. 5
Field study figure 4 Combined methane signal for system 1 (left) and system 2 (right). 6
Field study figure 5 Combined ethane signal for system 1 (left) and system 2 (right). 6
Field study figure 6 Combined propane signal for system 1 (left) and system 2 (right). 7
Field study figure 7 Combined Gross Heating Value for system 1 (left) and system 2 (right). 7