Before presenting the results, at this point, the test carrier shall be described. The test engine is a large-bore single-cylinder engine operated by natural gas. For reasons of confidentiality, a more detailed description of the test carrier must, unfortunately, be omitted at this point.
The results section consists of two types of results. One part focuses on the challenges and respective results achieved when these challenges were analyzed and finally overcome. The second part described the actual engine test bench experiments and comparisons to a state-of-the-art method for LOC measurement.
3.1. Challenge 1: Amount of Tracer
As mentioned previously, the tracer is manufactured by conducting a hydrogen/deuterium-exchange reaction on synthetic base oil. This reaction is based on the findings described in [
13,
14] but has been further improved in the course of the present research project. One of the first obvious challenges is the amount of tracer (deuterated base oil) needed in order to spike a complete oil filling of a large engine with deuterium. In order to achieve a deuterium signal in the exhaust gas, clearly distinguishable from the background, around 1% (
w/
w) of deuterium needs to be added to the oil. Assuming a hydrogen-deuterium exchange ratio of around 70–80 at%, this means around 5% (
w/
w) of tracer need to be added to the oil. For passenger car engines, this means, around 200 to 600 g of base oil have to undergo the hydrogen/deuterium-exchange process. However, large single-cylinder engines installed at research facilities have oil conditioning systems and, therefore, large systems circulating big amounts of lubricating oil. In this case, the lubricating system contains around 200 L of engine oil. This means an amount of roughly 10 L of synthetic base oil has to be deuterated. This entails two challenges: cost and time for the reaction. These challenges were solved using a pressure reactor with a volume of 5 L, so 1 L of oil could be deuterated at once. In addition, the heavy water used as a deuterium source and the palladium catalyst was recycled after the second reaction, further lowering the cost.
Continuous monitoring not only of the efficiency of the reaction but also of the stability and concentration of deuterium in the lubricating oil of the engine is crucial for the successful application of this newly developed method. This monitoring was carried out using a portable Fourier transform infrared spectrometer (FTIR) from the company eralytics GmbH. The eralytics ERASPEC OIL (
Figure 2) is capable of determining a variety of parameters such as total base number (TBN), total acid number (TAN), contaminants, degradation products by predefined methods. In addition, it is possible to receive the complete infrared spectrum. Different algorithms for determining the deuterium concentration in the oil have been defined and evaluated. By calibration with
1H-nuclear magnetic resonance (NMR) spectroscopy, the most accurate and promising algorithm in terms of linearity has been chosen for further consideration.
The following figures show typical infrared (IR) spectra of complete engine oil without (
Figure 3) and with (
Figure 4) deuterium tracer added. After deuteration, a significant peak in the highlighted region between 1900 and 2400 wavenumbers is visible. Additionally, the area and the new baseline chosen by the algorithm are highlighted.
Comparisons to
1H-NMR measurements have shown there is a strong linear correlation (coefficient of determination or R2 score of 0.9984) between the concentration of deuterium and the area underneath the respective peak in the IR spectrum (
Figure 5). The blue dots represent the collected measurement data for each concentration, whereas the red line depicts the calculated regression line.
3.2. Challenge 2: Signal Bias by Unburned Methane
Cross-sensitivities towards oxygen, carbon dioxide and nitrogen, such as described in [
15,
16] have been investigated in earlier publications. It turned out, that the effects on the calculated LOC are almost negligible and a mathematical correction is feasible, even though the parameters for respective corrections seem to differ between literature sources and experimental investigations.
As described in [
17], methane (CH
4) can as well interfere with the water isotope spectra and bias the measurement of the
1H/
2H-ratio. Therefore, the isotopic water vapor analyzer has a built-in algorithm correcting for this effect. According to [
17], the deviation should be linear; however, the CH
4 measurement of the analyzer is being calibrated for 2 ppm CH
4 in dry air. As in the exhaust gas of the engine, a volume share of around 1000 to 2000 ppm of unburned CH
4 is expected; the built-in correcting algorithm might not work properly.
As described in [
18], the reactivity of platinum for the conversion of methane is low compared to other materials such as palladium or rhodium. Therefore, the platinum catalyst used for diesel and gasoline fuels needed to be exchanged for palladium.
Figure 6 the CH
4 concentration measured by the analyzer using a platinum (Pt) or a palladium (Pd) catalyst. Due to the dilution of the exhaust gas with dry air, the CH
4 volume fraction in the sample gas is already much lower than in the raw engine exhaust gas (around 300 ppm instead of 1500 ppm); however, this still leads to a significant bias of the
1H/
2H signal of the analyzer.
As palladium is very sensitive to poisoning by sulfur or phosphorus coming from the engine oil or fuel, the reactivity might decrease after some time. In the course of the experiment (50 engine operating hours), however, the catalyst did not seem to be deteriorating. According to [
18], rhodium as a catalyst would be less sensitive to poisoning; however, the reactivity with methane is slightly lower. At temperatures of 500 °C total oxidation of methane can be reached with both palladium and rhodium. In terms of cost, palladium still is a more economical option than rhodium, even if it has to be exchanged more often. As described in [
18], platinum reaches a maximum conversion efficiency of 35%, even at 500 °C. As for all catalysts, the temperature is a crucial parameter for reactivity; the catalytic converter is (pre)-heated to 450 °C. This temperature was limited by the catalytic material itself minus the temperature increase due to the energy released during the conversion of methane.
Additionally, the dilution with dry air increases the reactivity by both increasing the share of excess oxygen and lowering the humidity of the exhaust gas. Further information on the catalytic materials and correlations with humidity and temperature is [
19] highly recommended.
3.3. Challenge 3: Background Fluctuations
As deuterium (2H) is a stable, non-radioactive isotope, it is abundant in every compound containing hydrogen. Even though its natural abundance of 0.0156 at% is very low compared to the other stable hydrogen isotope protium (1H) with 99.98 at%, its occurrence in fuel and in the humidity of the ambient air has to be taken into account when calculating the lubricating oil consumption. The abundance in ambient air can be easily measured with the installed water isotope analyzer by just disconnecting it from the exhaust gas sample line. However, the procedure for fuels is more complicated. Liquid fuels can only be measured with the help of isotope-ratio mass spectrometry (IRMS) or 1H-NMR. For liquid fuels, the deuterium concentration is assumed to be equal to the natural abundance.
However, for gaseous fuels—especially methane—the actual source has an impact on the respective deuterium abundance. According to
Figure 7 ([
20]), the isotopic hydrogen ratio (delta
2H, or δ
2H) may fluctuate in a range from roughly −100‰ to −400‰ (the isotopic ratio of hydrogen is usually compared to the Vienna Standard Mean Ocean Water, which is defined to be 0‰). A negative ratio characterizes a deuterium-depleted sample compared to ocean water.
For the city of Graz, where the present experiments have been carried out, energy suppliers obtain the natural gas from the Baumgarten gas hub. The gas itself is a mix of various fossil natural gas sources, mainly from Russia and Ukraine, and a certain proportion of Austrian natural gas and purified biogas. The latter, in particular, naturally has very light isotope contents. The proportions are not constant over time but are recorded by the energy supplier. The gas itself is temporarily stored in underground reservoirs with a total capacity of about 92 TW (conversion 11.5 kWh/Nm3), so it can be assumed that the temporal changes in isotopic composition are relatively stable over the course of one day. In the longer term, the isotopic composition is probably already subject to certain fluctuations. However, there are records at Gas Connect Austria about temporal changes of the share of the different sources.
For an accurate online oil consumption measurement, this means that the composition of the intake gas mixture has to be monitored in certain intervals. The investigation of these background fluctuations will be a topic of further research in order to find out about the maximum length of monitoring intervals. With the current setup, continuous monitoring of the background during engine operation is not possible without additional effort, as the catalyst is designed for volume fractions of roughly 0–4000 ppm of methane. If this fraction exceeds a certain value (e.g., when sampling the inlet gas/air mixture), the temperature generated at the catalyst due to the reaction enthalpy will exceed 500 °C, which could harm the catalyst.
Potential solutions might be an intermittent measurement of the inlet concentration by increasing the air-fuel ratio in order to reach methane volume fractions of 1500 ppm or by using a tube furnace instead of the catalyst, which could enable continuous monitoring throughout the engine operation. However, two downsides come along with using a tube furnace. First, the sample gas flow must be low, in order to reach a high catalytic efficiency, which would imply long measurement cycle times. Second, in the past, it has turned out that a tube furnace is not robust enough to permanently withstand the rough conditions on an engine test bench (temperature changes, vibrations, etc.). Therefore, for the presented prototype, a pellet-type catalyst has been installed.
For the measurement campaign described in this paper, background measurements have been carried out only once before the start of the actual measurement, as depicted in
Figure 8. In this case, the hydrogen/deuterium-ratio was first measured at ambient air. Then, the sample line was connected to the engine inlet in order to measure the isotopic signature of the air/fuel mixture. This was done using a tube furnace instead of the catalytic reactor to prevent damage to the reactor. During the test, the engine was operated at a constant load and speed.
The isotopic ratio in the background of the intake air/fuel mixture consists of two compounds: the humidity in the intake air, which, during those experiments, was set to only 0.1 g/kg, thus, has a very low influence on this value. Therefore, the 1H/2H-ratio is mainly driven by the water coming from the combustion of natural gas. The average 1H/2H-ratio in the background (mainly natural gas) was found to be −225‰, meaning the used natural gas is deuterium depleted, increasing the selectivity of the oil consumption measurement.
In addition to the preparatory measurements described above, the measurement program included the recording of several load curves to check the reproducibility of the method. The results of the deuterium method were compared to the SO
2-method, which is based on the use of sulfur as a tracer. The sum of collected measurement data is depicted in
Figure 9. Both diagrams depict the mean effective pressure on the abscissa. On the ordinate, the lower diagram shows the lubricating oil consumption (LOC), whereas the upper diagram shows the LOC divided by the engine power, the so-called brake specific lube oil consumption.
At a qualitative level, both measurement methods show very similar trends. The quantitative deviation of the average values is as well satisfactory in most of the operating points. However, both methods show fluctuations to a certain extent. The coefficient of variation is much higher for the SO2 method. This is mainly caused by fluctuations of total sulfur in the natural gas due to the odorization processes of the gas supplier. In combination with the use of a low-sulfur oil (total sulfur 2300 ppm), low fluctuations in the intake mixture and SO2-measurement in the exhaust gas led to large deviations in the LOC measurement. For the SO2 method, sulfur concentrations in the lubricating oil of more than 5000 ppm are required to achieve satisfactory measurement accuracy. However, the measurement uncertainty of the total sulfur analysis for fuel and oil is generally a major challenge in the application of this method.
Although the signal from the deuterium method can also be distorted by background fluctuations, their effect on the LOC measurement does not appear to be that great. The theoretical measurement accuracy of the experimental setup, calculated by Gaussian error propagation of the measurement of all mass flows and their respective deuterium concentration, was found to lie in a range between 0.01 g/kWh and 0.03 g/kWh depending on the operating point.
Two major improvements could be made to the deuterium method with little effort. First, more tracer can be added to the engine oil, further increasing the selectivity and accuracy of the method. Second, the isotope pattern of the aspirated gas-air mixture could be monitored more regularly to avoid the signal being distorted by the origin of the gas.
At this point, for the sake of completeness, results from passenger car engine (PCE) tests shall be displayed. In previous experiments, engine characteristic maps of passenger car engines have been recorded using the same method [
3]. In order to compare both results, a load curve has been cut out from the engine characteristic map, as depicted in
Figure 10. As seen before, the bullets represent actual measurement points, whereas the line represents average values.
According to literature, but also to past experiments at the facilities at the LEC, the brake-specific lubricating oil consumption should follow the shape of a bathtub-curve, when depicted in relation to the mean effective pressure. This shape can be seen for both the large engine as well as for the passenger car engine, though it seems to be more pronounced for the PCE. This study strongly emphasizes the usefulness of the variable “brake specific oil consumption”, as the reference to engine power allows a comparison of different engine types, regardless of size or fuel used. For both engines, the brake-specific lube oil consumption (BSLOC) lies within a range of roughly 0.1 to 0.3 g/kWh.