The Flow of Lubricant as a Mist in the Piston Assembly and Crankcase of a Fired Gasoline Engine

For many years, researchers addressing various aspects of automotive engines have been aware of oil droplets in the gas flows through the piston assembly and crankcase [1‐7]. The phenomenon has been generically termed misting, and it has been suggested that misting affects the transport of lubricant through the piston assembly [4, 7], lubricant transport to the combustion chamber [7‐9] and lubricant degradation [1, 3, 7, 10]. Lubricant transport to the combustion chamber is understood to affect emissions [7, 11], oil consumption [9, 11, 12] and combustion, particularly via Low-Speed Pre-Ignition (LSPI) [8, 13‐16]. LSPI is understood to be a combustion event involving fuel and lubricant droplets [8, 13‐18], and is sensitive to lubricant formulation [13, 15, 16, 18, 19]. These droplet flows and their effects are not well understood, so this work aimed to characterise the nature of droplet flows in the engine.

The droplets are thought to be formed by five principal mechanisms, Fig. 1:

Previous work by the authors has reproduced mechanisms 3 and 4 on a laboratory simulator [31]. This work showed that the simulated droplets contain three characteristic droplet diameters: Mist-sized droplets were found in the ranges < 18 μm and 18–135 μm, termed minor and major, respectively, of which the latter were the most commonly occurring. Spray-sized droplets were seen in the range 135–1000 μm. The proportions of mist and spray droplets varied significantly with oil flow rate. Where the mist-sized droplets were predominant, entrainment at a component edge was the primary mechanism of droplet formation. Where the spray-sized droplets were predominant, blow-through was the primary mechanism. This work also indicated that the viscosity of the lubricant and the molecular structure of viscosity modifiers (VMs) had the greatest effect on the tendency of a lubricant to form droplets. An increase in lubricant viscosity generally decreased the tendency to form mist (entrainment from a component edge) but increased the tendency to form spray (blow-through) [31]. The presence of a viscosity modifier had the same effect as increasing the lubricant viscosity, even between lubricants of equal viscosity. This effect was greater when a high-molecular weight linear polymer was used rather than a high molecular weight star polymer [31].

In this study, a method of measuring the size distribution of droplet flows in the crankcase of a fired engine was developed. This was then used to investigate the effect of varying the viscosity of the lubricant and the concentration of viscosity modifier.

Other published works have studied the presence of droplets in engines. Uy et al. [10] considered the effects of aerosols and filtration on fired diesel engines, measuring droplets/particulates < 1 μm and < 10 μm using electrostatic precipitation: The composition of these droplets/particulates were shown to be soot, wear particles and some components expected in the lubricant—Differences were found between the bulk sump lubricants and the droplets, perhaps suggesting degradation or decomposed lubricant from the piston assembly [32, 33]. Clark and Tatli sampled particulates from the crankcase of a diesel engine, using either a condensator [34], or a electrical mobility particle sizer [35], which were 20–400 nm depending on condition [34, 35]. Johnson et al. sampled droplets generated in either a fired or motored diesel engine using either a high-speed camera [36] or a electrical mobility particle sizer in the range of 5 nm–19 μm [37]. Both droplet number and diameter were shown to be load sensitive: 133 nm was characteristic at low load, 30 nm at high load [37]. Spray-sized droplets were seen to form in the crankcase from rotating components [36]. Behn [38] sampled droplets in the range of 0.1–10 μm through the cylinder wall, where the characteristic diameter was ~ 1 μm and concentration was around 1000 ppm. The concentration of oil in these droplets was considerably lower than the concentration of fuel. Dollmeyer et al. [39] measured particulates with a characteristic size of 0.3 μm in a diesel engine. However, based on simulation work by the present authors [31], Johnson et al. [36] (who produced droplets with 10 μm and 320–1000 μm characteristic diameters is a rotating atomiser) and Begg et al. [40] (who produced droplets with characteristic diameters between 20-40 μm and up to 120 μm using a scaled, simulated crankcase), droplets with other characteristic diameters are likely to be present in the crankcase too. Wang et al. [41] injected ~ 1-mm-diameter lubricant droplets into a simulated marine engine combustion chamber. These droplets broke up into smaller droplets: in one case, gas flows ripped small droplets from the surface, comparable to the ‘rolling’ mechanism identified in our previous simulation work [31]; and in another case, larger droplets rapidly disintegrated into smaller droplets. The resulting ‘child’ droplets were > 40 µm in size.

This work was performed on a Ricardo Hydra—a laboratory-based single cylinder, indirect injection, gasoline engine, Fig. 2. The cylinder head is based on a General Motors 2.0 l, 4 cylinder automotive engine. The key properties of this particular version of the Hydra are given in Table 1. The engine is connected to a dynamometer and the engine can be either motored or fired. Several modifications have been previously made to this engine to allow better control and measurement of key tribological parameters: The sump for the crankcase is external to the engine. Thus, the crankcase is nominally dry and a separate lubrication circuit is used for the valve train, which enables lubrication of the crankcase and piston assembly to be studied separately.

In previous work [31], representative mist flows were produced in a bespoke laboratory rig and measured in terms of their droplet size distributions and volumetric flow rate using a laser diffraction particle sizer, a Malvern Spraytec 2000. Due to the success of this approach, and in the interests of consistency and validation of the laboratory methodology, the same measurement apparatus was used in this research. The most suitable way to measure the droplets in an engine was to extract gas from the crankcase of the running engine and pass this flow through the beam of the particle sizer, Fig. 2. The gas flow was extracted through a 105-mm-diameter hole in the anti-thrust side wall of the crankcase and passed through a flexible duct into the enclosure that contained the particle sizer. The end of the duct was such that the beam of the particle sizer intersected the centre of the gas flow at the horizontal approximately 60 mm from the outlet, Fig. 3. The gas flow was extracted from the enclosure through a 155-mm-diameter duct that was attached to the extraction system in the engine test cell. The enclosure contained a breather vent to relieve any positive or negative pressure. The extraction system removed gas from the enclosure at a rate of 500 l/min, which is much greater than the blow-by flow from the engine, which, for example, at 1500 rpm, 33% load and 50% throttle is 11.8 l/min (measured using an AVL 442 orifice blow-by flow meter with one damping chamber, Fig. 4). This has two main implications. Firstly, the flow path into the enclosure with the highest flow rate was through the enclosure breather. However, as the blow-by flow is highly pulsatile, the exact flow rate could not be determined. Thus, the flow rate from the crankcase could also not be determined exactly. Whilst it would have been desirable to be able to know the exact flow rate of gas from the crankcase, there were concerns with such a new system about the possibility of concentrating oil droplets and fuel vapour in the proximity of a hot engine and electronic equipment and the potential over-exerting a back-pressure on the crankcase. Thus, the higher than desirable extraction rate was decided upon to prevent these scenarios from occurring.

Secondly, as the exact flow rate from the crankcase could not be determined, the flow rate of the lubricant as droplets could not be determined either. The laser diffraction particle sizer can determine this by deriving a volumetric oil/air ratio from the scattering data and, if the total volumetric flow rate through the detector was known, the oil flow rate could have been calculated.

Therefore, because of these two factors, this investigation focussed on the measurement of the size distributions of the droplet flows in the engine. This was done with reference to the engine speed, load, lubricant base oil viscosity and viscosity modifier concentration. The testing matrix is shown in Table 2. Measurements were also taken at an intermediate speed condition of 3200 rpm, but are not reported here due to resonances in the extraction apparatus which caused instability in the readings under these conditions.

The test procedure was designed to produce engine conditions where thermal and tribological equilibrium could be reached, such that the loss of lubricant to the extraction system would not be damaging to the engine or render the results unrepresentative. An abnormal increase in component and coolant temperatures was seen as an indicator of lubricant starvation. The temperatures of the coolant inlet and outlet, the cylinder head and the crankcase were monitored. During all the tests, none of these showed significant variation from normal. Thus, it is concluded that the lubricant supply to the piston assembly was not significantly affected.

The engine was flushed with fresh lubricant before each test. Flushing was performed with a normal crankcase side plate, i.e. without the extraction point. A fresh 3 l sump of the test lubricant was connected, the oil filter changed and the engine was motored for 30 min at 1500 rpm and 50% throttle. After 30 min, another fresh 3 l sump of test lubricant was substituted and motored for a further 30 min under the same conditions. Thus, after a total of an hour, a third sump of fresh test lubricant was connected and the engine run at the test conditions until thermal equilibrium was reached. The engine was then stopped for 5 min, whereupon the side plate was replaced and the engine connected to the measurement and extract apparatus. The engine was then run until the previous thermal equilibrium was reached. When thermal equilibrium was reached, non-droplet-size measurements were recorded as an average over a 30-s period. Droplet size measurements were taken over a 5-min period and averaged. Experiments with variations in sample duration indicated that 5 min gathered sufficient data to dampen variation and that longer sampling periods provided no further improvement in data robustness: The droplet size distribution and oil/air ratio were at a generally steady state from around 2–3 min after the engine reached a constant speed (i.e. before sampling was started), indicating that control over the engine conditions by this method was reasonable. The engine was turned off after 5 min of sampling and allowed to cool before the procedure was restarted.

Previous work on lab-based simulators [22, 31, 42] showed that the lubricant properties that have the greatest effect on the tendency to form droplets were the lubricant viscosity and the presence of polymeric viscosity modifiers in the formulation. Therefore, these two factors were the basis for the lubricant matrix for this series of tests. The test engine was designed for SAE 20 lubricants, so the matrix was based around these: API Group III basestocks were used for their narrower molecular weight distributions compared to Group I and Group II. The matrix is shown in Table 2. Base oils of different viscosity grade were used to evaluate the effect of viscosity (SAE 5, SAE 10 and SAE 20). Different concentrations of a high molecular weight star-type viscosity modifier (VM) were included in an SAE 20 base oil to evaluate the effect of viscosity modifier presence, concentration and blend viscosity. VM contents denote the %wt of VM concentrate i.e. including a diluent oil too. The polymer concentration is proprietary but is within a representative range for a crankcase lubricant. To protect the engine, an overbased calcium sulfonate detergent (1%wt) and a succinimide dispersant (1%wt) were included in all test lubricants.

To aid comparison with the previous work [31], the lubricants investigated here were tested in the laboratory simulator rig used in this work. This rig consisted of a venturi droplet generator that produces flow representative of the flows seen in the piston assembly. The lubricant that did not form droplets and ran out of the system is weighed to determine the proportion of that which does form droplets. The flow rate of droplets formed as a percentage of the total lubricant flow was termed the Droplet Formation Tendency. For a more detailed description of this apparatus, method and interpretation of droplet formation phenomena, see the previous work by the current authors [31]. Figure 5 shows the droplet formation tendency of these lubricants at a range of flow rates. This shows that, at low lubricant flow rates, where greater resistance to droplet formation manifests as lower % Oil Misted, the droplet formation tendency of lubricants without viscosity modifiers was significantly influenced by their viscosity: The lower the viscosity, the greater the tendency to produce droplets. Polymer-containing lubricants showed significantly reduced droplet formation tendency. At higher lubricant flow rates, the polymer-containing lubricants showed a more linear increase in droplet formation tendency with flow rate, characteristic behaviour of ‘blow-through’, where greater resistance to droplet formation under shear causes lubricant to accumulate in the venturi and droplets are formed by air flow through the lubricant rather than over it. Counterintuitively, under these conditions, greater resistance to droplet formation under shear manifests as higher % Oil Misted, i.e. the transition from shear-driven, thin film droplet formation to a bulk flow phenomenon occurs at a lower flow rate. These lubricants had significantly different viscosities but all exhibited similar droplet formation behaviour. This indicated that the droplet formation tendency of these lubricants was most influenced by the presence of polymeric viscosity modifiers and appeared to be independent of polymer concentration in this range. This, in turn, indicated that the increased viscoelasticity imparted by the polymeric viscosity modifiers was the main mechanism. These, therefore, were the properties considered when interpreting the data obtained from the engine.

Figure 6 shows the droplet size distributions for polymer-free lubricants at all engine conditions. Based on the repeated tests conducted, the 95% confidence intervals for each of the droplet distribution parameters are shown in Table 3. The statistical significance of all the observations in this work were made with reference to these statistical limits.

As in the laboratory simulation rig, three characteristic droplets diameter ranges were observed in the engine. However, the characteristic diameter ranges were significantly different. These differences in distribution were observed even when comparing the distributions of the same lubricants in both the simulator rig and the engine, Fig. 6, [31]. In the engine, droplet diameters of 0–30 μm characterised the minor misting region, 30–250 μm characterised the major misting region and 250–1000 μm characterised the spray region. In the previous study, the ranges were identified as 0–18 μm, 18–135 μm and 135–1000 μm, respectively [31]. It was observed in Sect. 3 that the presence of detergent and dispersant chosen for this study did not significantly affect the misting properties of the lubricant. So, the likely reasons for the differences in droplet size ranges between the engine and simulator are as follows:

Aerosol-sized droplets (0.1–1.0 μm diameter) were rarely observed in the distributions. They may have been present but were not significant in their volume frequency due to their size. Droplets and particulates in this size range have been identified by other authors [10, 34, 35, 37, 39], albeit using different engines and droplet sizers, and were understood to be formed by the condensation of fuel/oil vapour with or without soot particulates [5, 10, 35], leading to high concentrations of fuel relative to the rest of the system [38]. Droplets in the 5–15 μm diameter range were more abundant. It is unlikely that the 0.1–1.0-μm-diameter droplets in other studies and the 5–15-μm-diameter droplets here represented the same mechanisms e.g. greater condensation rates in the present study led to a greater droplet diameter. It is more likely that these represented different formation mechanisms. Mist-sized droplets and spray-sized droplets were observed, Fig. 6. The ratios of mist-sized droplets to spray-sized droplets by way of volumetric proportion, indicated by the relative peak heights and areas, were similar to the ratios observed in the laboratory simulator [31].

The measured distributions did not definitively indicate the source mechanism of the droplets in the crankcase. Instead, the distributions appear to be a combination of droplets from different source mechanisms. The trimodal distributions correlate with those formed in the laboratory rig by flow over a component edge. However, some variations in the distributions, as explained below, suggest that some droplets originated elsewhere, e.g. blow-through of oil pockets or inertial throw-off from the con-rod and crankshaft [26, 27, 40].

Mist-sized droplets produced by the SAE 10 lubricant had a unimodal distribution, Fig. 6, but were bimodal for SAE 5 and SAE 20. The root cause was not apparent, although it was repeatable. The distribution for SAE 10 at 4500 rpm and 75% load appeared to show two characteristic sizes with overlapping distributions. However, this did not appear to be the case for the 4500 rpm and 33% load condition. This variation may have been caused by differences in the molecular weight distribution of the base oil, an artefact of the droplet formation process that produced variability in the droplet agglomeration or release from a surface, or a combination of all these.

Therefore, comparisons between lubricants of different viscosity were made between SAE 5 and 10, i.e. bimodal mist-sized droplet distributions: Firstly, at 4500 rpm, the characteristic diameter of the major mist droplets was greater for the higher viscosity lubricant, Table 4. This agrees with Dasch et al. [22] and the laboratory simulator [31], whereby higher viscosity produced larger characteristic droplet sizes. Characteristic diameters of minor mist or spray-sized droplets did not vary significantly. Secondly, the combined volumetric proportion of minor and major mist-sized droplets was lower at 4500 rpm for the lower viscosity lubricant, Table 5. Simulation work measured higher mist-sized droplet proportions with lower viscosity. This was not a contradiction because droplets in the crankcase probably have several source mechanisms, i.e. a lower viscosity lubricant might simultaneously produce a greater quantity of spray-sized droplet by inertial mechanisms, e.g. from the crankshaft [40], reducing the relative proportion of mist-sized droplets. Because characteristic diameters (i.e. volume) of major mist droplets were greater with higher viscosity lubricant, but minor mist characteristic diameters were unchanged, the ratio of major to minor mist volumetric proportion was greater with higher lubricant viscosity, Table 6.

With regard to engine conditions, the relative lack of correlation between droplet size distributions for different lubricants at 1500 rpm indicated clear speed dependence of droplet formation. The narrow and relatively non-Gaussian distribution around some characteristic sizes at 1500 rpm could represent either greater uniformity of droplet diameter or a smaller sample size of droplets. At 4500 rpm, droplet size distributions showed little significant variation with load. However, one statistically significant observation was that characteristic diameters of minor mist droplets increased at higher loads, when such droplets were present. Droplets may have grown by condensation of fuel/oil vapour: Higher piston assembly temperatures at higher loads increased evaporation rate, and greater temperature difference between piston assembly and sump increased condensation rate. Minor mist droplets will have been most greatly affected by condensation due to high surface-area-to-volume ratio.

Figure 7 shows the variation in droplet size distribution for lubricants containing varying concentrations of viscosity modifier. Because the viscosity modifier was added to the same SAE 20 base oil, there were no lubricants with the same dynamic or kinematic viscosity at either crankcase (~ 60 °C) or piston assembly temperature (> 100 °C [43]). Overall, the proportion of mist-sized droplets was lower when viscosity modifier was present. This effect was significant at 10% and 15% but marginal at 5% concentration, especially at low load. This could represent a reduction in mist formation or an increase in spray, perhaps both.

Considering engine conditions: The proportion of mist was significantly lower at 75% load than at 33–50%, Table 5. Within this, major mist proportion at 50% load was lower than at 33% load, most clearly with 10–15% VM concentration, but was insignificantly different at 75%. Referring to Fig. 5 (where laboratory scale testing indicated primarily viscosity-dependent droplet formation in non-VM lubricants, and primarily viscoelasticity-defined droplet formation in VM-containing lubricants): Because non-VM lubricants did not show this load dependence, it suggests that the load-dependent differences in droplet formation with VM-containing lubricants in the engine are a viscoelastic effect. Perhaps the higher piston assembly temperatures at 50–75% load increased the viscoelastic reduction in droplet formation, i.e. when polymer molecules extended. This effect may not have been as great at 5% VM due to reduced total capacity for viscoelastic energy storage. There were greater similarities in the droplet distributions between 1500 and 4500 rpm conditions. However, the only statistically significant change was higher characteristic diameter spray droplets at 1500 rpm: These correspond more closely to spray droplets at 33% load, i.e. lower temperatures and gas flow rates. Overall, characteristic droplet diameter peaks showed lower variation around the mean at 1500 rpm.

There were extremely low proportions of minor mist proportions at 75% load, at both 1500 rpm and 4500 rpm: The proportion was higher in lubricants without VMs, suggesting a viscoelastic effect. However, minor mist proportions were greater for the same lubricant at 33–50% load, i.e. VM was not the only factor. Perhaps viscoelasticity affected particular flow mechanisms that were more influential at higher load. Unlike with non-VM oils, there was no significant change in these droplet diameters with load. Major mist droplets at 75% load had smaller characteristic diameters than at 33–50% load. This was more pronounced at 4500 rpm than 1500 rpm. This change was only observed for lubricants containing VM: Characteristic major mist diameters for SAE 20 base oil containing 5% and 10% VM were smaller than for SAE 20 base oil alone, Table 4.

Greater VM concentrations had progressively greater effect, but not linear. A minimum VM concentration appeared to be needed to affect different droplet formation mechanisms. At 75% load, this appeared to be between 5 and 10% VM. At 33–50% load, this appeared to be between 10 and 15%; for example, suppression or elimination of minor mist droplets was seen for 10% VM at 75% load but for 15% VM at 33% and 75% load.

After lubricant droplets are formed, they continue to interact with the gas flows in which they are entrained. Droplets can deform and break up into smaller droplets and/or coalesce into larger droplets as they interact with each other. Two key parameters were used to describe droplet stability: Firstly, the Weber Number [44]:where ρ_G is the density of the gas phase, v is the relative velocity between the droplet and the gas flow around it, d is the diameter of the droplet and σ is the surface tension of the droplet. This is the ratio of surface tension forces to the inertial forces. We is considered the most influential descriptive parameter for droplet breakup, even in combination with other parameters [45].

Secondly, the Ohnesorge Number [44]:where ρ_L is the density of the droplet liquid. This is the ratio of viscous forces to the surface tension forces. When Oh is greater than 1, the viscosity of the droplet is considered to have the dominant effect over the droplet breakup process.

Different droplet breakup mechanisms have been observed:

Table 7 shows various models and the dimensionless parameters that describe the transitions between these mechanisms. The Cascade Atomisation and Breakup [46] model was selected for this study because it covers all three breakup mechanisms and was validated against non-evaporating fluids.

The following generalised flows were considered and the assumed conditions are shown in Table 8:

Cascade Breakup Model criteria were applied to droplets with the mean droplet diameter in each characteristic droplet diameter range as defined above: Minor Mist = 0.1–30 μm, Major Mist = 30–250 μm and Spray = 250–1000 μm.

Table 9 shows the outcomes of the Cascade Breakup model applied to the crankcase. Table 10 shows the outcomes of the Cascade Breakup model for the piston assembly. In the crankcase, under both conditions, breakup of spray droplets would occur at relative gas-droplet velocity of approximately 17 ms⁻¹. The Weber number of these flows implies that these droplets would be in the bag breakup regime. At the 100 °C condition, spray Ohnesorge Number was 0.03–0.15, i.e. greater dependence on surface tension than viscosity. Conversely, at local gas temperature, spray Ohnesorge Number was 3.1–18.2, i.e. high dependence on viscosity. Detailed thermodynamic analysis would be required to determine the characteristic condition. Under all crankcase conditions, the major mist and minor mist droplets would not readily break up. Changes to these droplets would occur by any coalescence, deposition or condensation. Latter mechanisms aside, droplet from the crankcase would not be significantly changed during extraction and measurement.

Crankcase conditions did not vary greatly with load, so the greatest cause of changes in breakup parameters was variation in droplet size diameter, i.e. engine conditions influence indirectly by affecting droplet size. For instance, at 1500 rpm and 75% load, the SAE 10 and SAE 20 droplets would break up at around 15 ms⁻¹, due to their larger initial diameter. 5% VM lubricant at 4500 rpm and at 75% load was the only condition where droplet diameter affected the transition to breakup, where smaller characteristic diameters would make droplets stable up to 20 ms⁻¹.

Comparing the major mist droplets of the SAE 20 and SAE 20 + 5% VM at hypothetical 100 °C conditions, the presence of VM increased the Ohnesorge Number [44] by 62% at 33% load and 196% at 75% load. A change of similar magnitude would also be seen at the crankcase local temperature condition. This represents the largest difference in Ohnesorge Number between any of the evaluated conditions. Thus, there would be a large difference in the viscosity-dependent behaviour of the droplets formed by these two lubricants. This further suggests that there was change in the balance of the droplet formation mechanisms caused by the presence of VM.

In the piston assembly, minor mist Ohnesorge Numbers under all conditions were greater than 1, implying high viscosity dependence. However, major mist flows had Oh = 0.5–1.1, i.e. significantly lower viscosity dependence. Minor and major mist droplets at both conditions (68 ms⁻¹ and 145 ms⁻¹) would be expected to break up readily. Minor mist droplets at 68 ms⁻¹ would be in the bag breakup regime and many major mist droplets at 145 ms⁻¹ would be in the catastrophic breakup regime. Minor mist droplets at 145 ms⁻¹ and major mist droplets at 68 ms⁻¹ were in either the bag or shear regime depending on original size, but are close to the transition We of 80. Essentially, droplets would quickly break up in the piston assembly. This may increase lubricant evaporation from smaller droplets with their higher surface-area-to-volume ratio.

Comparing measurements from the laboratory simulator [31] and the engine:

These findings have implications for engine tribology:

Lubricant droplet flows in the crankcase of a fired gasoline engine have been extracted and droplet size distributions measured: