Printer Friendly

Investigation of engine oil base stock effects on low speed pre-ignition in a turbocharged direct injection SI engine.


Automotive manufacturers have realized vehicle fuel economy benefits in recent years by downsizing engines and employing turbocharged direct injection technology. Downsizing the engine increases fuel economy by reducing the vehicle weight, frictional losses, and pumping losses in the engine. To compensate for the loss in power associated with engine downsizing, manufacturers have employed turbocharged direct injection. It is reasonable to expect these engine designs will continue to find favor with automotive manufacturers and consumers.

While the performance benefits of turbocharged direct injection are significant, this technology presents new challenges for fuel and lubricant manufacturers. One challenge of great importance is low speed pre-ignition (LSPI). LSPI occurs when some fraction of the fuel-air mixture in the combustion chamber ignites prior to the spark plug firing. This pre-ignition disrupts the engine timing and can create dangerously high pressures in the combustion chamber as the piston compresses the expanding combustion products. The high pressures and disrupted timing can lead to knocking in the engine, and in the worst case, produce catastrophic engine failure after only a few LSPI events. Turbocharged direct injected engines are particularly susceptible to LSPI when operating at low engine speeds (RPM typically < 3000) and high brake mean effective pressures (BMEP).

LSPI differs from conventional knocking in several ways. Principally, LSPI refers to pre-ignition that occurs at low engine speeds and high loads. Furthermore, unlike conventional end-gas knock, LSPI is random in nature and occurs unpredictably even while the engine operates at LPSI-prone conditions. Several reports have been published aiming to determine the exact causes of LSPI and yet no single cause has been conclusively identified. Factors such as fuel quality, engine operating conditions, and lubricant formulation have all been shown to influence the frequency and severity of LSPI events. Because of the unique nature of LSPI and the potential damaging impact it can have on engine hardware, the next generation International Lubricants Specification Advisory Committee (ILSAC) GF-6 and General Motors (GM) dexos1[R] engine oil specifications will include testing to ensure robust LSPI performance for licensed lubricants.


There have been many investigations into LSPI published in the past five years. These studies have generally employed similar methodologies for generating LSPI in the laboratory and identifying its causes. Most research follows the following protocol. The researchers fit a test engine with pressure transducers in the cylinders and operated the engine in low speed, high load conditions. The researchers collect pressure data over time, identifying pre-ignition events by transient periods of extremely high pressure. (One technical challenge is the lack of a single definition or threshold for distinguishing LSPI from other abnormal combustion events.) The researchers may supplement data collection with high-speed imaging of the combustion and may compare imaging data to computer simulations of combustion chemical kinetics and fluid dynamics. The researchers vary some conditions of the experiment and judge the effect on the frequency of pre-ignition events. Some examples of conditions explored in these studies are spark plug timing, fuel injection strategy, fuel chemical composition, engine oil composition, and engine coolant temperature.

The investigations most relevant to this paper are those that explored the influence of the engine oil on LSPI. We will summarize the most relevant research here.

Dahnz et al. [1] found that LSPI events are intermittent and alternate with regular combustion cycles. Combining this evidence with observations from combustion chamber imaging, they concluded that pre-ignition is related not to persistent hot spots in the engine (which would lead to consecutive rather than intermittent pre-ignition events), but rather shorter-lived contaminants entering the combustion chamber and initiating combustion. The authors hypothesized that the LSPI mechanism begins with fuel impinging on the cylinder walls and mixing with the lubricant film that coats its surface. The resulting fuel-oil mixture has different physical properties than the pure engine oil that make it more able to form droplets that eject from the wall or the piston grooves and enter the combustion chamber. Because fuel-oil mixture has a lower ignition temperature than the fuel itself, these droplets ignite prematurely. An unexpended finding in this paper was that decreasing the engine coolant temperature (thereby lowering temperature at the cylinder walls) had the effect of increasing pre-ignition frequency.

Zahdeh et al. [2] proposed a similar hypothesis for LSPI cases based on their experimental results. They found that fuel injection strategies meant to minimize fuel impingement reduced the frequency of pre-ignition events. This seems to support the hypothesis described above. Similar to Dahnz et al., they found that lower coolant temperatures increase pre-ignition frequency. The authors tested three different engine oils formulated with various base stocks (hydrocracked petroleum-based stocks and synthetic PAO stocks) and various concentrations of lubricant additives. The engine oil with PAO base stock and lower concentrations of additives produced the least amount of pre-ignition, although it is difficult to conclude whether this is due to the base stock, the additives, or some combination of both.

Takeuchi et al. [3] focused their studies on engine oil effects on LSPI. The authors blended and tested a variety of engine oils with various base stocks and varying concentrations of detergent, friction modifier, and antiwear additives. The authors identified two trends that characterize the base stock effect on pre-ignition event frequency. First, the base stocks' pre-ignition tendencies ranked in order of their API group. The Group I base stock produced the most pre-ignition events and Group IV the least, and the results ordered as Group I > Group II > Group III > Group IV. Second, a higher viscosity Group IV base stock produced more pre-ignition than a lower viscosity Group IV. Seeking a more fundamental explanation for these observations, the authors found that the auto-ignition temperature (measured by calorimetry) of the base stocks at 10 atm pressure correlates well with pre-ignition event frequency. This led the authors to hypothesize that the engine oil effect on LSPI is a better described as a chemical (combustion-related) phenomenon than a physical (evaporation or otherwise) one.

Welling et al. [4] further explored LSPI and its relationship to fuel and engine oil ejection into the combustion chamber. The authors modified an engine to directly inject engine oil into the combustion chamber to more easily produce pre-ignition events. The authors found no significant correlation between base oil group and preignition event frequency. Other attempts to classify the physical properties of the lubricants' influence on LSPI produced results that are difficult to interpret.

To summarize, there are numerous factors that appear to affect LSPI. Some of these factors are mechanical, such as engine speed and load, fuel injection strategy, and engine cooling. In fact it is possible that purely mechanical solutions to LSPI exist. Other factors involve the fuel and the lubricant properties and chemistry. Because the science of LSPI is still developing, it is difficult to synthesize the various experimental findings into one comprehensive hypothesis. Nevertheless, it is useful to find common themes in the work and collect them to form a conceptual model: A mixture of fuel and engine oil collects in the piston crown land crevice over a period of time during engine operation. This collection mechanism is accompanied by sporadic ejection of small (~ 100 um) droplets of some composition that have sufficient character to auto-ignite, thus causing premature ignition of the fuel-air mixture already present in-cylinder. This "droplet-hypothesis", as well as other theories involving deposit formation, forms the basis of understanding the causes of LSPI in the engine.


Our experimental apparatus was the LSPI stationary engine test developed by Southwest Research Institute (SwRI) for the Preignition Prevention Program (P3). This test method employs a GM Spark Ignited L4 2.0 L Ecotec[R] turbocharged gasoline direct injection engine. Our test protocol identified low speed pre-ignition events from in-cylinder pressure data obtained during prescribed engine speed and load set points, and included analyses of both peak in-cylinder pressures as well as 2% Mass Fraction Burned metric derived from the calculated heat release rate. The protocol identifies a pressure or heat release excursion greater than 4.7 standard deviations from the mean as an LSPI event. The exact operating procedure used in the experiment is proprietary to the P3 consortium.

Figure 1 describes the engine speed and load set points in the protocol. The test protocol repeats these test conditions four times, giving a total of 16 separate "counts" of the LSPI frequency per 25,000 engine cycles for each lubricant composition. The "Low Load/Speed" conditions generated too few LSPI events for our test oils to warrant a detailed analysis. Therefore this study considers only the data obtained at the "High Load/Speed" conditions. We typically observed between 0-60 pre-ignition events over the duration of a test.

The goal for this apparatus is to produce an experimental setup for producing and measuring LSPI in the laboratory. Because LSPI is highly sensitive to engine design and engine operating conditions, quantitative results are specific to the particular engine and protocol. However by employing a real production engine, we hope that the results of our experiments are qualitatively descriptive of LSPI in real driving conditions in the field.


In our research we observed that LPSI frequency can vary over time in a test engine even as test oil and engine conditions are held constant (Figure 2). We also observed bias in results from different engines.

It is not uncommon in engine testing for test results to drift over time even when the test protocol includes careful measures to break in the engine and flush the oil between runs. To address this bias, industry standard ASTM tests implement a statistical control system employing reference oils run periodically in test engines [5] and/or engine hour correction factors [6] to account for engine hours bias ("engine hours" refers to the number of operating hours logged on an engine).

It is outside the scope of this study to implement such a formal statistical control system. However we choose to test a reference oil at periodic intervals during our experiments to benchmark engine severity over time. This reference oil is a fully-formulated oil similar to the test oils in our experiments. We observed enough variation of reference oil results to justify implementing a correction factor to normalize our LSPI results. The goal of this correction factor is to correct for the observed engine hours effect and engine-to-engine bias.

Equation 1 shows the equation we used to normalize LSPI results based on reference oil results. The term in the denominator of Equation 1 is the expected value of the reference oil LSPI frequency at the appropriate engine hours of the test oil run. We calculated this number by linear interpolation of the reference oil LSPI results collected before and after the test oil in the particular test engine. The numerator in Equation 1 is the calculated mean of all observations of the baseline oil (128 data points), and is what we consider as the reference level performance for the baseline oil. The units of Raw LSPI Frequency in this equation are number of LSPI events per 25,000 engine cycles.

LPSI Frequency = (Raw LPSI Frequency) 26.8/interpolated reference oil LPSI frequency (1)

The cause of test engine drift is uncertain. One possible cause is gradual wear in the cylinder liners. Cylinder liners are typically honed with a cross hatched pattern to help retain oil for proper lubrication. As the engine wears this hone reduces in capability to hold engine oil, therefore leading to reduced LSPI due to less available oil entering the combustion chamber to contribute to LSPI. Furthermore, the test method itself may have a significant impact on the drift observed in the engine. The engine operates in a regime that is designed to cause LSPI, which can lead to gradual damage to the engine with time. While the exact causes of the engine drift are unclear, accounting for this effect is a critical step before attempting to draw conclusions from the data.

One unfortunate byproduct of this engine hours correction is that it leaves us without a way to estimate test precision--we had initially intended the reference oil runs as repeat data points for this purpose. With the reference oil results now serving as the source of a correction factor, we are left with no repeat runs. Therefore it is impossible to place error bars on our data. Instead, we will rely on a statistical analysis to judge the significance of base stock property effects on LSPI.


To measure the effect of lubricant base stock choice on LSPI we blended 11 engine oils consisting of various base stocks and base stock blends with a common set of lubricant additives. The additive chemistry in these oils is similar to that found in many commercially available premium engine oils, and is identical in all the test oils. Additionally, we kept the viscosity modifier type and concentration constant in all these oils. Table A1 summarizes the properties of the base stocks in our study and the LSPI results for their corresponding finished oil.

The goal of this analysis is to measure the relationship between base stock properties and LSPI performance. Generally the properties of interest to base stock manufacturers and oil marketers are viscosity, viscosity index, volatility, concentration of aromatics, and concentration of sulfur. These measurements do not fully describe the complex variety of base stock composition, but they are useful and common metrics and therefore we have chosen these as the basis of our analysis.

We deliberately selected base stocks spanning a wide range of properties. API Group I, II, III, and IV stocks are all represented in the design. The properties of the Group I stocks and some of the Group II stocks in this design would make them inadequate for use in commercial engine oils, however we decided it is useful to include these base stocks to test the extremes in properties. The aromatics concentrations of 500-600 mmol/kg (approximately 20% by weight) and the 27-29% NOACK volatility of the two Group I stocks in the design are both particularly high.

We have chosen to both present the results in a plot and (in the next section) attempt a more detailed data analysis. We plot the LSPI Frequency in Figure 3 as it varied with base stock kinematic viscosity, and mark the points by their API Group; viscosity and API group were the variables that ref [3] identified as affecting LSPI Frequency. The two highest viscosity base stocks in our design produced the most LSPI events in our experiments. Based on these two results, there appears to be a trend of increasing LSPI events with base stock viscosity (consistent with the finding of ref [3]), however there is no obvious trend with API Group (which differs from the findings of ref [3]). There also appears to be a fair amount of variance in LSPI that is not explained by the viscosity. It is worth investigating whether the variance observed in LSPI can be explained by other variables related to the base stock. In the next section we consider the data from a multivariate perspective.


In this section we will describe several statistical models we created to describe our LSPI data. A statistical model of this sort is a useful tool for measuring the effect of multiple predictor variables (base stock properties) on a response variable (LSPI frequency). To be clear, the purpose of this activity is not to produce a mathematical expression that predicts LSPI performance, but rather to measure the influence of various basestock properties on LSPI.

A challenge that arises when studying base stock effects on lubricant performance is the natural correlation that exists between base stock properties. In base stock manufacturing processes, distillation of the oil streams affects both the oil's volatility and viscosity and therefore these properties are often correlated. Likewise, solvent extraction and hydrocracking refinery processes tend to reduce the concentration of aromatics and sulfur and increase the viscosity index (VI) of oil streams. These correlations make it difficult to measure the isolated effect of an individual base stock property on its performance. Therefore, one must exercise statistical judgment when analyzing base stock properties and performance.

Another decision the statistician must make is whether the model variables require a mathematical transformation. Sometimes a log or power law transformation is beneficial in producing models with normally distributed error. In this case we investigated a square root transformation of the LSPI Frequency. The P3 consortium identified this square root transformation as potentially appropriate. When we applied a square root transformation to our own data, we found the same conclusions. In this paper we present only the results for the untransformed LSPI Frequency for the sake of simplicity.

Table A2 lists the correlation coefficients between the variables in our study. The base stock property with the highest correlation to LSPI Frequency is viscosity. Therefore we decided that base oil mix KV100 should be included in a statistical model of our data that relates base stock properties to LSPI Frequency.

Given this decision, the logical next question is what other base stock properties affect LSPI Frequency. We chose to include the Viscosity Index in our statistical model. Viscosity Index has little correlation to KV100 so this property contains different information about the base stock composition than viscosity that is worth investigating. Table A3 and Figure 4 shows the estimates and quality of fit for this model. In this model the effect of base stock viscosity at 100[degrees]C is judged a statistically significant effect (p>=0.011) and the viscosity index not statistically significant (p=0.437). This model measures a strongly positive effect of base stock viscosity on LSPI Frequency, meaning that higher viscosity oils produce more pre-ignition events.

We tested other assumptions and measurements in the data analysis by constructing a variety of statistical models of the data, however the model presented in Table A2 and Figure 4 remained the analysis we chose as the most robust and explanatory. Though this model includes only two specific base stock properties, it is possible to interpret KV100 as a more general measure of the viscosity cut of the base stock that distinguishes light stocks from heavy stocks, and the viscosity index as a more general measure of the base stock quality (in other words, how highly refined it is). From this perspective, the analysis indicates that viscosity differences between base stocks drive larger differences in pre-ignition than differences in base stock quality among Group I-IV stocks within the range of our experimental testing.

We considered a second model including variables for KV100, Viscosity Index, and NOACK Volatility. NOACK Volatility is of interest because the engine oil's role in pre-ignition may involve an evaporation step. We can attempt to quantify this effect by including the NOACK variable in the statistical model. Unfortunately, adding NOACK to the model drastically changed the estimates for KV100 and Viscosity Index, and produced a Variance Inflation Factor [7] greater than 10. These are symptoms of too much multicollinearity in the model variables [7] which can produce misleading interpretations. Therefore we concluded from this exercise that it is impossible to reliably measure the isolated effects of these three variables in our dataset due to their correlations.

We considered a third model including variables for KV100, Viscosity Index, and base stock aromatics concentration. Base stock aromatics are of interest because we expect they should affect the combustion reaction rates of the engine oil. However the inclusion of base stock aromatics in the model left the conclusions unchanged; the KV100 effect remained statistically significant with a similar estimate, and the Viscosity Index and Aromatics effects are not statistically significant.

With these three models we examined the data from several different perspectives. Unfortunately we found it impossible to independently measure the effect of more than a few basestock properties on LSPI events. However we find strong evidence that base stock viscosity tends to increase the number LSPI events, and base stock quality's effect is less dramatic, and was not statistically significant in our data set.

The model described by Figure 4 exhibits a fair amount of error even after accounting for LSPI effects from base stock viscosity and base stock quality. It is possible this variance represents the experimental error in our apparatus and test methodology; it is always a challenge to achieve high precision in experiments in real engines. Alternately it is possible that some real variance in LSPI remains unexplained in this model--remember that the base stock properties we considered do not fully explain the variety of base stock composition. It is possible that base stock properties such as concentration of multi-ring aromatics, surface tension, or measures of volatility at different temperature cut points could explain some of the remaining variance. We did not investigate the effect of these properties in our analysis.


The mechanism of pre-ignition has received a fair amount of research in the past few years and several studies support a conceptual model of LSPI where injected fuel impinges on cylinder walls and mixes with the engine oil film coating that wall. The resulting mixture ejects into the combustion chamber and initiates combustion prematurely. It is now established that the engine oil plays a role in low speed pre-ignition in turbocharged gasoline direct injection engines, and different engine oils can show a variety of pre-ignition tendencies.

Although it is clear that engine oil formulation affects pre-ignition tendency, investigations into the specific effects of lubricant additives and base stock physical and chemical properties are still accumulating in the literature and have yet to reach consensus. This paper presented the results of an investigation into the role of base stock properties on engine oil pre-ignition tendency. We formulated 11 engine oils with various base stocks and a common additive system and measure pre-ignition tendency in a test engine. This is far from a comprehensive study however we made an effort to select Group I, II, III, and IV base stocks showing a variety of physical and chemical properties.

While it is difficult to measure the isolated effect of various base stock properties on performance due to the natural correlations between these properties, we presented evidence that base stock viscosity at 100[degrees]C has a statistically significant effect on LSPI and base stock quality (measured either by viscosity index or base stock aromatics concentration) does not. We measured a strong positive effect of viscosity in increasing the frequency of LSPI events, meaning that higher viscosities base stocks produced more preignition. Our measured effect of base stock viscosity agrees with that presented by Takeuchi et al. [3] Our measure of the effect of base stock quality on pre-ignition seems to agree with Welling et al. [4] but disagree with Takeuchi et al., who found less pre-ignition events for high quality Group III and IV base stocks.

In pursuit of an explanation for the base stock effects we measured in this paper, it is tempting to hypothesize a physical effect rather than a chemical one. The concentration of aromatics in the base stock should drive the most dramatic chemical reactivity differences between base stocks, however we could not measure a significant effect of base stock aromatics on LSPI. Attempting to build upon the conceptual model of LSPI offered in references [1] and [2], it is possible the influence of base stock occurs through its role in generating the lubricant films on the piston and cylinder surfaces. Higher viscosity lubricants produce thicker films, which could lead to a larger number of fuel-oil droplets produced when injected fuel impinges on the walls.

Both references [1] and [2] report that lowering the engine coolant

temperature increase the frequency of LSPI events. Because the coolant temperature lowers temperature at the cylinder walls in the vicinity of the lubricant film, and oil viscosity increases when cooled, it is possible this coolant phenomenon is related to the engine oil viscosity effect we measured in this paper.

We conclude with an appreciation for the complexity of LSPI and the recognition that our studies were limited to one particular engine, set of engine conditions, and engine oil additive technology. It is possible that different choices of test hardware or additive technologies might have affected the study's conclusions. We encourage further research in this area studying the interaction between oil quality, engine hardware, and engine conditions in LSPI.


The goal of this study was to measure the engine oil base stock effect on low speed pre-ignition. We describe an experimental apparatus to study LSPI in the laboratory consisting of a test engine operated at severe conditions and a data collection system to identify pre-ignition events. We selected 11 base stocks spanning a wide range of physical and chemical properties and blended these with a common set of lubricant additives. The engine oils formulated with the highest viscosity base stocks produced the highest frequency of LSPI events, and of the various physical and chemical properties of the base stocks, viscosity had the highest correlation with pre-ignition events. We considered several statistical models of our data aimed at identifying any other base stock effects on LSPI, but none were judged statistically significant. We recommend that researchers investigating LSPI in test engines be aware of a possible drift of test results as the engines age--we found it necessary to run a reference oil periodically in our study to benchmark engine LSPI drift over time and apply a correction factor to our results to compensate for this drift.

Arthur Andrews, Raymond Burns, Richard Dougherty, Douglas Deckman, and Mrugesh Patel

ExxonMobil Research & Engineering


[1.] Dahnz, C., Han, K., Spicher, U., Magar, M. et al., "Investigations on Pre-Ignition in Highly Supercharged SI Engines," SAE Int. J. Engines 3(1):214-224, 2010, doi:10.4271/2010-01-0355.

[2.] Zahdeh, A., Rothenberger, P, Nguyen, W., Anbarasu, M. et al., "Fundamental Approach to Investigate Pre-Ignition in Boosted SI Engines," SAE Int. J. Engines 4(1):246-273, 2011, doi:10.4271/2011-01-0340.

[3.] Takeuchi, K., Fujimoto, K., Hirano, S., and Yamashita, M., "Investigation of Engine Oil Effect on Abnormal Combustion in Turbocharged Direct Injection--Spark Ignition Engines," SAE Int. J. Fuels Lubr. 5(3):1017-1024, 2012, doi:10.4271/2012-01-1615.

[4.] Welling, O., Collings, N., Williams, J., and Moss, J., "Impact of Lubricant Composition on Low-speed Pre-Ignition," SAE Technical Paper 2014-01-1213, 2014, doi:10.4271/2014-01-1213.

[5.] "Lubricant Test Monitoring System: ASTM Test Monitoring Center Requirements for Engine Test Stand/Laboratory Calibration," Test Monitoring Center, A Program of ASTM International, accessed June 9, 2014,

[6.] "ASTM D7589--12a Standard Test Method for Measurement of Effects of Automotive Engine Oils on Fuel Economy of Passenger Cars and Light-Duty Trucks in Sequence VID Spark Ignition Engine," ASTM International, accessed March 3, 2011,

[7.] Kutner, M.H., Nachtsheim, J.N., Neter, J. Applied Linear Statistical Models, Fifth Edition. New York: McGraw Hill, 2005. Pp. 406-408.


Readers may contact Arthur Andrews via email at

or mail to ExxonMobil Research & Engineering

600 Billingsport Road

Paulsboro, NJ 08066


The authors thank Southwest Research Institute for their assistance and expertise in conducting the LSPI experiments documented in this paper. Specifically, we acknowledge Tom Briggs and Garrett Anderson for helpful discussions and coordination of the Pre-Ignition Prevention Program. The authors also thank Smruti Dance and Kevin Kelly (ExxonMobil Research & Engineering) for their helpful comments.


Table A1. This table shows our LSPI results and some properties of
the test oils. The test oils are fully formulated oils consisting of
a base stock or base stock blend, and lubricant additives. The
lubricant additive system (including viscosity modifiers) is
identical for all oils in this table. The listed properties (KV100,
Sulfur, etc.) are properties of the base stock, not the formulated
oil. KV100 is the kinematic viscosity of the base stock measured by
ASTM D445. We calculated the viscosity index by ASTM D2270. The base
stock sulfur concentration was measured by ASTM D2622. The base stock
aromatics concentration was measured by a combination of ASTM D2007
as well as a proprietary method appropriate for low aromatics stocks
by UV-Vis spectroscopy. The NOACK volatility was measured by ASTM
D5800. LSPI Frequency is defined by Equation 1. Engine Hours is a
measure of the time logged on the engine at the time of the engine
oil test. In reference to Figure 2, all these tests were run on
Engine 3. BS-9 is a Group III base stock produced with a
gas-to-liquids synthesis process. The other Group III base stocks in
the design are highly refined mineral oils. BS-8 appears as "Group
III" in the table but is actually a blend of a Group III stock with
a high quality Group II stock. The experiment on Oil BS-10 at 420
engine hours experienced failure in the pressure transducer so we
neglected its result in the analysis. To address this issue, we
re-ran the oil after the final reference oil run, which produced the
result at 652 engine hours.

Oil Code   Base Stock  KV100    Viscosity   Sulfur   Aromatics
           Group       (cSt)       Index    (ppm)    (mmol/kg)

BS-1       Group II    4.605         118        0        21.7
BS-10      Group IV    5.839         143        0         0.0
BS-10      Group IV    5.839         143        0         0.0
BS-11      Group II    6.011         111        4        37.7
BS-2       Group II    4.648         115       10       158.4
BS-3       Group I     4.169          99     2280       605.3
BS-4       Group I         4          97     4304       517.3
BS-5       Group III   4.433         145        0         0.3
BS-6       Group II    4.469         112        7       132.1
BS-7       Group IV    4.504         133        0         5.2
BS-8       Group III   4.204         116        0         1.2
BS-9       Group III   4.522         134        0         0.7

Oil Code       NOACK     Raw LSPI        LSPI    Engine
           volatility   Frequency    Frequency   Hours
               (wt%)     (events
                        per 25,000

BS-1            14.3         42.7        25.5      385
BS-10            5.2         24.4        15.9      420
BS-10            5.2         19.4        33.5      652
BS-11            8.7         51.0        29.2      368
BS-2            14.9         26.3        17.9      440
BS-3            24.8         20.8        18.7      531
BS-4            27.1         22.6        18.0      497
BS-5            12.2         28.1        27.1      547
BS-6            21.2         36.7        22.8      402
BS-7            10.1         20.4        17.3      516
BS-8            18.2         24.0        18.1      482
BS-9            11.1         24.1        17.3      468

Table A2. This table shows the correlation coefficients (r) between
properties of the base stocks, and also the measured LSPI Frequency
of the engine oil. The correlation coefficient varies between -1 and
1. A high positive value indicates the two variables (the rows and
columns of the table) are highly correlated. A low negative value
indicates the two variables are inversely correlated. A value of zero
indicates no correlation. The presence in this table of coefficients
of magnitude greater than 0.7 indicates that base stock properties
tend to be highly correlated.

                         KV100    Viscosity   Sulfur   Aromatics
                         (cSt)      Index     (ppm)    (mmol/kg)

KV100 (cSt)               1.000     0.326     -0.441    -0.446
Viscosity Index           0.326     1.000     -0.642    -0.753
Sulfur (ppm)             -0.441    -0.642      1.000     0.884
Aromatics (mmol/kg)      -0.446    -0.753      0.884     1.000
NOACK volatility (wt%)   -0.749    -0.800      0.752     0.819
LSPI Frequency            0.788     0.419     -0.334    -0.381

                           NOACK        LSPI
                         volatility   Frequency

KV100 (cSt)                -0.749       0.788
Viscosity Index            -0.800       0.419
Sulfur (ppm)                0.752      -0.334
Aromatics (mmol/kg)         0.819      -0.381
NOACK volatility (wt%)      1.000      -0.566
LSPI Frequency             -0.566       1.000

Table A3. This table shows the statistics model we chose to analyze
the experimental data. We chose to represent the LPSI results with a
linear model including terms for base stock KV100 and base stock
Viscosity Index. The KV100 effect is statistically significant
(p=0.011) while the Viscosity Index effect is not (p=0.437). The
Viscosity Index term in this model acts as a surrogate for base stock
quality--it is impossible to measure the individual effects of base
stock sulfur, NOACK volatility, and aromatics due to collinearity.
In these tables, columns headed by the acronym "VIF" list variance
inflation factors.

Summary of Fit

RSquare              0.650
RSquare Adj          0.563
Root Mean Square      3.77
Mean of Response      22.3
Observations (or        11
  Sum Wgts)

Analysis of             DF      Sum of      Mean     F Ratio
  Variance                     Squares    Square

Model                    2       211.4     105.7       7.443
Error                    8       113.6      14.2    Prob > F
C. Total                10       325.0              0.0149 *

  Estimates Term   Estimate   Std Error   t Ratio      Prob>     VIF
                                                     of (t)]

Intercept           -15.18       10.74     -1.41      0.1953
KV100 (cSt)         6.3861       1.937      3.30    0.0109 *    1.119
Viscosity Index    0.06359     0.07770      0.82      0.4368    1.119
COPYRIGHT 2016 SAE International
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2016 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Andrews, Arthur; Burns, Raymond; Dougherty, Richard; Deckman, Douglas; Patel, Mrugesh
Publication:SAE International Journal of Fuels and Lubricants
Article Type:Technical report
Date:Jun 1, 2016
Previous Article:Performance and emission characteristics of CI engine operated on Madhuca Indica biodiesel using multi-objective optimization by applying Taguchi...
Next Article:Review: fuel volatility standards and spark-ignition vehicle driveability.

Terms of use | Privacy policy | Copyright © 2020 Farlex, Inc. | Feedback | For webmasters