Downloaded from SAE International by University of Liverpool, Monday, August 13, 2018
2003-01-0725
An Improved Friction Model for Spark-Ignition Engines
Sloan Automotive Laboratory, Massachusetts Institute of Technology
Copyright © 2003 SAE International
ABSTRACT
A spark-ignition engine friction model developed by
Patton et al. in the late 1980s was evaluated against
current engine friction data, and improved. The model,
which was based on a combination of fundamental
scaling laws and empirical results, includes predictions
of rubbing losses from the crankshaft, reciprocating, and
valvetrain components, auxiliary losses from engine
accessories, and pumping losses from the intake and
exhaust systems. These predictions were based on
engine friction data collected between 1980 and 1988.
Some of the terms are derived from lubrication theory.
Other terms were derived empirically from
measurements of individual friction components from
engine teardown experiments. Recent engine
developments (e.g., improved oils, surface finish on
piston liners, valve train mechanisms) suggested that
the model needed updating. So modifications were
made to the piston ring tension and gas pressure
loading contributions to piston assembly friction, the
impact of liner roughness, and to the valvetrain
mechanism friction. Lubricant viscosity scaling was
added to the hydrodynamic terms in the rubbing friction
component models. The improved model now gives
reasonable estimates of individual friction component
and total spark-ignition engine friction mean effective
pressure. The inclusion of oil viscosity scaling with
temperature results in cold engine friction predictions
about twice the value for warmed up engines. This
agrees with the limited cold engine friction data available
in the literature.
INTRODUCTION
There is strong interest in improving spark-ignition
engine fuel consumption. Thus, accurate predictions of
the major components of engine friction are important.
However, models of engine friction are difficult to
validate since the total friction loss is the summation of
losses arising from many components in the engine.
These friction components respond differently to
variations in cylinder pressure, component
temperatures, and speed as engine load varies [1].
Daniel Sandoval and John B. Heywood
The engine friction model assessed and then improved
as described in this paper, was developed by Patton et
al. [2] to predict spark-ignition engine friction using basic
engine design and operating parameters as inputs.
However, the data used to develop this model date back
to the mid-1980s. It was thus important to compare this
model’s predictions to current engine friction data, and
determine the changes that needed to be made to
certain terms in this friction model. It was found that the
pumping losses over the intake and exhaust strokes,
and the terms representing the cylinder gas pressure
loading effects needed to be modified, and appropriate
adjustments were made. We have also expanded the
model to include lubricant viscosity as one of the model
variables, so that the effects of component temperatures
on engine friction during cold start and warm-up
transients can be predicted.
This paper is organized as follows. First, a brief
overview of the components of the Patton et al. engine
friction model is provided. Details of the original and
improved models are then presented to show and
explain the modifications that have been made to
specific model terms. The “old” and “new” model are
compared. Model predictions of friction components are
then compared to measured engine teardown friction
data from three current engines. The effects of
changing coolant temperature on model predictions,
from start-up to warmed-up conditions, are then
examined. The paper concludes with a summary of the
improvements made to the model.
ENGINE FRICTION BACKGROUND
Engine friction losses are usually divided into three main
categories: mechanical or rubbing losses, pumping
losses, and auxiliary component losses. Mechanical or
rubbing losses are those which result from relative
motion between solid surfaces in the engine; e.g., the
motion between a piston ring and a cylinder wall, or a
crankshaft journal in a bearing. Relative motion does
not require that the two solids be in contact with each
other; usually there is lubricant separating the surfaces.
Pumping losses are the work done by the piston as
gases are pushed out of and pulled into the cylinder
during the exhaust and intake strokes. In naturally-
aspirated engines this pumping work is the result of flow
resistances and, therefore, is a friction loss. Auxiliary
Downloaded from SAE International by University of Liverpool, Monday, August 13, 2018
Downloaded from SAE International by University of Liverpool, Monday, August 13, 2018
Downloaded from SAE International by University of Liverpool, Monday, August 13, 2018
Downloaded from SAE International by University of Liverpool, Monday, August 13, 2018
 Downloaded from SAE International by University of Liverpool, Monday, August 13, 2018

Table 2: Major Engine Parameters

Engine Type Current Current Prototype

Engine
Enaine Information
Layout/ # Cylinders V6
2 4
Valvetrain Type SOHC DOHC SOHC
Roller
Valvetrain Mechanism Direct Acting Finger Roller Finger Follower
Follower
Ratio 0.85 1.1 0.9
Compression Ratio 9.0 10.0 11.75

3ULRU0RGHO ,PSURYHG0RGHO parameters used in the friction model for these three
engines and The two current
 production engines were tested by motoring. This gave
 component friction data that allowed us to modify and
 compare the results for each component of the model.
 The prototype engine was used to compare the

pumping loss model at varying intake pressures. The

engines were tested at an engine coolant temperature of

The rest of this section shows the results for the
improved friction model.



CRANKSHAFT FRICTION
U\
LQ
RQ

HS
DV

HS
IW

HS
KD

Figure 2 shows that the improved friction model gave the

UD

OLD
J
VW

SP

WIP
P
HW
NV

[L
SL

RQ

WP
OY

DX
DQ

same trends but significantly under predicted the

VW

YD
SL
FU

crankshaft friction measurements for both engines. The

Figure 1: Component and total friction mep data (in crankshaft model terms may oversimplify the complexity
for engine at 2500 rpm - WOT, comparing original of crankshaft friction, since about 80% of the friction was
and improved models. accounted for by the hydrodynamic and turbulent
dissipation terms. also may be that testing the
No large differences are seen in the mechanical friction crankshaft without the pistons and connecting rods,
though with equivalent weights added to the crank
components (crankshaft, piston, and
throw, significantly increased the friction under these
valvetrain) since the viscosity scaling was only included in
conditions.
the hydrodynamic terms which make a small contribution
to the total friction for each of these components under
warmed-up engine conditions. However, there is a
'DWD/ 0RGHO/
difference of about 6 in the pressure term 'DWD/ 0RGHO/
due to the piston ring tension, surfaces roughness, and 
mixed lubrication factors that were updated in that model.

The auxiliary losses are closely comparable since the
new model only changes afmep at higher speeds. For 
the pumping model, the change made to the valve
N3D


pressure drop constant decreased the pumping mep 
loss by 25


DETAILED RESULTS 
    

ENGINE PARAMETERS USP

The modifications to the friction model were based on Figure 2: Comparison of crankshaft friction mep
friction data from two current production engines and a predictions and data for engine and engine.
prototype engine. Table 2 lists the key engine
 Downloaded from SAE International by University of Liverpool, Monday, August 13, 2018

RECIPROCATING FRICTION Table 3: Constants for Valvetrain Mechanism Terms

TYPE Cam Oscillating Oscillating
Terms without Gas Pressure Loading
Follower Hydrodynamic Mixed
Flat Roller
Figure 3 shows the friction losses for the piston group
SOHC
(piston skirt, piston rings, and connecting rod) without
finger 600 0.2 42.8
the gas pressure loading. The model for the 3.0L engine
follower 0.0227
matched well over the low engine speed range and then
at mid- to high-speeds rose to be a consistent amount SOHC 400 0.5 21.4
above the data. At high speeds, the difference between rocker 0.0151
the model and data was about 13 kPa (18%). For this arm
model, the piston ring hydrodynamic term dominated the
friction contribution. The model for the 5.4L engine SOHC 200 0.5 10.7
matched the measured data at low, and at high speed. direct 0.0076
The piston ring hydrodynamic term dominates the acting
reciprocating friction contribution due to the increase in
mean piston speed: see Eq. 7. DOHC* 600 0.2 25.8
finger 0.0227
Terms with Gas Pressure Loading follower

Data was not available to compare the gas pressure DOHC 133 0.5 10.7
loading effect on the reciprocating friction component direct 0.0050
since the data were collected with the cylinder head and acting
valvetrain mechanism removed.
OHV 400 0.5 32.1
VALVETRAIN FRICTION
*
Figure 4 shows valvetrain friction losses for the 3.0L and Determined by matching engine data for DOHC
5.4L engine, respectively. In the 3.0L engine, the model finger follower valvetrain mechanisms.
over predicted the valvetrain friction by about 20%, while
in the 5.4L engine, the model and the valvetrain friction 
data are in good agreement. Both models did maintain AUXILIARY FRICTION
a similar trend to the engine friction data. Note the
model had to predict friction for different types of Figure 5 shows the accessory friction losses for the 3.0L
valvetrain and interfaces. Table 3 shows the constants and 5.4L engine, respectively. The auxiliary friction
available for the different valvetrain mechanisms model indicates no difference between the engines since
commonly used in engines. The original Patton et al. accessory mep losses were based only on engine
constants were used for the 5.4L engine predictions. A speed. The modified auxiliary friction model fits the
set of constants was added to the original table of engine data well.
coefficients to be able to model the DOHC finger roller
follower of the 3.0L engine. These constants were
obtained by matching the model to the newer valvetrain
'DWD/ 0RGHO/
friction data. The first three constants were kept the
'DWD/ 0RGHO/
same as the constants for a SOHC finger follower

configuration. The oscillating mixed constant was set to
0.6 times the constant for the SOHC finger follower 
since this gave the best fit between data and model.

N3D

For the 5.4L engine, the difference in valvetrain friction 

at low engine speeds between the model and data may

result from the decrease made in the boundary
lubrication term factor of (1 + 1000/N) to (1 + 500/N). 
However, the data between 2000 and 5500 rpm are 
matched well.      
USP

Figure 3: Comparison of predictions of reciprocating

friction mep without loading to data, for 3.0L engine and
for 5.4L engine.
Downloaded from SAE International by University of Liverpool, Monday, August 13, 2018
 Downloaded from SAE International by University of Liverpool, Monday, August 13, 2018

/93XPSLQJ/RVVHV 
0RGHOYV'DWD 


'DWD0RWRUHGZLWK3LVWRQ*DV/RDGLQJ

0RGHO0RWRUHGZLWK3LVWRQ*DV/RDGLQJ 
'DWD
'DWD
  'DWD
 'DWD
 'DWD


)0(3N3D



N3D

 


 



          
 0$3N3D
          
USP

Figure 10: Comparison of predictions of pumping mep

with data, as a function of engine load, for engine.
Figure 8: Comparison of predictions of total mechanical
friction mep with piston gas loading, with data, for
engine.
ENGINE FRICTION DURING WARM UP
'DWD0RWRUHGZLWK*DV([FKDQJH/RVVHV
0RGHO0RWRUHGZLWK*DV([FKDQJH/RVVHV The viscosity scaling included in the new model allows
us to examine the effect of oil temperature on the

different components of engine friction. Shayler et al.
 [7] included a modification in the previous Patton et al.
model to account for the increased friction losses which

arise when the engine is cooler than fully warmed-up.
These additional losses were attributed primarily to the
N3D


effect of temperature on oil viscosity. In their tests, the


engine was allowed to soak down to room temperature
(20oC) prior to each fired friction test. A plot of friction

power loss against time from engine start-up was
 obtained. Several similar tests were carried out and
       these indicate that the friction power loss at engine start-
USP up is some two times higher than the warmed-up engine
friction power loss.
Figure 9: Comparison of mep predictions of total
mechanical and gas exchange losses, with data, for With the modified friction model, predictions can now be
engine. made for a given oil, and oil temperature. Figure 11
shows the model friction mep predictions as a function of
temperature. Note that the temperature used to
PUMPING LOSSES
calculate oil viscosity is the coolant temperature.
Figure 10 shows the pumping losses for the Comparing the total mechanical friction at 20 oC and at
90oC shows a cold engine friction power loss of about
engine. These data were used to evaluate the intake
2.1 times the hot engine friction power loss. This is
and exhaust friction mean effective pressure models.
close to the ratio determined by Shayler et al. from their
The total losses are the sum of the intake and
firing engine warm-up tests.
exhaust system, port and valve pressure drops. The
pmep data are plotted against increasing intake manifold
pressure (MAP) at constant speeds. The model shows SUMMARY
the same trends for pmep with increasing speeds and
increasing MAP as the data. The largest difference This paper documents improvements made to a
between the model and the data is at 1000 For multicylinder water-cooled spark-ignition engine friction
increasing speeds, this difference decreases model developed by Patton et al. in the late to
significantly and for 3000 the model matches improve the predictive accuracy of the original model.
the data well. At higher speeds and MAP, the exhaust Engine friction has decreased significantly over the past
system, and both intake and exhaust valve pressure 15 years. These model improvements were based on
drop terms, contribute significantly to the pumping friction data sets from three modern engines. The
losses. modifications were made assuming that the friction work
 Downloaded from SAE International by University of Liverpool, Monday, August 13, 2018

components fell into three main categories: 5. Chevron Supreme Motor Oil.
independent of speed (boundary lubrication), http://library.cbest.chevron.com/lubes.
approximately proportional to speed (hydrodynamic
friction), or to speed squared (turbulent dissipation). 6. Taylor, R. I., Brown, M.A., Thompson, D. M., and
Modifications were also made to the hydrodynamic terms Bell, J. C., 1994, The Influence of Lubricant
to include their dependence on differences in oil grades Rheology on Friction in the Piston Ring-Pack, SAE
and temperature. Changes to some of the boundary Paper 941981.
friction and turbulent dissipation terms were made
knowing that engine design improvements have lowered 7. Shayler, P.J., Christian, S.J., and Ma, T., A Model
these specific friction losses in modern engines. for the Investigation of Temperature, Heat Flow and
Friction Characteristics During Engine Warm-up,
Based on our evaluation of the improved model against SAE Paper 931153.
modern engine data, the following conclusion can be
drawn:

1. The experimental data suggest that total engine 7RWDO0HFKDQLFDO 5HFLSQRJDV

5HFLSJDV &UDQNVKDIW
friction has decreased by some as engine 
design has improved. Major areas where model

changes which reduce friction were needed were

piston friction and pumping losses.


2. Useful predictions of crankshaft, reciprocating, 

valvetrain, and auxiliary friction and pumping losses

N3D

for modern gasoline engines can now be made. The 
crankshaft friction model still gives significant 
differences between model and data, under 
predicting crankshaft friction. Since the other friction

components were more accurately predicted, and

crankshaft friction is relatively small, the sum of all the         
friction components gave an accurate prediction of :2LO7HPSHUDWXUH&
the total engine friction.

3. Adding oil viscosity scaling with square root of

temperature, to the hydrodynamic friction 7RWDO0HFKDQLFDO 5HFLSQRJDV
5HFLSJDV &UDQNVKDIW
component terms, allowed the friction losses from an

engine as a function of coolant temperature to be
assessed. The model predicted that the total 

mechanical friction mep at 20°C is about 2 times the 

friction loss at 90°C. This ratio is comparable to 

limited data available from firing engine warm-up 

N3D

tests. 

REFERENCES 

1. Wakuri, Y. et. al., Studies on Friction Characteristics 
of Reciprocating Engines, SAE Paper 952471. 
        

2. Patton, K. J., Nitschke, R. G., and Heywood, J. B., :2LO7HPSHUDWXUH&

Development and Evaluation of a Performance and

Efficiency Model for Spark-Ignition Engines, SAE
Paper 890836; SAE Trans., Vol. 98, pp. 1441-1461, Figure 11: Friction mep for varying oil temperatures
1989. (assumed equal to coolant temperature) at 2000 rpm for
engine (top) and engine (bottom).
3. Heywood J.B., “Internal Combustion Engine
Fundamentals” McGraw-Hill, New York, 1988, pgs.
712-745.

4. William, J.A., “Engineering Tribology” Oxford

University Press, New York, 1994, pg. 241.
Downloaded from SAE International by University of Liverpool, Monday, August 13, 2018
Downloaded from SAE International by University of Liverpool, Monday, August 13, 2018