Calculation of Friction in High Performance Engines

Calculation of Friction in High Performance Engines Ricardo Software European User Conference 21 Phil Carden Ricardo UK www.ricardo.com

Presentation contents Introduction Some general comments and issues Semi-empirical models FAST Piston assembly friction PISDYN and RINGPAK Fluid film bearing friction ENGDYN Crankcase pumping loss WAVE Valve train and timing drive friction VALDYN Case study Conclusions 2

Introduction Minimising friction in motorsport engines is vital because power is required elsewhere high friction often means high wear high friction leads to unnecessary heat generation and greater need for cooling Knowledge of friction level is required to evaluate potential of alternative low friction designs, coatings etc as input to performance simulation models Measurement of engine friction is highly problematic, particularly at high engine speed the results are often obtained too late to make key design changes but measurements are useful for validation of calculation methods Analytical tools are beginning to reach required level of fidelity but there are still some problems and mysteries 3

Which losses are included in engine friction? Piston ring/liner friction Piston skirt/liner friction Small end bearing friction Big end bearing friction Main bearing friction Thrust bearing friction Valve train friction Camshaft bearing friction Timing drive friction Balancer bearing friction Seal friction Definitely included Auxiliaries friction (oil pump, water pump, alternator, scavenge pumps etc) Optional Losses due to pumping of crankcase gas below pistons Windage losses in engine Oil churning losses in engine Power required to drive auxiliaries (oil pump, water pump, alternator) Transmission losses (gears, bearings, seals, windage, oil churning etc) Definitely not included Pumping of gas above piston 5

Stribeck curve Friction coefficient at any lubricated contact depends on relative speed, normal load, oil viscosity, shape of contacting parts, roughness of contacting parts, temperature, materials of contacting parts etc Boundary lubrication.22.2 Mixed lubrication Elastohydrodynamic lubrication Hydrodynamic lubrication.18 Cam/follower friction coefficient.16.14.12.1.8 Piston rings / liner Crankshaft bearings.6.4.2 2 4 6 8 1 12 14 16 18 2 Duty parameter (viscosity x speed / load) or specific film thickness (film thickness / surface roughness) 6

How to measure friction? Description Issues Motored test Simply connect engine to motor Drive the engine without firing and measure the torque Need to control temperature of oil and water carefully Test can be extended by progressively stripping engine to give approximate breakdown in friction by subsystem When the whole engine is motored then pumping loss cannot be separated from friction by measurement No combustion so gas load due to compression only No way to account for increased local temperatures due to combustion Fired test Operate engine normally Measure cylinder pressure (with very good knowledge of TDC) and brake torque as accurately as possible Calculate FMEP from IMEP BMEP Can extend test to make more detailed measurements FMEP is small compared to IMEP and BMEP so any measurement error is magnified Use of more intrusive instrumentation usually involved changes to components and this can affect durability 7

What kinds of calculation are performed? Advantages Semi-empirical approach Very quick to set up and run Can potentially account for most sources of loss Advanced CAE approach Potentially good accuracy for any contact if suitable model is available Can account for differences in load, detailed component geometry, oil grade, temperature etc Limitations Need lots of test data to develop tool Accuracy depends of how similar engine is to those in database Cannot accurately predict effect of detail design changes Potentially long set up times Potentially long run times Need different tools to analyse all contacts Some losses are very difficult to model Easy to lose sight of big picture 8

Some special problems of motorsport engines Friction testing usually calls for prolonged operation under stabilised conditions Many motorsport engines cannot be operated under stabilised conditions at high speed/load During a race the piston temperature varies significantly and can approach the limit at end of straights after ~15 seconds If the engine is held at full load the piston temperature will rise significantly affecting clearances, local oil temperature and so piston friction before failure Motorsport engines are often not run-in properly before use In typical passenger car engine friction reduces from initial value for at least 2-3 hours Motorsport engines are sometimes used straight from new build so friction level is reducing all the time average piston temperature (deg C) FMEP (bar) 45 4 35 3 25 2 15 1 5 2.2 2 1.8 1.6 1.4 1.2 1.8.6.4.2 Simulation of 1 lap at Monza 1 2 3 4 5 6 7 8 9 1 11 time (sec) Measurement on passenger car engine at 4 rpm 5% reduction in 4 hours 5 1 15 2 25 3 35 4 45 5 55 6 Time (hours) 9

Semi-empirical models - FAST FAST (Friction Assessment Simulation Tool) is a semi-empirical engine friction prediction program used internally by Ricardo FAST predicts friction loss across the engine speed range for individual subsystems and for the whole engine FAST uses semi-empirical equations Coefficients and exponents were developed by Ricardo to give improved agreement with the latest measured data obtained from motored teardown tests on modern passenger car engines and with other published measured data New equations were developed for systems not covered by published sources Latest version extended to enable prediction of friction under fired conditions FAST currently requires ~5 input values such as bore, stroke, bearing diameter, valve train type etc. to make a prediction of motored friction loss All input values can be obtained easily from drawings or components More detailed information such as cylinder pressure, piston mass, rod mass, ring tension etc. is required to calculate the increase in piston friction under fired engine conditions 11

Typical results for reciprocating group Predictions under motored conditions compared with measured data from motored teardown test Prediction under full load conditions shows significant increase in friction at top piston ring and skirt Gasoline reciprocating group friction - motored (bar) 1.9.8.7.6.5.4.3.2.1 CPMEP FIRST RING SECOND RING OIL CONTROL RING PISTON SKIRT BIG END BEARING MEASURED Gasoline reciprocating group friction - full load (bar) 1.9.8.7.6.5.4.3.2.1 CPMEP FIRST RING SECOND RING OIL CONTROL RING PISTON SKIRT BIG END BEARINGS 1 15 2 25 3 35 4 45 5 Engine speed (rpm) 55 6 65 7 1 15 2 25 3 35 4 45 5 Engine speed (rpm) 55 6 65 7 12

Typical results for other systems V6 gasoline crankshaft group friction (bar).35 WINDAGE Results.3 SEALS.3 accurate to MAINS BEARINGS.25 MEASURED DATA.25 +/-2% for.2.2 each.15 subsystem.1 and often.15.1 much better than this.5.5. 1 15 2 25 3 35 4 45 5 55 6 65 Engine speed (rpm) V6 gasoline valvetrain group friction (bar).35 TIMING BELT DIRECT ACTING HYDRAULIC VALVETRAINS CAMSHAFT SEALS CAMSHAFT BEARINGS MEASURED DATA 1 15 2 25 3 35 4 45 5 55 6 65 Engine speed (rpm) I4 gasoline oil pump friction (bar).2.18.16.14.12.1.8.6.4.2 FAST V2 Measured data 1 15 2 25 3 35 4 45 5 55 6 Engine speed (rpm) I4 gasoline alternator friction (bar).2.18.16.14.12.1.8.6.4.2 FAST V2 Measured data 1 15 2 25 3 35 4 45 5 55 6 Engine speed (rpm) 13

Typical results for high performance sports car engine For whole engine friction the results are generally accurate to within +/-1% FAST has not been validated for engines designed to rev higher than 8 rpm Whole engine - no load (bar) 2.4 2.2 2 1.8 1.6 1.4 1.2 1.8.6.4.2 1 Auxiliaries group Crankshaft group Valvetrain group Reciprocating group 15 2 25 3 35 4 45 5 55 6 Engine speed (rpm) 65 7 75 Whole engine - full load (bar) 2.4 2.2 2 1.8 1.6 1.4 1.2 1.8.6.4.2 8 1 Auxiliaries group Crankshaft group Valvetrain group Reciprocating group 15 2 25 3 35 4 45 5 55 6 Engine speed (rpm) 65 7 75 8 14

Piston assembly friction prediction Piston assembly friction involves losses at many sliding interfaces is very difficult to measure even if cost is not an issue Analysis can offer insights not possible by measurement Need for validated software but Losses and rings and skirt are inter-related Some key inputs are hard to determine Temperature of surfaces and oil Film thickness on liner Worn surface shapes of skirt and rings rings The lubrication regime, and so the effective friction coefficient,, can change dramatically during each half stroke skirt boundary at end of stroke ( = ~.1) hydrodynamic at mid stroke ( = ~.5) 16

Ricardo Software - PISDYN and RINGPAK PISDYN can calculate friction loss due to oil shearing effects and asperity contact at skirt/liner contact at small end bearing piston secondary motion wear loads Top Ring Groove Friction Force Crown Contact and Friction Asperity Contact Skirt Oil Film Hydrodynamics (Pressure) Cylinder Dynamics Secondary Motions Lubrication Elastic Deformation Cut Outs Bore Distortion Crown Land Profiles Bore Temps. (Viscosity effect) Thermal / Pressure Inertial Skirt Deformation RINGPAK can calculate friction loss due to oil shearing effects and asperity contact at each ring/liner contact wear loads oil consumption blow-by and blow-back Dispersed oil flow Ring dynamics Piston tilt Pin bearing eccentricity Cylinder oil evaporation Ring lubrication, friction Inter-ring gas flows Ring - bore conformability Oil supply 17

Model validation at low engine speed Predictions validated at low engine speed (up to 25 rpm) using force difference method Measured cylinder pressure, connecting rod strain, crankshaft angular position, liner temperature, piston temperature etc. Piston friction force calculated from gas force, inertia force and rod force Future plans for validation at higher speed as part of ENCYCLOPAEDIC project Results at 2 rpm full load 18

Experience at high speed Ricardo have used PISDYN and RINGPAK to calculate piston friction at very high engine speeds Credible ring/liner friction results obtained from RINGPAK But at high speed losses are dominated by piston/skirt asperity contact which is predicted to occur in each stroke even for a well-developed piston skirt profile Skirt/liner friction power loss shows strong sensitivity to component temperature governs clearance, component distortion, and local oil temperature oil supply assumptions component surface texture rings rings skirt 19

High Speed Case Study Predicted result shows asperity contact pressure at 15 rev/min Wear region above rings Each skirt has patches of wear, either side of the central region And Centrally under the rings 2

Presentation contents Introduction Some general comments and issues Piston assembly friction PISDYN and RINGPAK Fluid film bearing friction ENGDYN Crankcase pumping loss WAVE Valve train and timing drive friction VALDYN Case study Conclusions 21

Fluid film bearing friction Journal bearings normally operate in the hydrodynamic lubrication regime Losses are dominated by oil shearing effects Relatively easy to quantify using simple bearing analysis methods Can be used to study the effects of bearing dimensions, clearance, oil viscosity etc Surface contact (and so boundary or mixed lubrication) is possible during running-in process or if bearing is overloaded Friction prediction is theoretically possible using elasto-hydrodynamic analysis with asperity contact model but Measured worn axial bearing profiles and distorted shapes are needed for sensible results so not very useful for design 1 2 3 4 5 6 Total Main Bearing FMEP (bar).3.25.2.15.1.5. 4 6 8 1 12 14 16 18 No axial profile Bearing wear depth (µ) Engine Speed (rpm) Main 2 Oil A Oil B Axial profile rear Bearing axial position (mm) front 22

Camshaft bearing friction effect of ENGDYN model level Front camshaft bearings often exhibit edge wear due to loads from the timing drive This wear usually occurs during run-in period and then stabilises ENGDYN used to predict friction at camshaft bearings Simple rigid model gives power loss due to oil shear effects With flexible camshaft model edge contact leads to high friction loss When worn profile is introduced friction loss is similar to results from simple model power loss (W) 1 9 8 7 6 5 4 3 2 statically determinate loading 1 dynamic loading dynamic loading - with axial profile 6 8 1 12 14 16 18 2 camshaft speed (rpm) 23

Crankcase pumping loss - WAVE Crank case pumping incurs a power loss Power required to pump crankcase gas from one bay to the next as the pistons move up and down This can be significant loss on high speed engines so dry sump designs with scavenge pumps are often used to minimise this Ricardo and others have used 1D gas dynamics modelling software such as WAVE to predict crank case pumping loss with some success Complex geometry of crank case from CAD and then automatically reduced to 1D network 25

Crankcase pumping loss - WAVE BREATHER AREA CPMEP > Critical Breather Area Losses reduce with increase in area Max CPMEP at a Critical Breather Area Reduced resistance to inter-bay flow exchange -.2 -.4 -.6 < Critical Breather Area Losses increase with increase in area Increasing inter-bay flow exchange Reduced recovery of compression work CPMEP (bar) -.8 -.1 -.12 -.14 7 85 11 -.16 135 16 -.18 3 6 9 12 15 EFFECTIVE BREATHER AREA (mm 2 ) 26

Crankcase pumping loss - WAVE Inter-bay breathing holes removed Single external breather exit through balancer shaft center 4 scavenge points ENGINE SPEED (rpm) 6 8 1 12 14 16 18 2 22.. -.5 -.5 -.1-1. Wet Sump CPMEP (bar) -.15 -.2 -.25 -.3 -.35.38 bar 4.7 kw -1.5-2. -2.5-3. -3.5 POWER (kw) Dry Sump -.4-4. -.45-4.5 -.5-5. 27

Valve train friction prediction At system level valve train friction is reasonably well-understood The mean friction level of valve trains can be measured by driving the camshaft using an electric motor and measuring mean drive torque However These system level measurements mask the fact that the detailed sources of loss are not very well understood Especially since the measurements often include the timing drive as well as the valve train Motorsport engines nearly always have sliding contacts between cam and tappet to enable use of a lightweight follower and optimise high speed dynamics Friction loss is dominated by losses.1 at the these highly-loaded sliding contacts FMEP (bar).6.5.4.3.2 total gear losses camshaft bearings tappet/bore contacts cam/tappet contacts 9 1 11 12 13 14 15 16 Engine speed (rpm) 29

Cam/follower friction sliding contact VALDYN can calculate friction power loss at sliding contacts between cam/tappet or cam/follower Friction due to oil shear and asperity contact are calculated separately Friction is therefore dependent on oil viscosity as well as surface texture The model also considers friction between tappet and cylinder tappet rotation.6 Mean Effective Friction Coefficient.5.4.3.2.1 Oil A Oil B. 4 6 8 1 12 14 16 18 Engine Speed (rpm) 3

Timing drive friction - VALDYN VALDYN has the capability to calculate the losses in chain timing drives Models have been validated validated against measured data at passenger car engine speeds Results can be very dependent on tensioner design power loss (W) 18 16 14 12 1 8 6 4 2 link/guide roller/sprocket pin/bush cam bearings idler bearings measured 1 15 2 25 3 35 4 45 5 55 engine speed (rpm) 31

Timing drive friction - VALDYN VALDYN has gear models for calculation of contact forces and sliding speeds Friction coefficients can be entered to calculate losses at gear meshes But actual losses in high speed gears are mostly caused by windage and oil churning effects which cannot be modelled with confidence VALDYN can model belt drive dynamics but detailed sources of loss are not understood so cannot be predicted Belt material hysteresis? Rubbing losses between teeth and pulleys? Pumping air out during meshing? Camshafts Tensioner Crankshaft 32

Case study The following slides illustrate the use of CAE tools and the results of motored teardown test to understand the distribution of friction in a racing engine The engine was motored and a 4-stage teardown test was made 1 Whole engine with intake and exhaust systems 2 Intake and exhaust systems removed 3 Cylinder head removed including upper timing gears Cylinder head removed by plate to maintain bolt loads on structure Plate was non restrictive so no pumping losses 4 Pistons and rods removed Replaced by dummy weights All stages include driven oil pump, water pump and transmission 34

Case study - Valve train group friction Measured data obtained by subtraction Test 2 Test 3 Pumping loss under motored conditions calculated using WAVE FMEP (bar) 2 1.8 1.6 1.4 1.2 1.8.6 pumping loss total gear losses camshaft bearings tappet/bore contacts cam/tappet contacts measured data Camshaft bearing friction calculated using ENGDYN.4.2 Cam/tappet losses and tappet/bore losses calculated using VALDYN 9 1 11 12 13 14 15 16 Engine speed (rpm) 2 1.8 1.6 total gear losses Gear contact losses estimated using loads from VALDYN and friction coefficients Windage losses estimated FMEP (bar) 1.4 1.2 1.8.6.4.2 camshaft bearings tappet/bore contacts cam/tappet contacts measured data minus predicted PMEP 9 1 11 12 13 14 15 16 Engine speed (rpm) 35

Case study - Reciprocating component group friction Measured data obtained by subtraction Test 3 Test 4 Piston skirt friction loss and small end bearing loss calculated using PISDYN Piston rings friction loss calculated using RINGPAK FMEP (bar) 2 1.8 1.6 1.4 1.2 1.8 piston skirt oil control rings top piston ring crankcase pumping small end bearings big end bearings measured data Crankcase pumping loss calculated using WAVE.6.4.2 Big end bearing loss calculated using ENGDYN 9 1 11 12 13 14 15 16 Engine speed (rpm) 36

Crankshaft/transmission group friction Measured data obtained directly Test 4 Main bearing loss calculated using ENGDYN Windage losses estimated using simple approach Gear contact losses estimated using simple approach Oil pump losses and water pump losses estimated FMEP (bar) 2 1.8 1.6 1.4 1.2 1.8.6.4.2 9 1 11 12 13 14 15 16 Engine speed (rpm) water pump gear windage oil pump seals Rolling element bearing losses calculated using calculated loads and friction coefficients balancer windage rolling element bearings gear meshes crankshaft windage main bearings measured data Seals losses estimated using suppliers database 37

Whole engine motored friction loss (1) With motored pumping losses included 5 4.5 pumping loss timing gears camshaft bearings Additional loads at main bearings due to inertia of pistons and rods accounted Allowance in calculations for additional loads at pistons and bearings due to motored gas force with cylinder head on FMEP (bar) 4 3.5 3 2.5 2 1.5 tappet/bore contacts cam/tappet contacts piston skirt oil control rings top piston rings crankcase pumping small end bearings big end bearings water pump gear windage oil pump seals balancer windage 1 rolling element bearings gear meshes.5 crankshaft windage main bearings measured data 9 1 11 12 13 14 15 16 Engine speed (rpm) 38

Whole engine motored friction loss (2) Same data but with results grouped by subsystem Pumping losses excluded At 15 rpm Crankshaft 25.1% Transmission 7.9% Pumps 11.4% Crankcase pumping 8.5% Pistons 33.9% Valve trains 13.2% FMEP (bar) 5 4.5 4 3.5 3 2.5 2 1.5 valve trains and drive piston assemblies crankcase pumping oil and water pumps transmission (inc balancer) crankshaft measured minus pumping loss 1.5 9 1 11 12 13 14 15 16 Engine speed (rpm) 39

Full load fired friction loss by analysis Next step was to use CAE tools to model the following under fired engine conditions Piston rings Piston skirt Small end bearing Big end bearings Main bearings The following were accounted for Increased cylinder pressure Increased component temperature and effect on clearances Increased oil temperature and effect on oil viscosity 4

Full load fired friction loss at piston group (1) Fired conditions lead to FMEP (bar) slightly higher loss at small end bearings no significant change at top piston rings reduced loss at oil control rings significantly increased loss at piston skirt (particularly at lower engine speeds) 2.5 2.25 2. 1.75 1.5 1.25 1..75.5.25. motored full load fired piston skirts 9 1 11 12 13 14 15 16 Engine speed (rpm) FMEP (bar) FMEP (bar) FMEP (bar).2.18.16.14.12.1.8.6.4.2.2.18.16.14.12.1.8.6.4.2.2.18.16.14.12.1.8.6.4.2 small end bearings 9 1 11 12 13 14 15 16 Engine speed (rpm) top piston rings motored full load fired 9 1 11 12 13 14 15 16 Engine speed (rpm) oil control rings motored full load fired 9 1 11 12 13 14 15 16 Engine speed (rpm) motored full load fired 41

Full load fired friction loss at piston group (2) Predicted losses at piston skirt are highly sensitive to skirt/liner clearance which is sensitive to assumed component temperatures Results shown with +/-5 micron clearance No change in component or oil temperatures Model could be refined with better knowledge of component temperatures FMEP (bar) 2.5 2.25 2. 1.75 1.5 1.25 1..75.5.25. 9 1 11 12 13 14 15 16 Engine speed (rpm) motored full load fired (nominal) full load (5um reduced clearance) full load (5um increased clearance) 42

Full load fired friction loss at crank train bearings Fired conditions lead to slightly higher loss at lower speeds but similar losses at high speed as inertia forces dominate FMEP (bar) FMEP (bar).3.25.2.15.1.5 9 1 11 12 13 14 15 16.3.25.2.15.1.5 Main bearings Big end bearings motored full load fired Engine speed (rpm) 9 1 11 12 13 14 15 16 Engine speed (rpm) motored full load fired 43

Whole engine full load fired friction loss Results grouped by subsystem Pumping losses excluded Measured line is whole engine motored data for comparison At 15 rpm 5 4.5 4 3.5 valve trains and drive piston assemblies crankcase pumping oil and water pumps transmission (inc balancer) crankshaft measured minus pumping loss Crankshaft 22.2% Transmission 6.9% Pumps 1.% Crankcase pumping 7.5% FMEP (bar) 3 2.5 2 Pistons 41.9% 1.5 Valve trains 11.5% 1.5 9 1 11 12 13 14 15 16 Engine speed (rpm) 44

Conclusions Overall friction level can be estimated easily based on historical data If more complex calculations are performed there are many problems All major contacts involve asperity contact losses and oil shear losses Asperity contact losses dominate if dimensions from drawings are used but as components wear during run-in period these losses reduce This makes true prediction difficult but if worn component geometry and surface roughness data are used as model inputs then better results are possible Some interesting insights can be obtained by use of relatively inexpensive motored strip test combined with use of advanced CAE tools But to make further progress we need measured data from fired engines at high speed to validate software improved models of mixed lubrication including 3D surface texture effects 46

Thank you for your attention Any Questions? 47