DETAIL ENGINE FRICTION ESTIMATION USING EXPERIMENTALLY-SIMULATION APPROACH

Transcription

1 XLVII. INTERNATIONAL SCIENTIFIC CONFERENCE OF THE CZECH AND SLOVAK UNIVERSITIES AND INSTITUTIONS DEALING WITH RESEARCH OF INTERNAL COMBUSTION ENGINES SEPTEMBER 5-6, 2016 BRNO, CZECH REPUBLIC BRNO UNIVERSITY OF TECHNOLOGY, FACULTY OF MECHANICAL ENGINEERING INSTITUTE OF AUTOMOTIVE ENGINEERING DETAIL ENGINE FRICTION ESTIMATION USING EXPERIMENTALLY-SIMULATION APPROACH Miloslav Emrich 1, Michal Takáts 2 Abstract Higher mechanical efficiency of the combustion engine and lower energy consumption of the engine accessories helps to pass latest demands on CO2 emissions reduction. This article describes experimentally-simulation approach to estimate particular friction losses of the engine cranktrain. Single cylinder research engine was used as source of experimental data for model calibration. Commercial software GT-Suite by Gamma Technologies was used for detailed friction losses calculation. 1. INTRODUCTION New demands for lower CO2 emissions can be achieved by combination of highly efficient engines, hybridization and energy recuperation, lower mass and better aerodynamics of new vehicles. Increase of Engine efficiency is achieved by downsizing, down speeding, combustion process optimization, friction losses reduction etc. Increase in engine specific power by downsizing leads to higher load of components and influences friction losses. Prediction of friction losses by numerical simulation can be used in the early stage of new engine design to find the best compromise between the engine load, speed and mechanical efficiency. The goal of this work is to estimate in detail partial friction losses of a real single cylinder diesel engine. Experimental results were used for calibration of the mathematical model made that was created in commercial software GT-Suite. 2. BODY OF PAPER Single cylinder research diesel engine AVL (Bore 85mm, Stroke 90mm, DOHC, 1 st order balancers) was used as source of calibrating data. The test bed is 1 Miloslav Emrich, Czech Technical University in Prague, Faculty of Mechanical Engineering, Vehicle Center of Sustainable Mobility, Přílepská 1920, Roztoky, Czech Republic, Miloslav.Emrich@fs.cvut.cz 2 Michal Takáts, Czech Technical University in Prague, Faculty of Mechanical Engineering, Vehicle Center of Sustainable Mobility, Přílepská 1920, Roztoky, Czech Republic, Michal.Takats@fs.cvut.cz

2 equipped with an external source of pressurized oil and cooling water including conditioning system. These units kept temperatures in narrow limits set by user, which is very important especially in case of the oil temperature. External compressor can be used for overboosting of the engine. There are only two accessories powered by the engine common rail pump and 1st order balancers. The test bed is equipped with the low and high speed data acquisition system. The low speed DAQ is used for measuring of pressures and temperatures of oil, water, fuel, air, and exhaust gasses in time domain. The high speed DAQ is used for measuring in-cylinder pressure, intake and exhaust manifold pressure (TPA), actual torque and speed irregularity. These data are saved with the dependency on the crank angle Experimental results The engine was performed under full load, partial load and under motored condition including disconnection of accessories. All measurement were performed in three various RPM (1000, 1700, 2400), some also at 3100 RPM. Description of experiments covers Table 1. Table 1: Description of Experiments Common Rail Pump CommonRail replaced by Pulley Valvetrain and CommonRail Belt Balancers with Belt OverBoost Oil and Water Temperatures [degc] Friction mean effective pressure (FMEP) was evaluated (1) as the diference between the indicated mean effective pressure (IMEP) and the brake mean effective pressure (BMEP). High speed in-cylinder pressure measuring was used for IMEP (2) estimation as the mean value over 200 cycles. Piezoresistiv sensor in the intake port was used for zero level correction of the in-cylinder pressure. TDC was estimated under motored conditions with the loss angle 0,9 deg for each RPM. BMEP is calculated from engine torque (3). FMEP = IMEP BMEP (1) p dv IMEP (2) V cycle Z Full Load Curve 50, 70, 90 Part Load Curve (IMEP=4 bar) 50, 70, 90 Motored 50, 70, 90 Motored_without_CR 50, 70, 90 Motored_without_CR_overboosted 90 Motored_without_CR&VT 50, 70, 90 4 Mt BMEP V Z (3) (p actual in-cylinder pressure, VZ engine displacement 511 cm 3 ; Mt engine torque) New Castrol 5W-40 was used for all experiments. Oil viscosity has significant influence on friction losses of the engine. For this reason, the oil and water temperature was held on the same constant value during measurement by external conditioning units. RPM dependency of FMEP measured at 90degC (oil and water) is in Figure 1. Reasonable results were obtained under fired conditions (continuous lines) higher RPM and

3 higher load causes increase of friction. The same results were obtained at motored conditions (dashed lines). The engine was also motored with supercharging, pressure in the intake manifold was 1.7 bar and 2.2 bar absolutely. External compressor unit with the steady state vessel was used for this purpose. Figure 1 also shows influence of component disconnecting. Common rail pump was replaced by freewheel pulley with two low friction ball-bearings (line: Motored, NO CR). The lowest curve shows FMEP for motored engine without any accessories except 1st order balancers (connected all the time). In this experiment, the timing belt was removed and the valves were fixed in their closed positions. Figure 1: FMEP at constant temperature of oil and water (90 degc) Figures 2 and 3 show FMEP vs. RPM as a temperature dependency under motored and fired conditions. Figure 4 describes calculated FMEP for valvetrain and for Commonrail pump. These values are inaccurate because they were calculated as the difference of two high numbers motored engine and motored engine with removed CR pump. Moreover, inconsiderable energy for fuel compression during combustion had to be neglected. Other solution was not available at the moment, but these components will be measured separately at specialized test benches in the near future. Figure 2: Motored Engine at various temperatures, case of uncoupled Common Rail Pump and Valvetrain

4 Figure 3: FMEP for engine under two loads at various oil and water temperatures Figure 4: Valvetrain and CommonRail FMEP evaluated from experimental data Measurement uncertainty Measurement uncertainty (MU) is very often neglected and experimental results are published without calculation or discussion about this issue. Experimental results are expected to show engine FMEP dependency on the temperature, load and speed. FMEP is obtained as the difference between two large values (1) which principally leads to high uncertainty. Discussion can be split into IMEP and BMEP uncertainty. IMEP uncertainty is influenced by accuracy of piezoelectric pressure transducer (temperature drift, linearity), its position in cylinder, accuracy of piezoelectric amplifier, voltage transducer and TDC determination. There is very small potential for improvements, because very accurate components of this measurement chain were used. IMEP uncertainty was qualifiedly estimated as uimep=5 kpa. BMEP is calculated from the engine torque (3). Torque measurement uncertainty using torque flange is influenced by temperature effects on the zero signal, temperature effect on the sensitivity, hysteresis, linearity deviation, repeatability, reproducibility, determination of sensitivity, mechanical remanence (hysteresis of the zero signal), parasitic loads, rotational speed, and adaptation parts [4]. Table 2 contains torque flange operation parameters, distribution factor and sectional measurement uncertainty calculated according to [4]. Total MU is calculated as a root of sum of second power sectional measurement uncertainties.

5 The main result is total torque measurement uncertainty umt=1.15 Nm in operational area ( Nm). Note: Torque required for motoring the engine without accessories is approx. 4.5 Nm. Table 2: Torque measurement uncertainty for torque flange HBM T40 Very simplified FMEP uncertainty estimation can be done base on equation (5). This equation is used for calculation of indirectly measured quantity in case when only one measurement is performed. It consider only uncertainty B-type, uncertainty A-type is omitted. u y = ( y ) 2 u 2 X 1B + ( y ) 2 2 u 1 X 2B (5) 2 where uy is combined uncertainty; X1, X2 are measured quantities (in our case IMEP and torque Mt); u1b and u2b are standard measurement uncertainty of quantities X1 and X2. u FMEP = ( FMEP IMEP )2 u IMEP 2 + ( FMEP Mt )2 2 u Mt (6) Numerical result of equation (6) is total measurement uncertainty ufmep=19,4 kpa. Extended measurement uncertainty with confidence interval 95,45% (coverage factor=2) is ufmep=38,8 kpa. Measurement accuracy can by improved by torque flange with lower range, holding ambient temperature close to the reference value 23 degc and of course minimizing parasitic loads. Example: using Torque flange 200Nm at ambient temperature 23 degc leads to half extended measurement uncertainty ufmep=19,8 kpa GT-SUITE Friction model GT-Suite is a software product of Gamma Technologies, LLC. It consists of preprocessor, processor and postprocessor. Graphical interface is used to build models in pre-processor (Figure 5). Presented friction predictive mechanical model consists of rigid bodies, beams and joints. Differential equations are numerically solved using fifth order explicit Runge-Kutta method. The time step changes throughout the simulation

6 for the solution of a mechanical system. The purpose of the adaptive time step control is to achieve minimum computational time while maintaining the desired solution accuracy. The model convergence is very fast, solution time is usually in the range of 1.5 and 2 minutes per each case. Check [1] for detailed description of solution methods. Figure 5: Mathematical model in GT-Suite graphical pre-processor The model consists of main mechanical parts and friction couples. Fly-wheel is connected through a virtual shaft (consisting of rod and two rubber elements) to the dynamometer with constant angular speed. Torsional stiffness and Torsional Damping Coefficient was set base on real shaft parameters. This configuration allows to simulate speed irregularity of the engine (Figure 6). The model load is described by measured in-cylinder pressure. Control elements (Gain and Delay) were used to delay and reduce in-cylinder pressure acting to piston rings. Forty percent of in-cylinder pressure acts on the second ring with the offset of 90 degrees CA. The main geometry and mass parameters were measured or obtained from the testbed supplier. There are geometrical parameters generally not measurable during engine operation, such as actual piston shape, cylinder bore distortion, actual local components and oil temperatures etc. Bore distortion was chosen base on results published in [3]. The piston skirt shape was chosen as a perfect cylinder. Detailed description of solved equations would exceed allowed article size. Only basic assumptions and main equations will be described here Journal Bearings Bearing Oil Film Calculations are based on Mobility Approach. It means that known forces are used to calculate states. Vibrations and secondary motions are ignored and transmitted forces are obtained from rigid-body kinematic solutions with perfect joints. Reynolds equation is solved for varying eccentricity and load vector Piston Rings The core functionality for the lubrication-friction model includes solution of equations governing ring cross-section radial and toroidal twist motions, ring oil film hydrodynamics and ring-cylinder metal-to-metal (asperity) contact pressures.

7 Equations for radial force and toroidal moment balances on the ring cross-section are solved, in order to predict its radial and toroidal twist motions, and the instantaneous film thickness. The Reynolds equation, governing the hydrodynamics of the oil film between the ring face and the cylinder, is solved for the oil film pressure distribution, and (through its integration over the ring face) the force and moment on the ring cross-section due to the oil film pressure. Equations for the Greenwood-Tripp [2] asperity contact model are solved, with the same resolution as the Reynolds equation, for the distribution of asperity contact pressure distribution on the ring face, and (through its integration over the ring face) the force and moment on the ring cross-section due to the asperity contact pressure. The effect of surface roughness on oil film hydrodynamics is modeled through calculation and application of the Patir-Cheng pressure and shear flow (correction) factors in the Reynolds equation. A fast-running, simplified ring-bore (elastic) conformability analysis, is performed to estimate the fraction of bore circumference which "carries" the ring load, and is used as an "attenuation" factor for the cylinder reactions in ring cross-section force and moment balances. Axial motions (up-down) of the rings within their grooves have also been left outside the model, as their effect on ring friction is indirect and thought to be less significant Piston Skirt GT-SUITE allows modelling of lubrication and friction of the piston skirt against cylinder bore at a level of fidelity required for prediction of skirt friction. The core functionality for the lubrication-friction model includes solution of equations governing piston thrust and rotation motions, skirt oil film hydrodynamics and skirt-cylinder metal-to-metal (asperity) contact pressures. Equations for lateral force and rotation moment on the piston are solved, in order to predict its eccentricity and tilt motions, and the instantaneous film thickness. Solution of the Reynold equation including effect of surface roughness is mostly same as for piston rings. A skirt deformation is omitted in this case Model optimization Presented model was built based on real known geometry, masses and other known parameters. A lot of sensitivity analysis was done to check the influence on results. Unknown parameters were chosen from recommended values. Only one parameter optimization was run to achieve target FMEP. The ring oil film temperature was optimization variable. Optimized oil ring temperature was limited to expected range 91 degc to 169 degc. Final values were linearized and partial results are presented in Figure 7. The skirt oil film temperature was set as constant to 92 degc, in case of the oil sump temperature 90 C and 65 degc, in case of the oil sump temperature 50 degc. Coincidence of measured and simulated speed irregularity was checked to confirm validity of the model (Figure 6). Predicted FMEP compared with experimental data is presented in Figure 8 and Figure 9. Very good coincidence was achieved at higher speed. The biggest difference is at 1000 RPM, but these differences can be covered with measurement uncertainty.

8 Figure 6: Comparison of engine speed irregularity Figure 7: Ring Oil Film Temperature (Oil Sump temperature 90 degc) Figure 8: Comparison of experimental data and numerical simulation results (oil and water 50degC) 2.4. Results Figure 9: Comparison of experimental data and numerical simulation results (oil and water 90degC) Total Friction losses and share losses are presented in Figure 10. This chart is a combination of simulated and measured data (valvetrain and commonrail pump losses obtained from experiments only). Detail share friction at 2400 RPM and full load are presented in pie chart (Figure 11).

9 Figure 10: Friction Losses and share Losses combination experiment and simulation Figure 11: Detail share losses at 2400 RPM and Full Load These results correspond with commonly published values. Piston rings and piston skirt share is 55% of all losses. Bearings share is approximately 23% and the valvetrain and commonrail pump accounts for the remaining 22%. Other common accessories, such as the oil and water pump, alternator, AC compressor were not covered in this calculation. Various Crank Angle dependency data are also available in results. As an example total Friction Power and Force of piston rings is presented in Figure 12 and Figure 13. Integral value of total friction power presented in Figure 12 is 260W. Hydrodynamics lubrication is minimized (under 10%) at this regime due to high load and low speed. Figure 12: Friction Power loss of piston rings at 1000 RPM, full load, 90degC

10 Figure 13: Total Friction Force of piston rings at 1000 RPM, full load, 90degC 3. CONCLUSION Simple numerical model has been calibrated base on experimental data measured on single cylinder research engine under various load and temperature conditions. It has been found very good coincidence. This model has very fast computation time under two minutes per case. This kind of presented model seems to be suitable as part of GT-Suite thermodynamical models, where empirical equation like Chen-Flynn is often used. This empirical equation predicts friction losses base on maximal combustion pressure and mean piston speed only. These parameters are not able to capture other important influences on friction losses of the combustion engine unlike presented friction prediction model. Moreover thermodynamical model can be source of boundary conditions for friction model, especially temperatures of piston, piston rings and cylinder which influences oil temperature. This ideas of combination thermodynamical and mechanical model will be studied in close future. 4. ACKNOWLEDGEMENT This research has been realized using the support of Ministry of Education, Czech Republic, research program Národní program udržitelnosti I NPU I (LO1311), project MŠMT-7778/2014. This support is gratefully acknowledged. This research has been realized using the support of EU Regional Development Fund in OP R&D for Innovations (OP VaVpI) and Ministry for Education, Czech Republic, project # CZ.1.05/2.1.00/ Acquisition of Technology for Vehicle Center of Sustainable Mobility. This support is gratefully acknowledged. 5. REFERENCES [1] GT-SUITE Mechanics Theory Manual ver.7.4. Gamma Technologies, Inc [2] Greenwood, I. and Tripp, J.H. (1971). The Contact of Nominally Flat Surfaces", Proc. I. MechE, Vol 185, pp [3] Lombardi A. A Study of Cylinder Bore Distortion in V6 Aluminium Alloy Engine Blocks. Master of Applied Science thesis. Ryerson University, Toronto, Canada [4] Wegener, G. and Andrae, J., 2006, "Measurement uncertainty of torque measurements with rotating torque transducers in power test stands", Measurement, Vol. 40, Issues 7/8, pp