Analysis of knocking phenomena in a high performance engine Federico Millo, Luciano Rolando 1 st GTI Italian User Conference March 18 th, 2013 Turin
OUTLINE Introduction Experimental setup Results & discussion Conclusions 2
Introduction The main downsizing driver: CO 2 reduction targets Fleet standards 2015 130 g/km CO 2 95 g/km CO 2 Proposed fleet standards 2020 130 g/km by 2015 95 g/km by 2020
Introduction Technologies for improving fuel economy of gasoline engines: Downsizing + Turbocharging Gasoline Direct Injection (GDI) Variable Valve Actuation (VVA) Electrification and Hybridization..
BMEP POLITECNICO DI TORINO - ENERGY DEPARTMENT Introduction BMEP vs. Specific Brake Power European Gasoline Engines 2009 25 bar bmep 100 kw/ltr Specific Brake Power [kw/dm 3 ] Source: Indagine sui principali parametri prestazionali nei motori ad accensione comandata autoveicolistici di attuale produzione, P. Paniccia, BSc Thesis, Politecnico di Torino, 2010
SURGE OVERSPEED Introduction Scavenging CHOKING The use of GDI in turbocharged engines allows a reduction of the octane request, thus permitting to increase compression ratio, boost level, spark timing: significant performance improvements can thus be achieved, allowing an effective engine downsizing. However the low end torque of a turbocharged engine is usually limited by compressor surge. The use of GDI, combined with VVT, allows a cylinder scavenging effect, with significant improvements in the low-end torque performance. (Source: Andriesse et al. The New 1.8 ltr DI Turbo-Jet Gasoline Engine from FPT,17. Aachener Kolloquium Fahrzeug und Motorentechnik, 2008)
Introduction Knock Mega knock: Immediate effects causing damages to spark plug, rings, piston. 250 200 150 100 50 0 330 360 390 However, although the achievement of high boost levels at low engine speed definitely improves engine low end torque performance, the likelihood of engine knock increases dramatically. Furthermore, the risk of pre-ignition or mega knock, with pressure peaks reaching or exceeding 200 bars, is also significantly increased, due to the high power density. Therefore, reliable knock predictive models are necessary to support the design and calibration activities of new turbocharged high performance engines.
Introduction Increase in the complexity of calibration of GDI turbocharged engines Increase in the number of calibration parameters: Lambda Boost level Spark Advance Intake Valve Opening Exhaust Valve Opening Traditional one parameter at a time calibration approach unsuitable Interactions between calibration and design parameters choices (eg. between boost and compression ratio) Possibilities offered by the continuous development of CAE tools to carry out the system optimization on a virtual test bench
OUTLINE Introduction Experimental setup Results & discussion Conclusions 9
Experimental Setup Engine: FIAT T-Jet family Production Engine N of cylinder 4 In line Displacement 1368 cm 3 Bore 72 mm Stroke 84 mm Injection System PFI Turbocharger IHI RHF3 Pistons Forged Compression Ratio 9.8 Max Nominal Torque 230 Nm@3000 RPM Max Nominal Power 150 CV@ 5750 RPM Racing Engine N of cylinder 4 In line Displacement 1368 cm 3 Bore 72 mm Stroke 84 mm Injection System PFI Turbocharger FGT - Garrett GT 1446 Pistons Forged Compression Ratio 9.4 Max Nominal Torque 270 Nm@3000 RPM Max Nominal Power 180 CV@ 5750 RPM 10
Experimental Setup 4 Sensorized Spark Plugs: Kistler 6115 Accelerometer: Bosch 0261231148 Additional Sensors: 4 Thermocouples (k-series) in the intake runners, 1 Lambda sensor mounted downstream of the turbine Turbocharger speed
BMEP [bar] ID/IDm [-] Relative A/F [-] POLITECNICO DI TORINO - ENERGY DEPARTMENT Test Matrix Engine Speed: 2500-3000 - 3500-4000 - 5000-6000 [rpm] Relative A/F Ratio: 0.7 0.8 0.9 [-] Boost Pressure: 2000 2200 2400 [mbar] 0.9 24 22 20 18 16 14 12 10 Spark Advance 0 2 4 6 8 10 12 14 16 Spark Advance [ ] 3 2.5 2 1.5 1 0.5 0 Norm Knock intensity 2000 3000 4000 5000 6000 2000 2200 For each operating conditions 3 different spark advance settings were tested : Knock Limited Spark Advance (KLSA), KLSA +2, KLSA -2 For each op.cond. and spark timing 200 consecutive engine cycles were acquired 0.8 0.7 Engine Speed [RPM] 2400 Boost Pressure [mbar] 12
Test Matrix Additional tests were carried out evaluating the effects of: Different Gasoline: Composition [% m/m] Racing Gasoline Lower Heating Value [MJ/kg] 41.41 Octan Number [R.O.N.] 102 Density [kg/m3] 758.4 Carbon: 83.72 % Hydrogen: 13% Oxygen: 3.28% Composition [% m/m] Unleaded Gasoline Lower Heating Value [MJ/kg] 44.47 Octan Number [R.O.N.] 95.7 Density [kg/m3] 724.6 Carbon: 86.45 % Hydrogen: 13.55% Temperature of the intake manifold downstream of intercooler: T 1 = 44 C T 2 = 55 C 13
OUTLINE Introduction Experimental setup Results & discussion Conclusions 14
Three Pressure Analysis (TPA) Knocking phenomena were analyzed by means of the Three Pressures Analysis (TPA). The TPA represents a simulation based methodology to analyze experimental data and to determine quantities that are difficult or impossible to measure directly, such as: Apparent burn rate Residual fraction Trapping ratio Valve mass flow profiles Focuses on a cylinder, cuts-off rest of system Replacing it by measured port pressures Input exp. cylinder pressure to get comb. rate Valid only for steady state operating points Single cylinder model (typically) Provides as output combustion and knock metrics such as, for instance: Crank angle at knock onset Unburned mass fraction at knock onset, etc. 15
Knock metrics In cylinder pressure MAPO Maximum Amplitude Pressure Oscillation Band pass filtered pressure
MAPO (atm) MAPO (bar) MAPO (bar) POLITECNICO DI TORINO - ENERGY DEPARTMENT Chun, Heywood, SAE-890156 Chun, Heywood, SAE-890156 Borg, Borg, Alkidas, Alkidas, SAE-2008-01-1088 Unburned Mass Fraction at Knock Onset Unburned Mass Fraction at Knock Onset Unburned Mass % at Knock Onset Knock metrics: why use also engine block vibration signal? No correlation was initially found between knock intensity and unburned mass fraction at knock onset, in good agreement with literature data. Lack of correlation could be due to incorrect knock metrics?
Test results POLITECNICO DI TORINO - ENERGY DEPARTMENT Knock metrics: why use also engine block vibration signal? In cylinder pressure frequency spectra Knock free Knock free Incipient Light Knock Incipient Knock Knock Medium- Heavy Knock Knock Heavy Knock f m,n [khz] f 1.0 f 2.0 f 0.1 f 3.0 f 4.0 theoretical 8.1 13.5 16.9 18.6 23.5
Knock intensity: block vibration vs. in-cylinder pressure correlation P boost = 2.2 [bar] - λ = 0.8 [-] S.A: 4 Engine Speed: 2500 RPM Engine Speed: 3500 RPM Engine Speed: 5000 RPM After a proper tuning of the bandpass filtering frequencies, a good correlation between block vibration based and in cylinder pressure based knock intensity measurements was found. 19
MAPO (atm) MAPO (bar) MAPO (bar) POLITECNICO DI TORINO - ENERGY DEPARTMENT Knock intensity vs. unburned mass fraction at knock onset Chun, Heywood, SAE-890156 Borg, Alkidas, SAE-2008-01-1088 Unburned Mass % at Knock Onset Unburned Mass Fraction at Knock Onset A good correlation between knock intensity and unburned mass fraction at knock onset was also found, differently from data previously reported in Unburned Mass Fraction at Knock Onset literature. Engine Speed 2500 RPM Engine Speed 5000 RPM
Test Results Statistics Knock intensity thresholds Knock Free - S.A. = 0 Knock threshold Knock threshold Incipient Knock - S.A. = 2 Knock threshold Knock threshold Light Knock - S.A. = 4 Knock threshold Knock threshold 21
Knock prediction: phenomenological models Different models available in literature (Douaud&Eyzat, Franzke, Worret) have been included in the EngCylKnock object They are generally based on the assumption that end gas autoignition will occur when the condition dt/ = 1 will be reached (where t is the elapsed time from the start of end-gas compression and τ is the induction time) Douaud & Eyzat - MaMass fraction burned - MaMass fraction burned Knock Free cycle Knocking cycle - Mass fraction burned - Induction time integral - Mass fraction burned - Induction time integral
Knock prediction: phenomenological models Very good correlation between prediction of crank angle at knock onset based on Douaud&Eyzat model and experimental values was found.
Conclusions & Future Work After a proper tuning of the knock intensity metrics and of the knock prediction models, a good correlation could be found between experimental measurements and simulation results, both in terms of knock intensity ad in terms of knock onset prediction. Further investigations will be carried out in order to evaluate the effects of: internal EGR different fuel properties (e.g. E85) mixture inhomogeneity 24
AKNOWLEDGMENTS The valuable support provided to the research activity by Centro Ricerche Fiat, Fiat Powertrain Technologies Racing, Kistler Italy and Gamma Technologies is gratefully acknowledged. The authors would also like to thank in particular mr. Fabrizio Mirandola (formerly at FPT racing) for his precious and constant support during the experimental activities.
Analysis of knocking phenomena in a high performance engine Federico Millo, Luciano Rolando 1 st GTI Italian User Conference March 18 th, 2013 Turin