MECCA_21-01_web Range Extender ICE Multi -Parametric Multi -Objective Optimization MIKULÁ! ADÁMEK, RASTISLAV TOMAN MECCA 01 2021 PAGE 11 10.14311/mecdc.2021.01.02 Range Extender ICE Multi -Parametric Multi -Objective Optimization MIKULÁ! ADÁMEK, RASTISLAV TOMAN MIKULÁ! ADÁMEK CTU in Prague, Faculty of Mechanical Engineering; Technická 4, Praha 6, 166 07, Czech Republic; E -mail: mikulas.adamek@fs.cvut.cz RASTISLAV TOMAN CTU in Prague, Faculty of Mechanical Engineering; Technická 4, Praha 6, 166 07, Czech Republic; E -mail: rastislav.toman@fs.cvut.cz ABSTRACT Range Extended Electric Vehicles (REEV) are still one of the suitable concepts for modern sustainable low emission vehicles. REEV is equipped with a small and lightweight unit, comprised usually of an internal combustion engine with an electric generator, and has thus the technical potential to overcome the main limitations of a pure electric vehicle – range anxiety, overall driving range, heating, and air -conditioning demands – using smaller battery: saving money, and raw materials. Even though several REx ICE concepts were designed in past, most of the available studies lack more complex design and optimization approach, not exploiting the advantageous single point operation of these engines. Resulting engine designs are usually rather conservative, not optimized for the best effi ciency. This paper presents a multi -parametric and multi -objective optimization approach, that is applied on a REx ICE. Our optimization toolchain combines a parametric GT -Suite ICE simulation model, modeFRONTIER optimization software with various optimization strategies, and a parametric CAD model, that fi rst provides some simulation model inputs, and second also serves for the fi nal designs’ feasibility check. The chosen ICE concept is a 90 degrees V -twin engine, four -stroke, spark -ignition, naturally aspirated, port injected, OHV engine. The optimization goal is to fi nd the thermodynamic optima for three different design scenarios of our concept – three different engine displacements – addressing the compactness requirement of a REx ICE. The optimization results show great fuel effi ciency potential by applying our optimization methodology, following the general trends in increasing ICE effi ciency, and power for a naturally aspirated concept. KEY WORDS: RANGE EXTENDER, RANGE EXTENDED ELECTRIC VEHICLE, HYBRID ELECTRIC VEHICLE, BATTERY ELECTRIC VEHICLE, INTERNAL COMBUSTION ENGINE, SPARK -IGNITION, THERMODYNAMIC OPTIMIZATION, GENETIC ALGORITHM SHRNUTÍ Elektrické vozidlo s prodlou!en"m dojezdem (REEV) je pova!ováno za jednu z mo!ností, jak vyrobit cenov# dostupn" automobil s nízk"mi emisemi $kodlivin a skleníkov"ch plyn%. Hlavní v"hoda tohoto konceptu spo&ívá v malé (lehké) baterii, která by m#la svojí kapacitou pokr"t v#t$inu !ivotního cyklu, men$í baterie té! sni!uje cenu vozidla. Aby u!ivatel nebyl omezen krátk"m dojezdem, je pro v"jime&né p'ípady vozidlo vybaveno tzv. prodlu!ova&em dojezdu. V#t$inou se jedná o pístov" spalovací motor s generátorem, jeho! pomocí se nabíjí hlavní baterie. V"vojem takového systému se zab"vala 'ada v"robc%, v#t$ina návrh% se v$ak zakládala pouze na zku$enostech v"vojá'% a v"sledné motory nebyly optimalizovány pro jejich provozní podmínky. (lánek pojednává o návrhu spalovacího motoru pro prodlu!ova& dojezdu pomocí mnoho -kriteriální a mnoho -cílová optimalizace pln# parametrického termodynamického modelu, v kooperaci s CAD modelem. CAD model je pou!it jako zdroj vstup% pro termodynamick" model a sou&asn# ke kontrole realizovatelnosti v"sledk% optimalizace. Navrhovan" motor je &ty'dob", atmosférick", dvouválec do V, s rozvodem OHV a nep'ím"m vst'ikem paliva. Celkem jsou optimalizovány t'i varianty motoru, li$ící se zdvihov"m objemem. Cílem je pokud mo!no naplnit po!adavek na kompaktnost v"sledného motoru. V"sledky odpovídají trend%m pro zvy$ování ú&innosti a v"konu spalovacích motor% a vykazují velk" potenciál pro sní!ení spot'eby paliva spalovacího motoru. KLÍ"OVÁ SLOVA: PRODLU#OVA" DOJEZDU, ELEKTRICKÉ VOZIDLO S PRODLOU#EN$M DOJEZDEM. RANGE EXTENDER, ELEKTRICK$ AUTOMOBIL S PRODLU#EN$M DOJEZDEM, ELEKTROMOBIL, SPALOVACÍ MOTOR, ZÁ#EHOV$ MOTOR, TERMODYNAMICKÁ OPTIMALIZACE, GENETICK$ ALGORITMUS RANGE EXTENDER ICE MULTI -PARAMETRIC MULTI -OBJECTIVE OPTIMIZATION Range Extender ICE Multi -Parametric Multi -Objective Optimization MIKULÁ! ADÁMEK, RASTISLAV TOMAN MECCA 01 2021 PAGE 12 1. INTRODUCTION Range Extended Electric Vehicles (REEV) are one of the doctrines considered for modern low emission vehicles. Main advantage of this concept comes from using the battery as small as possible, this way helping to save the rare and expensive materials used in batteries, still without causing the “range anxiety” to the user. To fully exploit this advantage and to be economically viable, the Range Extender (REx) used in REEV, needs to focus on the price, overall mass, package dimensions, and NVH (Noise Vibration and Harshness). The most common type of range extender (and the only one considered in this paper) is the internal combustion engine (ICE). Several companies have developed a REx ICE to some extent, including OEMs like Lotus, TATA, MAHLE, and some others [1; 2; 3; 4; 5]. Design concepts from table 1 suggest, that most of REx ICEs are designed primarily to fulfi l the already mentioned requirements. Most of the companies try to keep the price as low as possible, which means sticking to traditional concepts of natural aspiration, indirect injection, and two valves per cylinder heads: the only major difference is AVL with its rotary REx concept [5]. All the evaluated Range Extender engines operate at stoichiometric conditions, allowing for the use of three -way catalyst to fulfi ll the emission regulation. Although most of the engine concepts were designed from scratch, TATA being the only difference [3], all the designs are mostly based on engineering teams’ experience, and internal OEM’s empirical tools, and correlations. Literature indicates that the effi ciency is subject to other parameters (especially to price). This way, the entire system’s effi ciency will always be poor, due to the double tank -to -wheel energy conversion. Nevertheless, the fuel consumption can be signifi cantly reduced (which is always a positive attribute) by the engine optimization for its specifi c operation, even with “the cheapest” possible setup. 1.1 DESCRIPTION OF THE GENERAL IDEA Our general idea is based on building a detailed fully parametric thermodynamic model of an ICE, equipped with simulation sub -models with appropriate predictive capabilities, and then running a multi -parametric, multi -objective optimization. The optimization result is then checked for its feasibility, using a parametric CAD model of the engine block. This not only provides a necessary engineering feedback, but also helps to capture some important trends, that can occur when changing the basic engine parameters automatically: in our study it is for instance the con -rod elongation when decreasing the bore/ stroke ratio RB/S leading to the crank -train enlargement. Range Extender engines are designed to provide a specifi c power output Pe (generally enough for the vehicle to achieve the highway speed), usually around 30 kW. The engine is expected to provide this power output at wide -open throttle conditions and for most of its lifetime. This single operating point feature makes the REx engines well suitable for a multi- -parametric thermodynamic optimization mentioned above, although the general method can be easily extended on more operating points, and therefore other ICE concepts. 1.2 GOALS OF THE PAPER Our department at Czech Technical University in Prague (CTU) has gathered vast amount of experience on ICE design optimization throughout several former projects [6; 7]. The main goal of this article is to describe an ICE design and optimization method, that combines CAE simulation tools with CAD structural design. The CAD model sets up the simulation input data, and subsequently checks the fi nal design for its feasibility. AVL MAHLE TATA Lotus KSPG AVL Engine confi guration I2 I2 I2 I3 V2 (90 deg) Rotary (single) [-] Valvetrain layout SOHC SOHC SOHC SOHC OHV - [-] Valves per cylinder 2 2 2 2 2 - [-] ICE displacement Vd 0.570 0.900 0.624 1.193 0.799 0.254 [L] Bore B 70.0 83.0 73.5 75.0 80.0 - [mm] Stroke S 74.0 83.0 73.5 90.0 79.5 - [mm] Bore/Stroke ratio RB/S 0.946 1.000 1.000 0.833 1.006 - [-] Compression ratio rc 11 10 10.3 10 N/A [-] ICE speed nICE 5000 4000 5500 3500 4500 5000 [RPM] Mean piston speed cs 12.333 11.065 N/A 10.500 11.925 [m/s] Brake power Pe 18 30 28 36.8 30 15 [kW] BSFC 250 250 N/A 241 N/A 260 [g/kWh] TABLE 1: Overview of Range Extender ICE concepts TABULKA 1: P'ehled uspo'ádání motor% pro Range Extender Range Extender ICE Multi -Parametric Multi -Objective Optimization MIKULÁ! ADÁMEK, RASTISLAV TOMAN MECCA 01 2021 PAGE 13 The article is divided into six chapters, that cover all the important aspects of our method, starting with the engine concept selection (in chapter 2); followed by the introduction of the thermodynamic model with its predictive sub -models, and preliminary design analysis in chapter 3. Chapter 4 then discusses the optimization setup, results, and design analyses. Chapter 5 presents some potential future developments of our methodology, and fi nally, chapter 6 contains important conclusions from our study. 2. ICE DESIGN CONCEPT Choosing the right base concept is extremely important, since it has a strong direct infl uence on NVH, package, mass, and price. Our concept selection consists of weighing the four criteria (the lower the score, the better) in table 2, for four ICE concepts. Price is the most important parameter (weight of 4), followed by the NVH, mass, and total package: input weighing data for the package, mass, and price were taken directly from a table prepared by Mahle in [8]. The inline two -cylinder engine without balance shaft is a baseline concept; the other assessed two -cylinder designs are the I2 with balance shaft, V -twin engine (V2) with 90 degrees between the cylinder axes, and a two -cylinder with opposed pistons – Boxer engine. The NVH data in table 2 are our GT -Suite simulation results of each assessed crank -train confi guration. We obtained the simulation inputs from parametric CAD model: each engine setup had equal bore, stroke, and conn -rod lengths. First, all variants were fi rst order statically balanced. The I2 engines were simulated for the three feasible ignition orders (0 – 360 degrees, 0 – 180 degrees, and 0 – 450 degrees), and the best one of these is the I2 baseline in the fi nal table 2 comparisons (thus achieving 100%). The NVH relative value then represents a fraction of imbalanced force between the tested concept and I2 baseline. To enable the averaging of unbalanced forces and unbalanced moments together, the moments are weighted by the estimated bearing spacing. Finally, table 2 suggests, that V2 concept seems to be the most promising for REx ICE, and it is therefore chosen for our subsequent studies. To sum -up our ICE design concept: later chapters of this article will deal with the optimizations of four stroke, V2, naturally aspirated, spark ignition engine with port fuel injection, and two valves per cylinder (OHV). 3. THERMODYNAMIC MODEL A fully parametric thermodynamic model of our fi nal design concept serves for the subsequent multi -parametric and multi- -objective optimizations. The simulation model of the REx ICE was built within the 0D/1D GT -Suite simulation platform, which allows for the simulation of a whole engine thermodynamic cycle. The engine is a virtual one, with a special attention placed on the use of suitable sub -models with predictive abilities. A sub -model without a proper predictive ability could mislead the optimization and guide it to unrealistic results. 3.1 MAIN ENGINE GEOMETRY The main parameters of the thermodynamic model are the cylinder bore B, engine operating speed nICE, and bore/stroke ratio RB/S. Valve design parameters are linked to the cylinder bore diameter, using empirical formulas from [9]: • intake valve diameter Dvin = 0.36B; maximum intake valve lift Lvin = 0.3Dvin; • exhaust valve Dvex = 0.3B; maximum exhaust valve lift Lvex = 0.3Dvex. The 1D intake and exhaust air paths are also fully parametric, sized accordingly to the cylinder bore B, using generic fl ow coeffi cients. Intake air path contains also an air fi lter, throttle, and intake manifold volume. Exhaust path then contains a simplifi ed model of a catalyst brick, to get a realistic exhaust back -pressure. 3.2 FRICTION SUB -MODEL Friction model has a key infl uence on the resulting ICE’s setup, mainly on the RB/S ratio, and ICE operating speed. A simple Chen -Flynn model which was used in previous research studies showed major fl aws when used in a multi -parametric optimization due to its lack of predictive ability [7]. Therefore, we applied CTU in -house friction model Vyva!, created by Macek et al. [10]. In fact, its implementation into GT -Suite thermodynamic model as a sub -assembly. This friction sub -model has three main parts: pressure part, mechanic part, and friction part. Pressure part consists of series of pipe objects, that simulate engine blow -by, and predict the pressure differences between the piston rings – essential for the piston ring and skirt friction forces calculation. Second – mechanic part, consists of detailed mechanical model of crank -train, and determines all velocities, accelerations, and forces acting on each Package Mass Price NVH Weighted average Parameter weight 1 2 4 3 --- I2 without balance shaft 100 100 100 100 100 I2 with balance shaft 100 115 110 31.1 86.3 V2 with 90° angle 128 105 102 32.7 84.4 Boxer 104 108 103 44.7 86.6 TABLE 2: Decision table for the ICE concept selection TABULKA 2: Rozhodovací tabulka pro v"b#r konceptu motoru Range Extender ICE Multi -Parametric Multi -Objective Optimization MIKULÁ! ADÁMEK, RASTISLAV TOMAN MECCA 01 2021 PAGE 14 crank -train member. The last – friction part, calculates friction powers, FMEPs, and friction forces for each member of crank- -train, and piston assembly (crank bearings, main bearings, piston pin, piston skirt, and piston rings), using results obtained from the respective mechanic and pressure parts. Friction coeffi cients necessary for fi nding the friction forces are calculated using a simple model based on a mathematical description of Stribeck curve, that expresses a friction coeffi cient’s dependence on load, speed, and oil viscosity. Vyva! model also contains empiric relations for oil, and fuel pump losses. Finally, the friction loss in valve train is calculated by Bishop’s formula [9]. 3.3 COMBUSTION SUB -MODEL Another key area in spark -ignition ICE optimization is the simulation of SI combustion. We decided to use predictive phenomenological combustion model called EngCylCombSITurb (SITurb) available in GT -Suite. SITurb model calculates a differential equation for entrained mass rate of unburned gas with its main equation 1, where "u are unburned mixture density, and Af a fl ame area. SL and ST then represent the laminar and turbulent fl ame speeds. piston rings), using results obtained from the respective mechanic and pressure parts. Friction coefficients necessary for finding the friction forces are calculated using a simple model based on a mathematical description of Stribeck curve, that expresses a friction coefficient’s dependence on load, speed, and oil viscosity. Vyvaž model also contains empiric relations for oil, and fuel pump losses. Finally, the friction loss in valve train is calculated by Bishop’s formula [9]. 3.3. Combustion sub-model Another key area in spark-ignition ICE optimization is the simulation of SI combustion. We decided to use predictive phenomenological combustion model called EngCylCombSITurb (SITurb) available in GT-Suite. SITurb model calculates a differential equation for entrained mass rate of unburned gas with its main equation 1, where 𝜌𝜌2 are unburned mixture density, and 𝐴𝐴3 a flame area. 𝑆𝑆4 and 𝑆𝑆5 then represent the laminar and turbulent flame speeds. 𝑑𝑑𝑀𝑀- 𝑑𝑑𝑑𝑑 = 𝜌𝜌2𝐴𝐴3(𝑆𝑆4 + 𝑆𝑆5) (1) These two flame speed parameters – 𝑆𝑆4 and 𝑆𝑆5 – limit the flame kernel development: during the initial phases, when the kernel size is still small, the entrainment rate is limited by the laminar flame speed 𝑆𝑆4 (equation 2); equation 3 then accounts for the flame transition into a turbulent one, with 𝑢𝑢´ representing the mean fluctuating turbulent velocity, 𝑅𝑅3 the flame radius, and 𝐿𝐿6 the turbulent length scale. 𝑆𝑆4 = @𝐵𝐵7 + 𝐵𝐵8(𝜙𝜙 − 𝜙𝜙7)9C ∙ E 𝑇𝑇2 𝑇𝑇:G ; ∙ E 𝑝𝑝 𝑝𝑝:G < ∙ (1 − 2,06𝐷𝐷𝐷𝐷𝐷𝐷:.>>?0@) (2) 𝑆𝑆5 = 𝐶𝐶A𝑢𝑢´O1 − 1 1 + 𝐶𝐶BP𝑅𝑅3 9 𝐿𝐿69⁄ RS (3) SITurb needs a priori information about the turbulent flow in combustion chamber (𝐿𝐿6 and 𝑢𝑢´ parameters from equation 3). The source of these quantities is another GT-Suite’s sub-model – EngCylFlow (Flow) – a K-k-ε kinetic energy cascade flow model, that predicts the in-cylinder charge motion and turbulence. More details on both SITurb and Flow models, their evolution, and calibration can be found in [11; 12; 13]. SITurb uses five different calibration parameters, four from equations 2 and 3 (𝐷𝐷𝐷𝐷𝑀𝑀, 𝐶𝐶B, 𝐶𝐶A, and 𝐿𝐿6), and Taylor length scale multiplier 𝐶𝐶C (present in the burnup rate equation 𝑑𝑑𝑀𝑀D 𝑑𝑑𝑑𝑑⁄ , as a multiplier for the Taylor microscale of turbulence). Flow then has its own set of four calibration parameters. So, in total we have nine calibration parameters. First, we performed a thorough sensitivity analysis on these nine total calibration parameters, comparing the combustion model (combination of SITurb and Flow) responses on ICE geometry, ICE operating conditions (load, speed, and cooled EGR content). The sensitivity analysis showed that the burn durations vary in accordance with our general experience: however general burn rates are lower, and the overall sensitivity on operating conditions is higher (load, speed, cooled EGR). After the sensitivity analysis we performed a calibration of the nine SITurb and Flow calibration parameters using an available set of measurement data with ICE load/speed dependencies. The set of these nine calibrated parameters was then used in all our subsequent REx ICE optimizations, discussed in next chapters. 3.4. In-cylinder heat transfer sub-model (1) These two fl ame speed parameters – SL and ST – limit the fl ame kernel development: during the initial phases, when the kernel size is still small, the entrainment rate is limited by the laminar fl ame speed SL (equation 2); equation 3 then accounts for the fl ame transition into a turbulent one, with u´ representing the mean fl uctuating turbulent velocity, Rf the fl ame radius, and Lt the turbulent length scale. (2) piston rings), using results obtained from the respective mechanic and pressure parts. Friction coefficients necessary for finding the friction forces are calculated using a simple model based on a mathematical description of Stribeck curve, that expresses a friction coefficient’s dependence on load, speed, and oil viscosity. Vyvaž model also contains empiric relations for oil, and fuel pump losses. Finally, the friction loss in valve train is calculated by Bishop’s formula [9]. 3.3. Combustion sub-model Another key area in spark-ignition ICE optimization is the simulation of SI combustion. We decided to use predictive phenomenological combustion model called EngCylCombSITurb (SITurb) available in GT-Suite. SITurb model calculates a differential equation for entrained mass rate of unburned gas with its main equation 1, where 𝜌𝜌2 are unburned mixture density, and 𝐴𝐴3 a flame area. 𝑆𝑆4 and 𝑆𝑆5 then represent the laminar and turbulent flame speeds. 𝑑𝑑𝑀𝑀- 𝑑𝑑𝑑𝑑 = 𝜌𝜌2𝐴𝐴3(𝑆𝑆4 + 𝑆𝑆5) (1) These two flame speed parameters – 𝑆𝑆4 and 𝑆𝑆5 – limit the flame kernel development: during the initial phases, when the kernel size is still small, the entrainment rate is limited by the laminar flame speed 𝑆𝑆4 (equation 2); equation 3 then accounts for the flame transition into a turbulent one, with 𝑢𝑢´ representing the mean fluctuating turbulent velocity, 𝑅𝑅3 the flame radius, and 𝐿𝐿6 the turbulent length scale. 𝑆𝑆4 = @𝐵𝐵7 + 𝐵𝐵8(𝜙𝜙 − 𝜙𝜙7)9C ∙ E 𝑇𝑇2 𝑇𝑇:G ; ∙ E 𝑝𝑝 𝑝𝑝:G < ∙ (1 − 2,06𝐷𝐷𝐷𝐷𝐷𝐷:.>>?0@) (2) 𝑆𝑆5 = 𝐶𝐶A𝑢𝑢´O1 − 1 1 + 𝐶𝐶BP𝑅𝑅3 9 𝐿𝐿69⁄ RS (3) SITurb needs a priori information about the turbulent flow in combustion chamber (𝐿𝐿6 and 𝑢𝑢´ parameters from equation 3). The source of these quantities is another GT-Suite’s sub-model – EngCylFlow (Flow) – a K-k-ε kinetic energy cascade flow model, that predicts the in-cylinder charge motion and turbulence. More details on both SITurb and Flow models, their evolution, and calibration can be found in [11; 12; 13]. SITurb uses five different calibration parameters, four from equations 2 and 3 (𝐷𝐷𝐷𝐷𝑀𝑀, 𝐶𝐶B, 𝐶𝐶A, and 𝐿𝐿6), and Taylor length scale multiplier 𝐶𝐶C (present in the burnup rate equation 𝑑𝑑𝑀𝑀D 𝑑𝑑𝑑𝑑⁄ , as a multiplier for the Taylor microscale of turbulence). Flow then has its own set of four calibration parameters. So, in total we have nine calibration parameters. First, we performed a thorough sensitivity analysis on these nine total calibration parameters, comparing the combustion model (combination of SITurb and Flow) responses on ICE geometry, ICE operating conditions (load, speed, and cooled EGR content). The sensitivity analysis showed that the burn durations vary in accordance with our general experience: however general burn rates are lower, and the overall sensitivity on operating conditions is higher (load, speed, cooled EGR). After the sensitivity analysis we performed a calibration of the nine SITurb and Flow calibration parameters using an available set of measurement data with ICE load/speed dependencies. The set of these nine calibrated parameters was then used in all our subsequent REx ICE optimizations, discussed in next chapters. 3.4. In-cylinder heat transfer sub-model (3) SITurb needs a priori information about the turbulent fl ow in combustion chamber (Lt and u´ parameters from equation 3). The source of these quantities is another GT -Suite’s sub -model – EngCylFlow (Flow) – a K -k -# kinetic energy cascade fl ow model, that predicts the in -cylinder charge motion and turbulence. More details on both SITurb and Flow models, their evolution, and calibration can be found in [11; 12; 13]. SITurb uses fi ve different calibration parameters, four from equations 2 and 3 (DEM, Ck, CS, and Lt), and Taylor length scale multiplier C$ (present in the burnup rate equation dMb/dt, as a multiplier for the Taylor microscale of turbulence). Flow then has its own set of four calibration parameters. So, in total we have nine calibration parameters. First, we performed a thorough sensitivity analysis on these nine total calibration parameters, comparing the combustion model (combination of SITurb and Flow) responses on ICE geometry, ICE operating conditions (load, speed, and cooled EGR content). The sensitivity analysis showed that the burn durations vary in accordance with our general experience: however general burn rates are lower, and the overall sensitivity on operating conditions is higher (load, speed, cooled EGR). After the sensitivity analysis we performed a calibration of the nine SITurb and Flow calibration parameters using an available set of measurement data with ICE load/speed dependencies. The set of these nine calibrated parameters was then used in all our subsequent REx ICE optimizations, discussed in next chapters. 3.4 IN -CYLINDER HEAT TRANSFER SUB -MODEL Prediction of in -cylinder energy loss due to heat transfer is calculated with a combination of two models. First, structure and surface temperatures are obtained with a predictive fi nite element (FE) GT -Suite sub -model EngCylTWallSoln. FE model requires a simplifi ed geometry of all the surfaces, together with coolant and oil boundary conditions. The cylinder structure geometry is also parametric and linked to the engine main geometry. Second, the heat transfer coeffi cient is determined using classical Woschni correlation without swirl [14]. This approach was successfully used in some previous CTU research projects on ICE multi -parametric optimization [6; 7]. 3.5 KNOCK SUB -MODEL Knock prediction is modeled using a basic GT -Suite model EngCylKnock (Knock), which is based on a standard calculation of knock induction time integral. Knock model obtains the induction time using a Kinetics -Fit -Gasoline correlation. This correlation predicts the induction time with combination of reduced iso -octane oxidation and reduced n -heptane oxidation mechanisms [15]. We did not calibrate the Knock model, because of the time constraints. However, our previous sensitivity studies showed, that the uncalibrated model tends to stay on the safe side in its response to ICE operating condition changes, and this we consider advantageous [16]. 3.6 STRUCTURAL DESIGN ANALYSIS There are two main reasons for the use of a parametric CAD model of our V2 engine block: • CAD model helps us prepare some of the simulation input data for the friction model Vyva!, that requires detailed knowledge of crank -train masses, and con -rod compensating moment of inertia with its coordinates. Range Extender ICE Multi -Parametric Multi -Objective Optimization MIKULÁ! ADÁMEK, RASTISLAV TOMAN MECCA 01 2021 PAGE 15 • Subsequently, it serves for the fi nal structural design’s feasibility check – after running the multi -parametric optimizations. There are some relations, that can be established only with this parametric CAD model, and which are important inputs for Vyva! friction model – as we have already discussed. First of those relations is the relation between the piston assembly’s mass and cylinder bore; the second one is the relation between the con -rod’s length l and its geometry parameters; third one is the relation of con -rod ratio $ with bore/stroke ratio RB/S, and bore diameter B in equation 4 (n and c are functions of cylinder bore; $ = S/(2*l). One can assume, that a designer would try to minimize the package volume using a con -rod as short as possible, because this is a parameter that plays a major role in determining the crank -train’s size, and therefore the total engine’s package volume. rediction of in-cylinder energy loss due to heat transfer is calculated with a combination of two models. First, structure and surface temperatures are obtained with a predictive finite element (FE) GT- Suite sub-model EngCylT allSoln. FE model requires a simplified geometry of all the surfaces, together with coolant and oil boundary conditions. The cylinder structure geometry is also parametric and linked to the engine main geometry. Second, the heat transfer coefficient is determined using classical Woschni correlation without swirl [1 ]. This approach was successfully used in some previous CT research pro ects on ICE multi-parametric optimization [ ; ]. 3. . noc sub-model nock prediction is modeled using a basic GT-Suite model EngCylKno k (Kno k), which is based on a standard calculation of knock induction time integral. Kno k model obtains the induction time using a K n -F - a ol n correlation. This correlation predicts the induction time with combination of reduced iso-octane oxidation and reduced n-heptane oxidation mechanisms [1 ]. We did not calibrate the Kno k model, because of the time constraints. owever, our previous sensitivity studies showed, that the uncalibrated model tends to stay on the safe side in its response to ICE operating condition changes, and this we consider advantageous [1 ]. 3. . tructural desi n analysis There are two main reasons for the use of a parametric CA model of our 2 engine block: • CA model helps us prepare some of the simulation input data for the friction model Vyvaž, that requires detailed knowledge of crank-train masses, and con-rod compensating moment of inertia with its coordinates. • Subsequently, it serves for the final structural design’s feasibility check – after running the multi-parametric optimizations. There are some relations, that can be established only with this parametric CA model, and which are important inputs for Vyvaž friction model – as we have already discussed. First of those relations is the relation between the piston assembly’s mass and cylinder bore; the second one is the relation between the con-rod’s length 𝐷𝐷 and its geometry parameters; third one is the relation of con-rod ratio with bore/stroke ratio 𝑅𝑅 , and bore diameter 𝐵𝐵 in equation ( and are functions of cylinder bore; = 𝑆𝑆 2𝐷𝐷⁄ ). ne can assume, that a designer would try to minimize the package volume using a con-rod as short as possible, because this is a parameter that plays a ma or role in determining the crank-train’s size, and therefore the total engine’s package volume. = − P 𝑅𝑅 − 0 R 0 0 29 9 + 1 ∙ 0 012 + + P𝑅𝑅 − 0 R ( ) Figure 1 shows this relation in comparison with data acquired directly from the CA model (marked by RE suffix in legend), for four different bore diameters 𝐵𝐵. ur correlation gives feasible results, and these are fed into Vyvaž sub-model ensuring higher optimization accuracy and results’ feasibility. Okomentoval(a): [A4]: “Con-rod” or “connecting rod” Okomentoval(a): [A5R4]: Done – „con-rod“ used everywhere Okomentoval(a): [A6]: („REL“ suffix in the legend) Okomentoval(a): [A7R6]: Done (4) Figure 1 shows this relation in comparison with data acquired directly from the CAD model (marked by “REL” suffi x in legend), for four different bore diameters B. Our correlation gives feasible results, and these are fed into Vyva! sub -model ensuring higher optimization accuracy and results’ feasibility. Figure 2 with con -rod ratio $ dependence on RB/S ratio then shows, that for RB/S ratios under 0.5, the con -rod length grows rapidly (lower $ ratio means longer con -rod length). This Figure also clarifi es how we chose the RB/S ratio lower/upper limits in table 3, that will be introduced in next chapter. 4. OPTIMIZATION SCENARIOS A general expectation in an optimization aimed for the best effi ciency natural aspirated ICE is, that it will lead to relatively large displacement. This is however in a strong disagreement with the basic REx requirement for a compact design package, and low overall mass. To test this expectation, we decided to optimize our REx ICE model in three different scenarios, with the same Pe target of 30 kW, at the best possible BSFC. The only difference between the scenarios is the engine displacement. In the fi rst scenario, called Vmax , the optimizer is not limited by the cylinder displacement, only by the minimum/maximum cylinder bore diameter B and bore/stroke ratio RB/S. Main point of this scenario is to test, whether the optimization will properly follow the expected trends in increasing the ICE effi ciency. Second scenario V05 concept uses a fi xed cylinder displacement of 0.5 L, hence the name. This engine’s total displacement is slightly larger than most REx ICEs analyzed in chapter 1, but it is probably the most common engine displacement in automotive industry. Last scenario is V035, should represent the design’s compactness requirement. Since total displacement of 0.7 L is rather small for 30 kW power output, we expect the optimized design to turn out as “sporty”. 4.1 OPTIMIZATION SETUP There are nine total parameters chosen for the optimizations, summed -up in table 3 (SA is the spark advance angle; the fi ring TDC represents the 0 deg CA in all angular parameters). The Vmax scenario optimizes all of them; in the case V05 and V035 scenarios bore and stroke values are linked together because of the “locked volume”, and the optimizer varies only the RB/S ratio. FIGURE 1: Con-rod length dependence on engine stroke (values obtained with relation 4 are marked with “REL”) OBRÁZEK 1: Délka ojnice v závislosti na zdvihu motoru („REL“ zna&í hodnoty získané pomocí vztahu 4) FIGURE 2: Con-rod ratio dependence on bore/stroke ratio RB/S OBRÁZEK 2: Závislost ojni&ního pom#ru na pom#ru zdvihu a vrtání RB/S Range Extender ICE Multi -Parametric Multi -Objective Optimization MIKULÁ! ADÁMEK, RASTISLAV TOMAN MECCA 01 2021 PAGE 16 Parameters FIVC and FEVO multiply the width of the base valve lift curves, and by that control the inlet valve opening event or exhaust valve closing events respectively (Note: unity values of FIVC and FEVO correspond to 210 deg CA from open to closed valve, at 1 mm lift). All the optimizations are run using modeFRONTIER’s pilOPT hybrid algorithm, that combines local and global search, and is recommended for multi -objective problems [17]. pilOPT algorithm is then searching for the thermodynamic optima, trying to fulfi ll also the Pe demand. Each scenario optimization took about 10 000 design iterations. As we have already discussed in our introductory part, we use a fully parametric CAD model of V2 engine block for the design feasibility check, since a thermodynamically optimal design is not necessarily also feasible from the structural design viewpoint. 4.2 OPTIMIZATION RESULTS Since the optimization is multi -objective, optimal results are received in a Pareto front, which is a set of optimal values, where achieving improvement in one of the objectives will worsen the other. In our case, designs with lower BSFC tend to be further away from the desired power output, and vice versa. The optimal designs were then selected from the Pareto front using standard criterial function F (equation 5), where the weight coeffi cient for BSFC (%i1) is 0.9, and for brake power Pe (%i2) it is 0.1. The fraction Xi/Xi, max in equation 5 represents a normalization of Pareto front values, so that the two different objective functions Xi can be combined into one equation; Xi, max is the maximum value of the ith objective function in Pareto set. arameter mm - - de C de C de C - - o er limit . 1 29 32 . . er limit 1 1. 1 -1 3 1.3 1.3 esolution .1 . .1 1 1 1 . . able 3 imits and resolutions of the optimization parameters abul a 3 Rozsahy a rozli en optimalizovan ch parametr arameters and 0 multiply the width of the base valve lift curves, and by that control the inlet valve opening event or exhaust valve closing events respectively ( ote: unity values of and 0 correspond to 21 deg CA from open to closed valve, at 1 mm lift). All the optimizations are run using modeFR TIER’s l T hybrid algorithm, that combines local and global search, and is recommended for multi-ob ective problems [1 ]. l T algorithm is then searching for the thermodynamic optima, trying to fulfill also the - demand. Each scenario optimization took about 1 design iterations. As we have already discussed in our introductory part, we use a fully parametric CA model of 2 engine block for the design feasibility check, since a thermodynamically optimal design is not necessarily also feasible from the structural design viewpoint. 4. . timi ation results Since the optimization is multi-ob ective, optimal results are received in a areto front, which is a set of optimal values, where achieving improvement in one of the ob ectives will worsen the other. In our case, designs with lower BSFC tend to be further away from the desired power output, and vice versa. The optimal designs were then selected from the areto front using standard criterial function (equation ), where the weight coefficient for BSFC ( ) is .9, and for brake power - ( 9) it is .1. The fraction 7 in equation represents a normalization of areto front values, so that the two different ob ective functions can be combined into one equation; 7 is the maximum value of the ith ob ective function in areto set. = 7 B ( ) 4. . . erall results Results of all three optimization scenarios are summarized in table and figure 3. While table contains the set of optimal (independent – labelled with ’) parameters, main thermodynamic outputs, and some dependent parameters; figure 3 then shows the comparison of valve lift curves for each scenario. max 05 035 99. 9 .1 . [mm] . 3 1. .9 [-] 21 32 [R M] 1 .2 12. 1 .1 [-] 22. 3 . 2 . [deg CA] 13 .3 . 9 . [mm] 2.11 1. . [ ] Okomentoval(a): [A R ]: Corrected Okomentoval(a): [A ]: he su scri t in the su tion e should e „i“ (5) 4.2.1 OVERALL RESULTS Results of all three optimization scenarios are summarized in table 4 and Figure 3. While table 4 contains the set of optimal (independent – labelled with ‘*’) parameters, main thermodynamic outputs, and some dependent parameters; Figure 3 then shows the comparison of valve lift curves for each scenario. The Vmax scenario engine turned out as the most effi cient, with the best BSFC value of 214 g/kWh. But it is also the largest of the three engines. The fi nal Vmax volume is 2.1 L, with a long stroke RB/S of 0.735, leading to a mean piston speed of 9.70 m/s. Large volume means that the engine was able to achieve demanded Pe with only 2150 RPM – following the ICE down -speeding trend. A look at the Figure 3 shows, that the Vmax engine uses inlet valve open for extremely short time, running in Miller cycle. Tendency to push towards Miller cycle is well known. However, in comparison to similar previous optimizations [7], the optimizer was given bigger freedom in valve timing parameters, resulting in “stronger” Miller cycle. The combination of large volume, low speed, and Miller cycle leads to rather small BMEP of 7.95 bar, and volumetric effi ciency of only 58.7 %. Although such REx engine would hardly fulfi l package requirement, lower BMEP engines tend to be robust and reliable, which might be useful for an ICE operating at WOT right after start -up. Vmax concept also has relatively large compression ratio of 14.2. This can seem extreme, but in combination with spark advance of 22 deg CA BTDC, and Miller cycle especially, it is feasible. The V05 scenario achieves worse BSFC results compared to the Vmax variant, with a short stroke RB/S ratio of 1.485, at higher operating speed 3200 RPM, but lower cs of 7.05 m/s. We have Parameter B [mm] RB/S [-] rc [-] nICE [RPM] SA [deg CA] IVO [deg CA] EVC [deg CA] FIVC [-] FEVO [-] Lower limit 60 0.5 6 1000 80 290 320 0.7 0.7 Upper limit 100 1.5 15 7000 -10 380 450 1.3 1.3 Resolution 0.1 0.005 0.1 50 1 1 1 0.004 0.004 TABLE 3: Limits and resolutions of the optimization parameters TABULKA 3: Rozsahy a rozli$ení optimalizovan"ch parametr% Vmax V05 V035 B* 99.50 98.15 75.74 [mm] RB/S* 0.735 1.485 0.975 [-] nICE* 2150 3200 4650 [RPM] rc* 14.2 12.7 14.1 [-] SA* 22.0 36.0 27.0 [deg CA] S* 135.37 66.09 77.68 [mm] Vd 2.11 1.00 0.70 [L] CS 9.70 7.05 12.04 [m/s] BSFC 214.0 229.8 238.5 [g/kWh] BMEP 7.95 11.25 11.07 [bar] &' 58.67 89.27 91.08 [%] &m 92.78 90.70 84.48 [%] TABLE 4: Results for all three optimization scenarios TABULKA 4: V"slední hodnoty pro v$echny optimaliza&ní scéná'e Range Extender ICE Multi -Parametric Multi -Objective Optimization MIKULÁ! ADÁMEK, RASTISLAV TOMAN MECCA 01 2021 PAGE 17 seen some similar behavior before e.g. in [7], but there it was caused probably by the simple Chen -Flynn friction model [9], that lacks the predictive abilities, and therefore preferred this short -stroke confi guration. In this case, however, we think it is because of achieving good volumetric effi ciency, due to the direct link between the bore size and inlet valves diameter. The fact that V05 engine achieves high BMEP of 11.25 bar despite using some level of Millerization supports our assumption. Compression ratio is 12.7 which is lower than in the previous Vmax scenario, due to the short stroke confi guration, that leads to the increased knock tendencies [1]. However, with such a short stroke, even this compression ratio might be technically challenging to achieve. The last V035 scenario was chosen with rather small volume for such power output; the engine BSFC, ICE operating speed, and mean piston speed grow accordingly. The optimal RB/S ratio is also a bit surprising, since the engine is slightly long stroke. The reason for this is probably the tendency of Vyva! friction model (according to its author) to overrate the piston skirt friction growing with engine bore. Another parallel explanation is that Bishop’s formula uses the ICE operating speed and valve diameter to determine the valvetrain losses: since the valve diameter is linked directly to the engine bore, smaller bore actually leads to relatively high mechanical effi ciency of 84.5 %, despite high RPMs and mean piston speed (4650 RPM, 12.04 m/s). Yet, the small inlet valve area, pushes the optimizer to change the valve timing approach, compared to the fi rst two scenarios: V035 engine uses a later IVC angle compared to other scenarios. This way the engine achieves suffi cient cylinder fi lling, although with slight backfl ow just before IVC, reducing effective compression ratio thus improving knock robustness. 4.4.2 PACKAGE COMPARISON From the package standpoint, all three design scenarios were also compared using parametric CAD model of V2 engine block, since comparing just their volume is not particularly accurate, because the engine size is strongly infl uenced by RB/S ratio, and con -rod length (Figure 4, and table 5). Table 5 also contains estimated masses of the main engine components. The Vmax engine is by far the largest of the three – as expected. It is rather interesting, that the two other scenarios V05 and V035 are very similar from the package standpoint, though V035 is slightly shorter. According to our CAD model, all designs are FIGURE 3: Valve lift curves of optimized engines OBRÁZEK 3: Ventilové zdvihové k'ivky optimalizovan"ch motor% FIGURE 4: Engine block CAD models (from left Vmax, V05, V035) OBRÁZEK 4: CAD modely blok% (zleva Vmax, V05, V035) Width [mm] Height [mm] Length [mm] Block Mass [kg] Piston assembly Mass [kg] Con-rod Mass [kg] Con-rod Length [mm] Vmax 574.8 356.1 192.3 13,15 0,73 0,70 219,9 V05 401.2 231.8 186.6 6,58 0,71 0,54 126,7 V035 389.4 232.5 160.9 5,88 0,36 0,57 144,5 TABLE 5: Engine blocks’ dimensions and estimated components’ masses TABULKA 5: Rozm#ry blok% motor% a odhadnuté hmotnosti hlavních komponent Range Extender ICE Multi -Parametric Multi -Objective Optimization MIKULÁ! ADÁMEK, RASTISLAV TOMAN MECCA 01 2021 PAGE 18 feasible; however, V05 variant is on the limit, and some of the design features would be a very tight fi t. 5. POTENTIAL FOR FUTURE DEVELOPMENT OF THE OPTIMIZATION METHODOLOGY The designs resulting from our optimizations are generally feasible and achieve high yet not unrealistic effi ciencies for OHV designs. Despite these positive results, there is still some further potential in the development of our optimization methodology, especially from the viewpoint of the structural design, specifi cally in two areas. The fi rst area is referenced to the fact, that we do not scale the main and conn -rod bearing diameters with bore size. And these are important inputs for Vyva! friction model. A change in the con -rod bearing diameter would also infl uence the relation developed for con -rod ratio calculation (equation 4). We have already prepared some possible corrections to this formula, but these will be applied together with the bearing size scaling. The second area is connected to the valve lift scaling. Currently we use a scaling coeffi cient F( for the respective intake/exhaust lift vectors (equation 6, where L( is the valve lift and D( the valve diameter) based on Heywood’s formula, fi rst introduced in chapter 3.1 on main engine geometry. i ure 4 Engine block CA models (from left , , br e 4 CA modely blok (zleva , , ) The engine is by far the largest of the three – as expected. It is rather interesting, that the two other scenarios and are very similar from the package standpoint, though is slightly shorter. According to our CA model, all designs are feasible; however, variant is on the limit, and some of the design features would be a very tight fit. idth mm ei ht mm en th mm loc ass iston assembly ass Con-rod ass Con-rod en th mm max . 3 .1 192.3 13,1 , 3 , 219,9 05 1.2 231. 1 . , , 1 , 12 , 035 3 9. 232. 1 .9 , ,3 , 1 , able Engine blocks’ dimensions and estimated components’ masses abul a Rozm ry blok motor a odhadnut hmotnosti hlavn ch komponent . otential for future de elo ment of the o timi ation methodolo y The designs resulting from our optimizations are generally feasible and achieve high yet not unrealistic efficiencies for designs. espite these positive results, there is still some further potential in the development of our optimization methodology, especially from the viewpoint of the structural design, specifically in two areas. The first area is referenced to the fact, that we do not scale the main and conn-rod bearing diameters with bore size. And these are important inputs for Vyvaž friction model. A change in the con-rod bearing diameter would also influence the relation developed for con-rod ratio calculation (equation ). We have already prepared some possible corrections to this formula, but these will be applied together with the bearing size scaling. The second area is connected to the valve lift scaling. Currently we use a scaling coefficient for the respective intake/exhaust lift vectors (equation , where 𝐿𝐿 is the valve lift and 𝐷𝐷 the valve diameter) based on eywood’s formula, first introduced in chapter 3.1 on main engine geometry. 0 𝐿𝐿 𝐷𝐷 = = 0 𝐷𝐷 𝐿𝐿 ( ) This simple relation gives decent result from the thermodynamic perspective. Further increase in the intake valve lift from the optimal value of the design has a small additional positive effect on BSFC, which is clear from figure . (6) This simple relation gives decent result from the thermodynamic perspective. Further increase in the intake valve lift from the optimal value of the Vmax design has a small additional positive effect on BSFC, which is clear from Figure 5. In practice, the lift curve is usually limited by the valve train dynamics. Valve acceleration is determined by the lift curve shape – this acceleration relation can be simplifi ed to just valve lift and engine speed. However, REx ICE operates only in one primary operating point and it is not controlled by the driver: this means, that there is little chance of over -speeding the engine. The lift curves can be therefore designed exactly for this one operating point, even possibly saving some extra fuel. 6. CONCLUSION Our paper presents a multi -parametric, multi -objective design optimization methodology, that was tested on a case of three design scenarios for a four -stroke, V -twin, natural aspirated, spark -ignition REx ICE, operating in a single point REEV’s operation. We defi ned our REx engine concept, aiming at the best possible balance of achievable mass, package, overall design simplicity and therefore low cost, based on the literature research and using a design selection table. The other layouts that we considered were inline two -cylinder variants with or without balance shaft, and an opposed pistons engine – Boxer. Thermodynamic 0D/1D model was built within the GT -Suite simulation environment, with a special emphasis on predictive ability of its sub -models. Our paper briefl y discusses some of these sub -models: CTU in Prague in -house built friction model, GT -Suite’s phenomenological predictive SI combustion model, knock model, and fi nally the in -cylinder heat transfer model. Apart from the GT -Suite simulation model, our methodology uses also a parametric CAD model of the engine block. This CAD model provides some important input data for the GT -Suite sub- -models to enhance the optimization accuracy, and it is also used for the optimization result structural design’s feasibility check. Finally, the multi -parametric and multi -objective optimizations of three different design scenarios featuring different engine displacement with the same power output goal were carried out in a modeFRONTIER optimization platform, using a hybrid algorithm pilOPT. Resulting REx ICE designs are realistic, following the current trends of ICE effi ciency increase (Millerization and down -speeding) and power increase for natural aspirated engine. Our future work will mainly focus on further enhancements of our optimization methodology, adding more details of the crank- -train mechanism design into the simulation sub -models. Future simulation studies will consider different engine layouts. ACKNOWLEDGEMENTS This work was realized using support of: • Technological Agency, Czech Republic, programme National Competence Centres, project # TN01000026 Josef Bozek National Center of Competence for Surface Vehicles. FIGURE 5: BSFC dependence on the inlet valve lift for Vmax scenario OBRÁZEK 5: Závislost m#rné spot'eby na zdvihu sacího ventilu pro scéná' Vmax Range Extender ICE Multi -Parametric Multi -Objective Optimization MIKULÁ! ADÁMEK, RASTISLAV TOMAN MECCA 01 2021 PAGE 19 • The Grant Agency of the Czech Technical University in Prague, grant No. SGS19/104/OHK2/2T/12. This support is gratefully acknowledged. LIST OF ABBREVIATIONS BTDC Bottom Top Dead Center CAD Computer Aided Design CAE Computer Aided Engineering CTU Czech Technical University in Prague EGR Exhaust Gas Recirculation FE Finite Element I2 Inline two -cylinder Engine I3 Inline three -cylinder Engine ICE Internal Combustion Engine NVH Noise Vibration and Harshness OEM Original Equipment Manufacturer OHV Over Head Valve REEV Range Extended Electric Vehicle REx Range Extender SI Spark Ignition SOHC Single Over Head Camshaft TDC Top Dead Center V2 V -twin Engine WOT Wide Open Throttle FMEP Friction mean effective pressure BMEP Brake mean effective pressure BSFC Brake -specifi c fuel consumption CA Crank angle LIST OF SYMBOLS Af Flame area Bm Maximum laminar speed B) Laminar speed roll -of value CK Flame kernel growth multiplier CS Turbulent fl ame speed multiplier C$ Taylor length scale multiplier Dvex Exhaust valve diameter Dvin Intake valve diameter Fv Valve lift scaling coeffi cient Lvex Maximum exhaust valve lift Lvin Maximum intake valve lift Lt Turbulent length scale Mb Burnup mass Me Entrained mass Pe ICE brake power RB/S Bore/stroke ratio Rf Flame radius SL Laminar fl ame speed ST Turbulent fl ame speeds Vd ICE displacement Xk,max Pareto set’s maximum value of the objective function Xk Optimization objective function Cs Mean piston speed nICE Engine speed p0 Reference pressure rc Compression ratio %k Criterial function weight coeffi cient &m Mechanical effi ciency &v Volumetric effi ciency "u Density of unburned gas )m Fuel/air equivalence ratio at maximum laminar fl ame speed B Bore DEM Dilution exponent multiplier Dil Mass fraction of the residuals in the unburned zone F Criterial function S Stroke SA Spark advance angle c Con -rod ratio relation coeffi cient 1 n Con -rod ratio relation coeffi cient 2 p Pressure u´ Mean fl uctuating turbulent velocity * Pressure exponent $ Con -rod ratio ) Fuel/air equivalence ratio REFERENCES [1] Turner J, Blake D, Moore J, et al. (2010) The Lotus Range Extender Engine. SAE International 2010-01-2208:34. https://doi.org/10.4271/2010-01-2208 [2] Andert J, Kohler E, Niehues J, Schurmann G (2012) KSPG RANGE EXTENDER: A NEW PATHFINDER TO ELECTROMOBILITY. MTZ WORLDWIDE 2012:7. https://doi.org/10.1007/s38313-012-0170-1 [3] Agarwal A, Lewis A, Akehurst S, et al. Development of a Low Cost Production Automotive Engine for Range Extender Application for Electric Vehicles. SAE International 2016-01-1055. https://doi.org/10.4271/2016-01-1055 [4] Atzwanger M, Hubmann C, Schoeffmann W, et al. Two -Cylinder Gasoline Engine Concept for Highly Integrated Range Extender and Hybrid Powertrain Applications. SAE International 2010-09-28. https://doi.org/10.4271/2010-32-0130 [5] Fraidl GK, Fisher R, Hubmann C, et al. (2009) Range Extender Module: Enabler for Electric Mobility. ATZautotechnology 2009:40 – 49. https://doi.org/10.1007/BF03247140 Range Extender ICE Multi -Parametric Multi -Objective Optimization MIKULÁ! ADÁMEK, RASTISLAV TOMAN MECCA 01 2021 PAGE 20 [6] Bogomolov S, Dolecek V, Macek J, et al. (2014) Combining Thermodynamics and Design Optimization for Finding ICE Downsizing Limits: 2014-04-01. SAE International 2014. https://doi.org/10.4271/2014-01-1098 [7] Toman R, Brankov I (2018) Multi-Parametric and Multi- Objective Thermodynamic Optimization of a Spark- Ignition Range Extender Ice. Sciendo 2018:459 – 466. https://doi.org/0.5604/01.3001.0012.4368 [8] Bassett M, Thatcher I, Bisordi A, et al. Design of a Dedicated Range Extender Engine. SAE International 2011-01-0862. https://doi.org/10.4271/2011-01-0862 [9] HEYWOOD, John. Internal combustion engine fundamentals. 1. New York: McGraw-Hill, 1988. ISBN 978-0070286375. [10] Macek J, Fuente D, Emrich M (2011) A Simple Physical Model of ICE Mechanical Losses. SAE International 2011-01-0610. https://doi.org/10.4271/2011-01-0610 [11] Mirzaeian M, Millo F, Rolando L (2016) Assessment of the Predictive Capabilities of a Combustion Model for a Modern Downsized Turbocharged SI Engine. SAE International 2016-01-0557. https://doi.org/10.4271/2016-01-0557 [12] Rastislav T, Jan M (2017) Evaluation of the Predictive Capabilities of a Phenomenological Combustion Model for Natural Gas SI Engine. Journal of Middle European Construction and Design of Cars: The Journal of Czech Technical University 2017:37 – 48. https://doi.org/10.1515/MECDC20170007 [13] Fogla N, Bybee M, Mirzaeian M, et al. (2017) Development of a K-k-# Phenomenological Model to Predict In- -Cylinder Turbulence. SAE International Journal of Engines 10:562 – 575. https://doi.org/10.4271/2017-01-0542 [14] Woschni G (1967) A Universally Applicable Equation for the Instantaneous Heat Transfer Coeffi cient in the Internal Combustion Engine. SAE International 670931. https://doi.org/10.4271/670931 [15] Ra Y, Reitz R (2008) A reduced chemical kinetic model for IC engine combustion simulations with primary reference fuels. ScienceDirect 2008:713 – 738. https://doi.org/10.1016/j.combustfl ame.2008.05.002 [16] Adámek M (2020) OPTIMALIZACE TERMODYNAMIKY A KONSTRUKCE SPALOVACÍHO MOTORU PRO RANGE EXTENDER. Diplomová práce [17] ModeFrontier User Guide: Release 2018R3, 2018 ed.. ESTECO