• Energy Research

It is well known that 3D CFD simulations can give detailed information about fluid and flow properties in complex 3D domains while 1D CFD simulation can provide important information at a system level, i.e. about the performance of the entire engine. The drawbacks of the two simulation methods are that the former requires high computational cost while the latter is not able to capture complex local 3D features of the flow. Therefore, the two simulation methods are to be seen as complementary, indeed a coupling of the two approaches can benefit from the pros of the two methods while minimizing the cons. In particular, with a multi-scale modeling approach (1D-3D) it is possible to simulate large and complex domains by modeling the complex part with a 3D approach and the rest of the domain with a 1D approach. This paper describes an optimization cycle analysis of the unsteady flow of a single cylinder, two stroke gasoline engine using advanced numerical tools, which are in turn validated by means of experimental measurements. In particular, a 3D model (based on STAR-CD code) of the entire engine and a 1D-3D integrated fluid dynamics model (based on GT-POWER 1D and Converge-LITE 3D codes) is developed and applied for the representation of the geometrical domain and for the prediction of performance and gasdynamics in the whole intake and exhaust systems. The methodology allows users to accurately predict and deeply understand unsteady phenomena in the whole engine and capture the wave motion, which strongly affects the muffler 3D design in small two stroke engines equipped with resonance pipes.

AbstractAs discussed in the part I of this paper, 3D models represent a useful tool for a detailed description of the mean and turbulent flow fields inside the engine cylinder. 3D results are utilized to develop and validate a 0D phenomenological turbulence model, sensitive to the variation of operative parameters such as valve phasing, valve lift, engine speed, etc.In part II of this paper, a 0D phenomenological combustion model is presented, as well. It is based on a fractal description of the flame front and is able to sense each of the fuel properties, the operating conditions (air-to-fuel ratio, spark advance, boost level) and the combustion chamber geometry. In addition, it is capable to properly handle different turbulence levels predicted by means of the turbulence model presented in the part I.The turbulence and combustion models are included, through user routines, in the commercial software GT-Power™. With reference to a small twin-cylinder VVA turbocharged engine, the turbulence/combustion model, once properly tuned, is finally used to calculate in-cylinder pressure traces, rate of heat release and overall engine performance at full load operations and brake specific fuel consumption at part load, as well. An excellent agreement between numerical forecasts and experimental evidence is obtained.

A numerical study using large-eddy simulations (LES) to reproduce and understand sources of cycle-to-cycle variation (CCV) in spark-initiated internal combustion engines (ICEs) is presented. Two relevantly different spark-ignition (SI) units, that is, a homogeneous-charge slow-speed singlecylinder research unit (the transparent combustion chamber (TCC)-III, Engine 1) and a stratifiedcharge high-revving speed gasoline direct injection (GDI) (Engine 2) one, are analyzed in fired operations. Multiple-cycle simulations are carried out for both engines and LES results well reproduce the experimentally measured combustion CCV. A correlation study is carried out, emphasizing the decisive influence of the early flame period variability (1% of mass fraction burnt (MFB1)) on the entire combustion event in both ICEs. The focus is moved onto the early flame characteristics, and the crucial task to determine the dominant causes of its variability (if any) is undertaken. A two-level analysis is carried out: the influence of global parameters is assessed at first; second, local details in the ignition region are analyzed. A comparison of conditions at combustion onset is carried out and case-specific leading factors for combustion CCV are identified and ranked. Finally, comparative simulations are presented using a simpler flame deposition ignition model: the simulation flaws are evident due to modeling assumptions in the flame/flow interaction at ignition. The relevance of this study is the knowledge extension of turbulence-driven phenomena in ICEs allowed by advanced CFD (Computational Fluid Dynamics) simulations. The application to different engine types proves the soundness of the used models and it confirms that CCV is based on enginespecific factors. Simulations show how CCV originates from the interplay of small- and large-scale factors in Engine 1, due to the lack of coherent flows, whereas in Engine 2 the dominant CCV promoters are local air-to-fuel ratio (AFR) and flow velocity at ignition. This confirms the absence of a generally valid ranking, and it demonstrates the use of LES as a development and designorienting tool for next-generation engines.

AbstractOne of the most critical challenges for the specific power increase of turbocharged SI engines is the low end torque, limited by two aspects. First, the big size of the compressor necessary to deliver the maximum airflow does not allow high boost pressures at low speed, due to the surge line proximity. Second, the flame front velocity may become slower than the end gas auto-ignition rate, thus increasing the risk of knocking.This study is based on a current SI GDI V8 turbocharged engine, modeled by means of CFD tools, both 1d and 3d. The goal of the activity is to lower by 20% the displacement, without reducing brake torque, all over the engine speed range.It was decided to adopt a smaller bore, keeping stroke constant. Obviously, the combustion chamber, the valves and the intake-exhaust ports have been re-designed, as well as the whole intake and exhaust system. Instead of the two turbochargers, one for each bank of cylinders, a triple-turbocharger layout has been considered.The development of the engine has been carried out by means of 1D engine cycle simulations, using predictive knock models, calibrated with the support of both experiments and CFD-3d simulations. A few operating conditions for the final configuration have been also analyzed by means of a 3-d CFD tool.The paper presents the results of this activity, and describes in details the guidelines followed for the development of the engine.

In the recent past engine knock emerged as one of the main limiting aspects for the achievement of higher efficiency targets in modern spark-ignition (SI) engines. To attain these requirements, engine operating points must be moved as close as possible to the onset of abnormal combustions, although the turbulent nature of flow field and SI combustion leads to possibly ample fluctuations between consecutive engine cycles. This forces engine designers to distance the target condition from its theoretical optimum in order to prevent abnormal combustion, which can potentially damage engine components because of few individual heavy-knocking cycles. A statistically based RANS knock model is presented in this study, whose aim is the prediction not only of the ensemble average knock occurrence, poorly meaningful in such a stochastic event, but also of a knock probability. The model is based on look-up tables of autoignition times from detailed chemistry, coupled with transport equations for the variance of mixture fraction and enthalpy. The transported perturbations around the ensemble average value are based on variable gradients and on a local turbulent time scale. A multi-variate cell-based Gaussian-PDF model is proposed for the unburnt mixture, resulting in a statistical distribution for the in-cell reaction rate. An average knock precursor and its variance are independently calculated and transported; this results in the prediction of an earliest knock probability preceding the ensemble average knock onset, as confirmed by the experimental evidence. The proposed model estimates not only the regions where the average knock is promoted, but also where and when the first knock is more likely to be encountered. The application of the model to a RANS simulation of a modern turbocharged direct injection (DI) SI engine with optical access is presented and the analysis of the knock statistical occurrence obtained by the proposed model adds an innovative contribution to overcome the limitation of consolidated "average knock" analyses typical of a RANS approach.

The evaluation of the steady-flow discharge coefficient of ICE port assemble is known to be very sensitive to the capability of the turbulence sub-models in capturing the boundary layer dynamics. Despite the fact that theintrinsically unsteady phenomena related to flow separation claim for LES approach, the present paper aims to demonstrate that RANS simulation can provide reliable design-oriented results by using low-Reynolds cubic k-e turbulence models. Different engine intake port assemblies and pressure drops have been simulated by using the CFD STAR-CD code and numerical results have been compared versus experiments in terms of both global parameters, i.e. the discharge coefficient, and local parameters, by means of static pressure measurements along the intake port just upstream of the valve seat. Computations have been performed by comparing two turbulence models: Low-Reynolds cubic k-e and High-Reynolds cubic k-e. The analysis leaded to remarkable assessments in the definition of a correct and reliable methodology for the evaluation of engine port breathing capabilities. Comparison between numerical results and experiments showed that the low-Reynolds cubic k-e model is unavoidable to correctly capture the influence of port feature variations on engine permeability. In particular, the deficiencies demonstrated by High-Reynolds cubic k-e turbulence model in resolving the influence of near-wall shear and adverse pressure gradient effect on boundary layer dynamics are completely overcome by the use of the Low-Reynolds formulation.

The paper presents a combined experimental and numerical activity carried out to improve the accuracy of conjugate heat transfer CFD simulations of a high-performance S.I. engine water cooling jacket.Due to the complexity of the computational domain, which covers both the coolant jacket and the surrounding metal cast (both head and block), particular care is required in order to find a tradeoff between the accuracy and the cost-effectiveness of the numerical procedure. In view of the presence of many complex physical phenomena, the contribution of some relevant CFD parameters and sub-models is separately evaluated and discussed.Among the formers, the extent of the computational domain, the choice of a proper set of boundary conditions and the detailed representation of the physical properties of the involved materials are separately considered. Among the latters, the choice between a simplified single-phase approach and a more complex two-phase approach taking into account the effects of phase transition within the engine coolant is discussed.The predictive capability of the CFD-CHT methodology is assessed by means of the comparison between CFD results and experimental measurements provided by the engine manufacturer for different engine operating conditions.At the end of the validation process, a methodology for the correct and cost-effective characterization of conjugate heat transfer is proposed, showing a reasonable trade-off between the predictive capability and the computational effort of the simulations.