Search

Accurate knowledge of fuel spray behavior is of utmost importance for liquid-fuel-based combustion systems. Fuel properties, injector geometry, operating conditions, and thermal state of the combustion chamber determine the fuel’s ability to mix and burn efficiently. Three-dimensional computational-fluid-dynamics models can reveal the complex dynamics of the injector’s internal flow, as well as the spray breakup, evaporation, mixing, and combustion. However, time and length scales of in-nozzle flow and ensuing spray can differ by several orders of magnitude, limiting the feasibility of a simultaneous solution of the entire chain of physics. This work explores an end-to-end approach to decouple the problem at the injector outlet via a static-coupling framework. Flowfields are sampled at the injector exit, stored into spatiotemporally resolved maps, and used to initialize a Lagrangian spray whose properties reflect the flow’s instantaneous state as predicted by the in-nozzle flow simulations. Comparisons against typical rate-of-injection results and qualitative validation against optical spray data highlighted the ability of static coupling to unveil spray physics that would otherwise be missed.

Improving the efficiency of internal combustion engines can result in lower operating costs and enable reduced total cost of vehicle ownership. On the other hand, it is of utmost importance that the development of engine technologies is also directed toward curbing climate-altering and harmful emissions. One such technology is gasoline compression ignition, which can offer increased fuel efficiency and reduced emissions and criteria pollutants. By means of modal analysis methods, this study focuses on characterizing the internal flow behavior of a production, eight-hole, medium-duty diesel injector operating with gasoline-like fuels under typical diesel injection conditions. High-resolution tomographic reconstructions of the injector tip geometry, along with needle motion measurements, were obtained via X-ray measurements. This information was combined with a previously validated computational setup to perform a series of in-nozzle flow computational fluid dynamics (CFD) simulations. The CFD results were analyzed to identify the connection between the injector’s global behavior and relevant flowfields such as velocity and vorticity. The results obtained with the X-ray geometry were then compared against those obtained with a nominal geometry to investigate the effect of the injector’s geometric features on its behavior. Variations to the K-factor and inlet ellipticity of the orifices were explored, and these parameters’ ability to reduce in-orifice cavitation was assessed by analyzing the predicted in-orifice fuel vapor volume fraction. Relatively large and persistent flow structures arising within the injector sac were identified as the main driver for orifice-to-orifice variability, as their location and size impacted the amount of fuel delivered by each orifice. Assessing the extent to which these energetic structures impact the injector performance is of interest; thus, it is desirable to investigate the most energetically significant flowfield structures as a function of modifications to specific features of the injector geometry. Modal decomposition tools, such as space-only proper orthogonal decomposition (POD) and spectral proper orthogonal decomposition (SPOD) were applied to gain insight into dominant coherent structures that persist over the early phase of the needle opening event. It was found that the geometrical modifications applied resulted in variations of the modal energy content, as well as in the leading dominant spatial mode features. Additionally, it was found that the dominant frequency associated with the most significant energetic content was largely invariant to the geometrical modifications, with the difference between realistic and modified geometries being mostly found in the modal energy content.

To achieve carbon neutral ambition, hydrogen (H2) has recently received significant attention as a zerocarbon fuel for internal combustion engines (ICEs) across transportation sectors. As a critical element in the analysis-led design process, a hydrogen kinetic mechanism needs to be thoroughly evaluated to support the development of high-efficiency H2-ICE combustion system concepts. In this study, recently published H2 kinetic mechanisms were reviewed and down-selected for evaluations against available laboratory data in ignition delay time (IDT) and laminar flame speed (LFS) measurements. The examination was subsequently extended to high-fidelity three-dimensional (3-D) computational fluid dynamics (CFD), spark-ignited, H2 engine simulations. Discrepancies identified at engine-relevant conditions led to a kinetics tailoring campaign based on the H2 mechanism developed by Burke et al. (2012). Selected reactions identified via global sensitivity analysis were optimized under the engine-relevant pressure-temperature conditions. The reaction rate coefficients were adjusted within the experimental and theoretical uncertainty limits by adopting a Monte-Carlo sampling approach as a searching algorithm to generate candidate mechanisms. Finally, the optimized mechanism was validated sequentially from low-dimensional (0-D and 1-D) to high-fidelity 3D CFD engine simulations. Overall, the optimized H2 kinetic model led to significantly improved predictions on capturing engine in-cylinder pressure trace and heat release rate.

A closed-cycle, three-dimensional (3D) computational fluid dynamics (CFD) analysis campaign was conducted to evaluate the performance of using spark plugs to assist gasoline compression ignition (GCI) combustion during cold idle operations. A conventional spark plug using single-sided J-strap design was put at a location on the cylinder head to facilitate spray-guided spark assistance. Ignition was modeled with an L-type energy distribution to depict the breakdown and the arc-to-glow phases during the energy discharge process. Several key design parameters were investigated, including injector clocking, number of nozzle holes, spray inclusion angle, number of fuel injections, fuel split ratio, and fuel injection timings. The study emphasized the region around the spark gap, focusing on flame kernel formation and development and local equivalence ratio distribution. Flame kernel development and the ignition process were found to correlate strongly with the fuel stratification and the flow velocity near the spark gap. The analysis results showed that the flame kernel development followed the direction of the local flow field. In addition, the local fuel stratification notably influenced early-stage flame kernel development due to varying injection spray patterns and the fuel injection strategies. Among these design parameters, the number of nozzle holes and fuel injection timing had the most significant effects on the engine combustion performance.

Gasoline compression ignition (GCI) is a promising powertrain solution to simultaneously address the increasingly stringent regulation of oxides of nitrogen (NOx) and a new focus on greenhouse gases. GCI combustion benefits from extended mixing times due to the low reactivity of gasoline, but only when held beneath the threshold of the high temperature combustion regime. The geometric compression ratio (GCR) of an engine is often chosen to balance the desire for low NOx emissions while maintaining high efficiency. This work explores the relationship between GCR, variable valve actuation (VVA) and emissions when using GCI combustion strategies. The test article was a Cummins ISX15 heavy-duty diesel engine with an unmodified production air and fuel system. The test fuel was an ethanol-free gasoline with a market-representative research octane number (RON) of 91.4–93.2. In the experimental investigation at 1375 rpm/10 bar BMEP, three engine GCRs were studied, including 15.7, 17.3, and 18.9. Across the three GCRs, GCI exhibited a two-stage combustion process enabled through a split injection strategy. When keeping both NOx and CA50 constant, varying GCR from 15.7 to 18.9 showed only a moderate impact on engine brake thermal efficiency (BTE), while its influence on smoke was pronounced. At a lower GCR, a larger fraction of fuel could be introduced during the first injection event due to lower charge reactivity, thereby promoting partially-premixed combustion and reducing smoke. Although increasing GCR increased gross indicated thermal efficiency (ITEg), it was also found to cause higher energy losses in friction and pumping. In contrast, GCI performance showed stronger sensitivity towards EGR rate variation, suggesting that air-handling system development is critical for enabling efficient and clean low NOx GCI combustion. To better utilize gasoline’s lower reactivity, an analysis-led variable valve actuation investigation was performed at 15.7 GCR and 1375 rpm/10 bar BMEP. The analysis was focused on using an early intake valve closing (EIVC) approach by carrying out closed-cycle, 3-D CFD combustion simulations coupled with 1-D engine cycle analysis. EIVC was shown to be an effective means to lengthen ignition delay and promote partially-premixed combustion by lowering the engine effective compression ratio (ECR). By combining EIVC with a tailored fuel injection strategy and properly developed thermal boundary conditions, simulation predicted a 2.3% improvement in ISFC and 47% soot reduction over the baseline IVC case while keeping NOx below the baseline level.

Measurements of fuel injectors via non-destructive X-ray techniques can provide unique insights about an injector’s internal surface. Using real measured geometry rather than nominal design geometry in computational fluid dynamics simulations can improve the accuracy of the numerical models dramatically. Recent work from the authors investigated the influence of the injector design on the internal flow development and occurrence of cavitation in a production multi-hole heavy-duty diesel injector operating with a straight-run gasoline for gasoline compression ignition (GCI) applications. This was achieved by evaluating a series of design parameters which showed that the intensity and duration of cavitation structures could be mitigated by acting on certain injector parameters such as K-factor, orifice inlet ellipticity, and sac-to-orifice radius of curvature. In the present work, the findings from the previous parametric study were combined to generate two attempts at improving the injector design and numerically evaluate their ability to suppress cavitation inside the orifices at three levels of injection pressure (1000, 1500, and 2500 bar), while operating with the same high-volatility gasoline fuel. Qualitative and quantitative analyses showed that, compared to the results obtained with the original X-ray scanned geometry, the improved designs were able to prevent fuel vapor formation at the two lowest injection pressures and avoid super-cavitation at the higher pressure. It was shown that these results were due to the strong influence that the orifice shape can have on the pressure and fuel vapor volume fraction distributions within the orifices. The informed design choices proposed in this study can therefore be vital for extending the durability and reliability of heavy-duty injectors for GCI applications.

A design optimization campaign was conducted to search for improved combustion profiles that enhance gasoline compression ignition in a heavy-duty diesel engine with a geometric compression ratio of 17.3. Three-dimensional computational fluid dynamics simulations were employed using the software package CONVERGE. A large-scale design of experiments (DoE) approach was used for the optimization. The main parameters explored include geometric features, injector specifications, and swirl motion. Both stepped-lip bowls and re-entrant bowls were included in the optimization effort in order to assess their respective performance implications. A total of 256 design candidates were prepared using the software package CAESES for automated and simultaneous geometry generation and combustion recipe perturbation. The design optimization was conducted for three engine load points representing light to medium load conditions. The design candidates were evaluated for fuel efficiency, emissions, fuel-air mixing characteristics, and global combustion behavior. Simulation results show that the optimum designs were all stepped-lip bowls, which exhibited better overall performance than re-entrant bowls due to improvements in fuel-air mixing, as well as reduced heat loss and emissions formation. Improvements in indicated specific fuel consumption of up to 3.2% were achieved while meeting engine-out NOx emission targets of 1–1.5 g/kW·hr. Re-entrant bowls performed worse compared to the baseline design, and significant performance variations occurred across the load points. Specifically, the re-entrant bowls were on par with the stepped-lip bowls under light load conditions, but significant deteriorations occurred under higher load conditions. As a final task, selected optimized designs were then evaluated under simulated full-load conditions.

The differences between a center-mounted and a side-mounted injector for gasoline direct injection (GDI) applications are analyzed through computational fluid dynamics (CFD). The Engine Combustion Network’s (ECN) axisymmetric 8-hole Spray G injector is compared to a 6-hole injector designed to be side-mounted in an engine. Nozzle-flow simulations are carried out with the commercial CFD software CONVERGE, injecting Euro 5 certification gasoline into a constant volume chamber. Low-load operating conditions are targeted, setting the injection pressure at 50 bar and the ambient pressure to be representative of very early pilot injections. The phase change is handled with the Homogeneous Relaxation Model (HRM), which is assessed and adapted to gasoline flash-boiling conditions. The simulation domains are generated leveraging real injector internal geometries obtained by micron-resolution X-ray tomographic measurements, which introduce manufacturing tolerances and surface roughness in the computational study. Steady needle lift conditions are analyzed. The near-field fuel density distributions and plume morphologies are evaluated, validated and compared to X-ray radiography measurements. A computational best practice is defined and single plume characteristics and variability trends are highlighted as functions of the geometry of the orifices. The plume-plume interaction dynamics are identified and assessed, underlining differences from center- to side-mounted injectors at strong flashing conditions. The obtained numerical framework allows the identification of near-nozzle injection characteristics such as single plume direction, cone angle, spray initial velocity and spatial fuel density distribution. The presented results represent a unique dataset for the initialization of more-affordable Lagrangian spray models, which differentiate the behavior of side-mounted and center-mounted injectors.

Starting compression ignition engines under cold conditions is extremely challenging, due to insufficient fuel vaporization, heavy wall impingement, and low ignitability of the fuel. For gasoline compression ignition (GCI) combustion strategies, which offer the potential for an enhanced NOx-PM tradeoff with diesel-like fuel efficiency, robust ignition and combustion in very cold conditions pose a significant challenge due to the low reactivity of gasoline fuels. Based on the previous understanding of the spray, ignition and combustion processes for a GCI engine under cold conditions, this study focuses on investigating the cold combustion performance of a heavy-duty GCI engine with glow plug ignition assist. Glow plugs, commonly used for low temperature cold starts in diesel engines, are used to pre-heat a segment of the mixture to improve its ignitability. Here, CFD studies are carried out to explore the influence of a spray-guided glow plug on the spray and combustion behavior of a GCI engine under cold operating conditions. In a prior study, the underlying CFD model has been validated using experimental data from a six-cylinder, 15 L heavy-duty diesel engine operating with a compression ratio (CR) of 17.3 at a 600 rpm cold idle condition with RON92 E0 gasoline. The energy intensity required by the glow plug to deliver stable combustion isparametrically studied. The size and location of the glow plug are also parametrically varied to evaluate their effects on the combustion process. The influence of the glow plug on the in-cylinder mixture distribution and the ensuing combustion process is also investigated. In particular, the localized fuel spray distribution and mixture formation near the glow plug are examined. The results reveal that the glow plug enhances GCI combustion under cold idle conditions and that the spray-guided glow plug improves fuel vaporization, leading to a rich mixture near the glow plug and an enhancement of the combustion efficiency. In addition, the effectiveness of the glow plug at a low ambient temperature of 0°C and a 200 rpm cold start condition is evaluated. These simulations suggest that the glow plug can improve the cold start performance of a GCI engine.

Open-cycle engine simulations of a passive pre-chamber operating inside a direct-injected gasoline engine are performed and analyzed in this study. A comprehensive three-dimensional computational fluid dynamics (CFD) model has been formulated with state-of-the-art physical sub-models to account for the complex processes of engine gas exchange, fuel injection and mixing, ignition, and combustion. Numerical results are validated against experimental measurements under low load condition with high internal EGR. Realistic modeling considerations are discussed to ensure proper fidelity. In particular, the mixture conditions and flow motions could present very different features between the pre- and main chamber, which requires comprehensive simulation of the full engine cycle and imposes challenges for combustion modeling. Practically validated chemical kinetics models are essential for proper prediction of cylinder pressure history of pre-chamber jet combustion systems. Detailed analysis is then carried out to highlight key processes associated with pre-chamber operation, including residual scavenging, fuel/air mixture formation, flow pattern and turbulence development within the pre-chamber, and the ignition of main chamber mixture by issued turbulent jets. Numerical evaluation of pre-chamber design variants has been attempted, and less commonly investigated geometry parameters such as swirl nozzles and nozzle umbrella angle are found impactful for pre-chamber ignition performances.

Gasoline compression ignition (GCI) engines are considered an attractive alternative to traditional spark-ignition and diesel engines. In this work, a Machine Learning-Grid Gradient Ascent (ML-GGA) approach was developed to optimize the performance of internal combustion engines. ML offers a pathway to transform complex physical processes that occur in a combustion engine into compact informational processes. The developed ML-GGA model was compared with a recently developed Machine Learning-Genetic Algorithm (ML-GA). Detailed investigations of optimization solver parameters and variable limit extension were performed in the present ML-GGA model to improve the accuracy and robustness of the optimization process. Detailed descriptions of the different procedures, optimization tools, and criteria that must be followed for a successful output are provided here. The developed ML-GGA approach was used to optimize the operating conditions (case 1) and the piston bowl design (case 2) of a heavy-duty diesel engine running on a gasoline fuel with a research octane number (RON) of 80. The ML-GGA approach yielded >2% improvements in the merit function, compared with the optimum obtained from a thorough computational fluid dynamics (CFD) guided system optimization. The predictions from the ML-GGA approach were validated with engine CFD simulations. This study demonstrates the potential of ML-GGA to significantly reduce the time needed for optimization problems, without loss in accuracy compared with traditional approaches.

Soot formation in an n-dodecane spray flame under diesel engine conditions, known as Spray A, is modelled with the transported probability function (TPDF) method. The approach employs an acetylene-based two-equation soot model coupled with a Reynolds-averaged turbulence model and a Lagrangian discrete phase spray model. The aims are to evaluate, in the context of soot, the predictive capability of the model, the effects of turbulence–chemistry interactions (TCI), and various available chemistry mechanisms. TCI effects are evaluated by comparisons between the TPDF model and simulations using a well-mixed model neglecting turbulent fluctuations. Five test cases having variations in ambient temperature and oxygen concentration are considered. Five chemical mechanisms are first compared to experiments in terms of their ignition delay (ID) and lift-off length (LOL) under ambient O2 and temperature variations. Three relatively new mechanisms exhibit good ID performance (with both TCI approaches), while two short mechanisms also provide good LOL performance in conjunction with the TPDF approach. The two short mechanisms are considered for further comparisons. The auto-ignition process is analysed by comparing TPDF simulations with measurements from schlieren and 355 nm planar laser-induced fluorescence, detecting CH2O and polycyclic aromatic hydrocarbons, with overall good qualitative agreement, though with some differences on low-temperature reactivity. The experimental comparisons for soot consider transient soot mass and KL in the baseline condition, and steady state soot volume fraction (SVF) fields and total soot masses for all five ambient conditions. In terms of comparisons to experiment, the transient stage of the soot mass development is not well captured. An analysis of the transient KL in the baseline case shows the soot-containing region is larger in the experiment than the model, with soot extending in the experiment much closer to the jet boundary, suggesting that the model underestimates gradients around the jet head. During the steady period, however, the SVF agrees quite well. The soot models with both chemistry and TCI approaches were able to reproduce the overall soot trends with varying ambient temperature and oxygen, though the effect of the ambient temperature on the soot mass was under-predicted, in particular in its variation from 900 to 1000 K. TCI effects on soot were in overall terms relatively minor, in part due to compensating errors. Neglecting TCI showed generally higher peak soot amounts, narrower soot distributions, and more downstream soot onset and soot peak locations. These differences between the models are explained with the help of a detailed analysis of the soot phenomena.

Fuel physical properties are known to influence in-nozzle flow behavior, in turn affecting spray formation in internal combustion engines. A series of 3D simulations was performed to model the internal nozzle flow in a five-hole mini-sac diesel injector. The goal of the study was to evaluate the behavior of two gasoline-like fuels (full-range naphtha and light naphtha) and compare them against n-Dodecane, selected from a palette used as a diesel surrogate. Simulations were carried out using a multi-phase flow representation based on the mixture model assumption with the Volume of Fluid (VOF) method, and including cavitation effects by means of the Homogeneous Relaxation Model (HRM). Validated methodologies from our previous studies were employed to account for full needle motion. Detailed simulations revealed the influence of the fuel properties on injector performance, injected fuel energy and propensity to cavitation. The three fuels were compared with respect to global parameters such as mass flow rate and area contraction coefficients, and local parameters such as pressure and velocity distribution inside the sac and orifices. Parametric investigations were also performed to understand the fuel response to changes in the fuel injection temperature, injection pressure, and geometry details. Cavitation magnitude was observed to be strongly associated with the values of saturation pressure. Owing to their higher volatility, the two gasoline-like fuels were observed to cavitate more than n-Dodecane across all the investigated conditions. While at full needle lift cavitation was reduced for all fuels, during the injection transients the gasoline-like fuels showed more propensity to cavitate inside the orifice and seat regions. This is expected to have a profound influence on nozzle erosion. Although full-range and light naphtha have lower densities compared to n-Dodecane, owing to their lower viscosity, the mass flow rate differences between the naphtha fuels and n-Dodecane were small. The analysis of fuel energy content showed that the higher lower heating value (LHV) of light naphtha helped compensate for the slightly lower total delivered mass.

A skeletal mechanism with 54 species and 269 reactions was developed to predict pyrolysis and oxidation of n-dodecane as a diesel fuel surrogate involving both high-temperature (high-T) and low-temperature (low-T) conditions. The skeletal mechanism was developed from a semi-detailed mechanism developed at the University of Southern California (USC). Species and reactions for high-T pyrolysis and oxidation of C5-C12 were reduced by using reaction flow analysis (RFA), isomer lumping, and then merged into a skeletal C0-C4 core to form a high-T sub-mechanism. Species and lumped semi-global reactions for low-T chemistry were then added to the high-T sub-mechanism and a 54-species skeletal mechanism is obtained. The rate parameters of the low-T reactions were tuned against a detailed mechanism by the Lawrence Livermore National Laboratory (LLNL), as well as the Spray A flame experimental data, to improve the prediction of ignition delay at low-T conditions, while the high-T chemistry remained unchanged. The skeletal mechanism was validated for auto-ignition, perfectly stirred reactors (PSR), flow reactors and laminar premixed flames over a wide range of flame conditions. The skeletal mechanism was then employed to simulate three-dimensional turbulent spray flames at compression ignition engine conditions and validated against experimental data from the Engine Combustion Network (ECN).

The effects of strain rate history on turbulent flames have been studied in the past decades with 1D counter flow diffusion flame (CFDF) configurations subjected to oscillating strain rates. In this work, these unsteady effects are studied for complex hydrocarbon fuel surrogates at engine relevant conditions with unsteady strain rates experienced by flamelets in a typical spray flame. Tabulated combustion models are based on a steady scalar dissipation rate (SDR) assumption and hence cannot capture these unsteady strain effects; even though they can capture the unsteady chemistry. In this work, 1D CFDF with varying strain rates are simulated using two different modeling approaches: steady SDR assumption and unsteady flamelet model. Comparative studies show that the history effects due to unsteady SDR are directly proportional to the temporal gradient of the SDR. A new equivalent SDR model based on the history of a flamelet is proposed. An averaging procedure is constructed such that the most recent histories are given higher weights. This equivalent SDR is then used with the steady SDR assumption in 1D flamelets. Results show a good agreement between tabulated flamelet solution and the unsteady flamelet results. This equivalent SDR concept is further implemented and compared against 3D spray flames (Engine Combustion Network Spray A). Tabulated models based on steady SDR assumption under-predict autoignition and flame lift-off when compared with an unsteady Representative Interactive Flamelet (RIF) model. However, equivalent SDR model coupled with the tabulated model predicted autoignition and flame lift-off very close to those reported by the RIF model. This model is further validated for a range of injection pressures for Spray A flames. The new modeling framework now enables tabulated models with significantly lower computational cost to account for unsteady history effects.

This study computationally investigates the potential of utilizing gasoline compression ignition (GCI) in a heavy-duty diesel engine to address a future ultra-low tailpipe NOx standard of 0.027 g/kWh while achieving high fuel efficiency. By conducting closed-cycle, full-geometry, 3-D computational fluid dynamics (CFD) combustion simulations, the effects of piston bowl geometry, injector spray pattern, and swirl ratio (SR) were investigated for a market gasoline. The simulations were performed at 1375 rpm over a load range from 5 to 15 bar BMEP. The engine compression ratio (CR) was increased from 15.7 used in previous work to 16.5 for this study. Two piston bowl concepts were studied with Design 1 attained by simply scaling from the baseline 15.7 CR piston bowl, and Design 2 exploring a wider and shallower combustion chamber design. The simulation results predicted that through a combination of the wider and shallower piston bowl design, a 14-hole injector spray pattern, and a swirl ratio of 1, Design 2 would lead to a 2–7% indicated specific fuel consumption (ISFC) improvement over the baseline by reducing the spray-wall interactions and lowering the in-cylinder heat transfer loss. Design 1 (10-hole and SR2) showed a more moderate ISFC reduction of 1–4% by increasing CR and the number of nozzle holes. The predicted fuel efficiency benefit of Design 2 was found to be more pronounced at low to medium loads.

Gasoline compression ignition (GCI) engines are considered an attractive alternative to traditional spark-ignition and diesel engines. In this work, a Machine Learning-Grid Gradient Algorithm (ML-GGA) approach was developed to optimize the performance of internal combustion engines. Machine learning (ML) offers a pathway to transform complex physical processes that occur in a combustion engine into compact informational processes. The developed ML-GGA model was compared with a recently developed Machine learning Genetic Algorithm (ML-GA). Detailed investigations of optimization solver parameters and variables limits extension were performed in the present ML-GGA model to improve the accuracy and robustness of the optimization process. Detailed descriptions of the different procedures, optimization tools and criteria that must be followed for a successful output are provided here. The developed ML-GGA approach was used to optimize the operating conditions (case 1) and the piston bowl design (case 2) of a heavy-duty diesel engine running on a gasoline fuel with a Research Octane Number (RON) of 80. The ML-GGA approach yielded > 2% improvements in the merit function, compared to the optimum obtained from a thorough computational fluid dynamics (CFD) guided system optimization. The predictions from the ML-GGA approach were validated with engine CFD simulations. This study demonstrates the potential of ML-GGA to significantly reduce the time needed for optimization problems, without loss in accuracy compared to traditional approaches.

Future demand for middle distillates in the commercial sector is expected to rise in step with increased economic activity in much of the world. Simultaneously, the passenger car fleet will experience various degrees of electrification and demand for gasoline is expected to fall. This divergence in demand will offer an opportunity to commercial vehicle owners if gasoline fuel streams can be burned efficiently in commercial engines.

Gasoline compression ignition (GCI) engine technology has shown the potential to achieve high fuel efficiency with low criteria pollutant emissions. In order to guide the design and optimization of GCI combustion, it is essential to develop high-fidelity simulation tools. Building on the previous work in computational fluid dynamic (CFD) simulations of spray combustion, this work focuses on predicting the soot emissions in a constant-volume vessel representative of heavy-duty diesel engine applications for an ultra-low sulfur diesel (ULSD) and a high reactivity (Research Octane Number 60) gasoline, and comparing the soot evolution characteristics of the two fuels. Simulations were conducted using both Reynolds Averaged Navier-Stokes (RANS) and Large Eddy Simulation (LES) turbulence models. Extensive model validations were performed against the experimental soot emissions data for both fuels. It was found that the simulation results using the LES turbulence model agreed better with the measured ignition delays and liftoff lengths than the RANS turbulence model. In addition, two soot models were evaluated in the current study, including an empirical two-step soot formation and oxidation model, and a detailed soot model that involves poly-aromatic hydrocarbon (PAH) chemistry. Validations showed that the separation of the flame lift-off location and the soot lift-off location and the relative natural luminosity signals were better predicted by the detailed soot model combined with the LES turbulence model. Qualitative comparisons of simulated local soot concentration distributions against experimental measurements in the literature confirmed the model’s performance. CFD simulations showed that the transition of domination from soot formation to soot oxidation was fuel-dependent, and the two fuels exhibited different temporal and spatial characteristics of soot emissions. CFD simulations also confirmed the lower sooting propensity of gasoline compared to ULSD under all investigated conditions.

This work evaluates different optimization algorithms for Computational Fluid Dynamics (CFD) simulations of engine combustion. Due to the computational expense of CFD simulations, emulators built with machine learning algorithms were used as surrogates for the optimizers. Two types of emulators were used: a Gaussian Process (GP) and a weighted variety of machine learning methods called SuperLearner (SL). The emulators were trained using a dataset of 2048 CFD simulations that were run concurrently on a supercomputer. The Design of Experiments (DOE) for the CFD runs was obtained by perturbing nine input parameters using a Monte Carlo method. The CFD simulations were of a heavy duty engine running with a low octane gasoline-like fuel at a partially premixed compression ignition mode. Ten optimization algorithms were tested, including types typically used in research applications. Each optimizer was allowed 800 function evaluations and was randomly tested 100 times. The optimizers were evaluated for the median, minimum, and maximum merits obtained in the 100 attempts. Some optimizers required more sequential evaluations, thereby resulting in longer wall clock times to reach an optimum. The best performing optimization methods were particle swarm optimization (PSO), differential evolution (DE), GENOUD (an evolutionary algorithm), and Micro-Genetic Algorithm (GA). These methods found a high median optimum as well as a reasonable minimum optimum of the 100 trials. Moreover, all of these methods were able to operate with less than 100 successive iterations, which reduced the wall clock time required in practice. Two methods were found to be effective but required a much larger number of successive iterations: the DIRECT and MALSCHAINS algorithms. A random search method that completed in a single iteration performed poorly in finding 1 Currently at Southwest Research Institute, San Antonio, Texas optimum designs, but was included to illustrate the limitation of highly concurrent search methods. The last three methods, Nelder-Mead, BOBYQA, and COBYLA, did not perform as well.

With a view to understanding ignition and combustion behaviours in diesel engines, this study investigates several aspects of ignition and combustion of an n-dodecane spray in a high pressure, high temperature chamber, known as Spray A, using data resulting from modelling using the transported probability density function (TPDF) method. The model has been validated comprehensively with good to excellent agreement in our previous work against all available experimental data including for mixture-fraction and velocity fields in non-reacting cases, and flame lift-off length and ignition delay in reacting cases. This good agreement encourages further investigation of the numerical model results to help understand the structure of this flame, which serves to complement the experimental information that is available, which is very limited due to the difficult experimental conditions in which this flame exists. For example, quantitative experimental measurements of local mixture-fraction, temperature, velocity gradients, etc. are not yet possible in reacting cases. Analysis of the model results shows that two-stage ignition is found to occur across the ambient temperature conditions considered: the first stage is rapidly initiated on the lean side where temperatures are high and sequentially moves to richer, cooler conditions. The first stage is extremely resilient to turbulence, occurring in a region of very low Damkohler number. The second stage of ignition occurs first in rich mixtures in a region behind the head of the fuel jet where mixture gradients are low, and appears to be influenced strongly by turbulence. Relative to a homogeneous reactor, it is delayed on the lean side but advanced on the rich side, suggesting entrainment and mixing from the early igniting lean regions into richer mixtures is an important moderator of the ignition process. The second-stage ignition front propagates at very high velocities initially, suggesting it is a sequential ignition moving according to gradients of ignition delay and/or residence time. The flame stabilises however on the lean side in a region of much lower velocity, where turbulent velocity fluctuations are sufficiently high such that turbulent transport influences the propagation. It stabilises in a region of low Damkohler number which implies that a competition of chemistry versus micro-mixing might also be involved in stabilisation. The stabilisation mechanism is investigated by an analysis of the transport budgets, showing the flame is stabilised by autoignition but moderated by turbulent diffusion. Further analysis of the flame index supports this stabilisation mechanism, and demonstrates the simultaneous existence of non-premixed and premixed combustion modes in the same flame. Analysis of the flow fields also reveals that local entrainment and dilatation are important flow features near the flame base.

An n-dodecane spray flame (Spray A from Engine Combustion Network) was simulated using a δ function combustion model along with a dynamic structure large eddy simulation (LES) model to evaluate its performance at engine-relevant conditions and to understand the transient behavior of this turbulent flame. The liquid spray was treated with a traditional Lagrangian method and the gas-phase reaction was modeled using a δ function combustion model. A 103-species skeletal mechanism was used for the n-dodecane chemical kinetic model. Significantly different flame structures and ignition processes are observed for the LES compared to those of Reynolds-averaged Navier − Stokes (RANS) predictions. The LES data suggests that the first ignition initiates in a lean mixture and propagates to a rich mixture, and the main ignition happens in the rich mixture, preferably less than 0.14 in mixture fraction space. LES was observed to have multiple ignition spots in the mixing layer simultaneously while the main ignition initiates in a clearly asymmetric fashion. The temporal flame development also indicates the flame stabilization mechanism is auto-ignition controlled. Soot predictions by LES present much better agreement with experiments compared to RANS, both qualitatively and quantitatively. Multiple realizations for LES were performed to understand the realization to realization variation and to establish best practices for ensemble-averaging diesel spray flames. The relevance index analysis suggests that an average of 5 and 6 realizations can reach 99% of similarity to the target average of 16 realizations on the mixture fraction and temperature fields, respectively. However, more realizations are necessary for the hydroxide (OH) and soot mass fractions due to their high fluctuations.