Search

Fuel efficient Homogeneous Charge Compression Ignition (HCCI) engine combustion phasing predictions must contend with non-linear chemistry, non-linear physics, near chaotic period doubling bifurcation(s), turbulent mixing, model parameters that can drift day-to-day, and air-fuel mixture state information that cannot typically be resolved on a cycle-to-cycle basis, especially during transients. Unlike many contemporary modeling approaches, this work does not attempt to solve for the myriad of combustion processes that are in practice unobservable in a metal engine. Instead, this work treads closely to physically measurable quantities within the framework of an abstract discrete dynamical system that is explicitly designed to capture many known combustion relationships, without ever explicitly solving for them. This abstract dynamical system is realized with an Extreme Learning Machine (ELM) that is extended to adapt to the combustion process from cycle-to-cycle with a new Weighted Ring-ELM algorithm. Combined, the above techniques are shown to provide unprecedented cycle-to-cycle predictive capability during transients, near chaotic combustion, and at steady-state, right up to complete misfire. These predictions only require adding an in-cylinder pressure sensor to production engines, which could cost as little as $13 per cylinder. By design, the framework is computationally efficient, and the approach is shown to predict combustion in sub-millisecond real-time using only an iPhone generation 1 processor (the $35 Raspberry Pi). This is in stark contrast to supercomputer approaches that model down to the minutiae of individual reactions but have yet to demonstrate such fidelity against cycle-to-cycle experiments. Finally, the feasibility of cycle-to-cycle model predictive control with this real-time framework is demonstrated.

An extended coherent flamelet model was implemented in the computational fluid dynamics code KIVA-3V to achieve high fidelity simulation of gasoline direct injection (GDI) combustion. The model allows for the identification of fuel economy improvements and emissions implications for this technology, in addition to the investigation of new operating strategies. An extensive validation of all aspects of the simulation with experimental results was performed. In the coherent flame model, the flame is represented by a transport equation for flame density, with modeled terms for the production and destruction. Stratification of the engine charge is incorporated by diagnostic equations for the unburned fuel, oxygen, and enthalpy. This allows the characterization of the gas properties on a conditionally-averaged basis, by separately averaging over the burned or unburned fraction in a computational cell. The conditionally-averaged burned gas temperature can then be used to calculate the emissions formation rates, leading to more accurate predictions. The coherent flame model was extended, to capture the effects of exhaust gas recirculation stratification on combustion and pollutant formation, by adding a new diagnostics equation for CO2 originating from exhaust gas recirculation. A near-wall flame treatment was also implemented to represent the realistic behavior of the flame near the walls more accurately. Since the literature lacks integrated studies thoroughly validating models developed specifically for GDI engines, the combustion model was integrated with updated models for spray breakup and spray wall impingement. This complete model was then used to simulate the engine cycle of an optical 4-valve GDI engine, corresponding to an experimental GDI engine. Laser-induced fluorescence (LIF) experiments, in combination with traditional engine diagnostics, allowed us to validate the simulation by comparing air motion, mixture formation, and combustion data over a range of speed and load conditions, in a realistic engine geometry, including moving valves. The thorough validation of all aspects of the model for the complete engine cycle demonstrates how the model was able to capture the important characteristics of the GDI engine, and pointed to areas that require further improvements.

The thrust toward improving vehicle fuel economy has stimulated the development of engine technologies including Variable Compression Ratio (VCR) piston designs. Pressure Reactive Piston (PRP) technology separates the piston into two pieces with a spring set located between the upper and lower pistons. The unique feature of PRP is that the upper piston reacts to the cylinder pressure during the power stroke, accommodating rapid engine load changes passively. This mechanism effectively limits the peak cylinder pressures at high loads without an additional control device, while allowing high compression ratio under low loads. The maximum compression ratio without knocking was examined, and the preload and spring constant of spring set were obtained by using a quasi-dimensional simulation program. Belleville spring was selected as the spring set of the PRP because of its compactness and ability to carry high load with small deflection. Dynamic analysis of the piston crown was performed to calculate spring deflection and instantaneous chamber volume. The baseline and PRP engine were tested on single cylinder SI engine. Dynamometer test results demonstrated that BSFC improvement of the PRP engine over the baseline ranged from 8 to 18% at part loads. Full load torque was developed with the PRP engine without knocking at a similar magnitude as the baseline. The PRP engine combustion is characterized by Reverse and Flattened Motion of piston crown near the Top Dead Center (TDC) and higher thermal efficiency. An analytic model for the geometric interaction between the spherical flame and the combustion chamber was newly developed in order to calculate the flame entrainment rate of unburned charge. The program was modified for simulating PRP motion to investigate the effect of the spring set on engine performance and emissions over various operating conditions. It was found that BSFC improvement of the PRP engine increased with engine speed and fuel conversion efficiency was gradually increased with the increase of the spring set preload until the spring set could not be compressed any more due to high preload. However, NO emissions were increased at part loads compared to the baseline due to higher compression ratio.

Ozone, the main component of photochemical smog, is formed when hydrocarbons in the atmosphere oxidize in the presence of sunlight and nitric oxide. Ozone causes health problems, damages crops and disrupts natural ecosystems. Despite remarkable progress over the past 30 years, automobiles continue to be a major source of hydrocarbon emissions and ozone pollution. Numerous studies have shown that variable valve actuation (VVA) can reduce engine fuel consumption and nitric oxide emissions. The objective of this study is to evaluate whether VVA controlled exhaust valve opening (EVO) and exhaust valve closing (EVC) can also be used to reduce hydrocarbon emissions. An automotive gasoline engine was set up and tested with different EVO and EVC timings under steady-state and start-up conditions. Crank angle based hydrocarbon measurements were taken at the exhaust valve, in the exhaust manifold and at the entrance and exit of the close-coupled catalyst. Hydrocarbon speciation was performed to gain insight into combustion and post-flame oxidation processes and to estimate local ozone production (LOP) attributable to exhaust hydrocarbons. The methodology to estimate LOP, which accounts for local atmospheric conditions, was developed specifically for this purpose. A carefully validated engine cycle simulation was utilized to identify promising EVO and EVC timings and to simulate crank angle based exhaust mass flow that was used with measured hydrocarbon concentrations to calculate hydrocarbon mass flow. The first strategy tested on the engine utilizes early EVO with standard EVC. Although this strategy increases exhaust gas temperature and reduces catalyst light-off time, the rapid drop in cylinder temperature increases cylinder-out hydrocarbons to such a degree that a net increase in hydrocarbon emissions and LOP results. The second strategy investigated on the engine utilizes early EVO with early EVC. This strategy reduces hydrocarbon emissions and LOP. Early EVO reduces catalyst light-off time by increasing exhaust gas temperature and early EVC retains the hydrocarbon-rich exhaust gas from the piston crevice in the cylinder for reburning in the next cycle. Start-up hydrocarbon emissions are reduced by 27% and start-up LOP is decreased by 25%.

The Homogeneous Charge Compression Ignition (HCCI) engine is under widespread investigation due to its potential to lower NOx and particulate emissions while maintaining high thermal efficiency. Nevertheless, HCCI operation presents several challenges as the combustion event is very unstable under certain conditions. Furthermore, the HCCI combustion mode typically suffers from reduced power density and high-unburned hydrocarbon emissions (∼15 g/kW-hr). Accurate and yet computationally expedient HCCI engine models are necessary to evaluate approaches and guide experiments for improving power density, pollutant emissions, and controllability. Therefore, the objectives of this work are twofold: (1) develop a fundamentally-based quasi-dimensional model that can be used to evaluate HCCI engine performance and emissions tradeoffs over a range of conditions and configurations; (2) experimentally characterize HCCI performance, emissions, and gas-side boundary conditions under large-bore, turbo-charged configurations. In this study, natural gas was used as the fuel, primarily because its characteristics complement the HCCI process. A full-cycle simulation was developed to model the HCCI engine process under turbo-charged operation. The quasi-dimensional model couples a core gas zone to boundary layer and ring-pack crevice regions. The thermal boundary layer model is derived from energy considerations. The ring-pack model accounts for both ring position and flow exchange within its subvolumes. Two engine platforms were utilized in this study for model validation and ultimate confirmation of HCCI operating strategies. Two-component fuel data, collected at Lund University on a Volvo TD100 engine, were used to evaluate the predictive capability of the C-4 chemical mechanism for different fuel-mixtures. In parallel, experimental data were acquired on a large-bore, Caterpillar 3500 single-cylinder engine that was operated in a turbo-charged configuration under a range of thermodynamic conditions, engine speeds and compression ratios. The study has shown that a physically based quasi-dimensional model has the ability to accurately predict both HCCI engine performance as well as unburned hydrocarbon (UHC) emissions. Experiments confirmed that the engine could be operated in a stable configuration up to equivalence ratios of 0.32. The NOx emissions were on the order of .05g/kW-hr while the combustion efficiency ranged from 85--92%. Nevertheless, the engine thermal efficiency ranged from 30--39%, well below what was expected for HCCI operation.

A series of combustion sub-models, which represent the physical mechanisms of ignition, turbulent combustion, transition, and soot formation were developed and investigated towards the goal of creating a robust and accurate computational engine design and performance evaluation tool. All sub-models were validated via either comparison with appropriate experimental measurements and/or alternative modeling results. In addition, all sub-models incorporated detailed chemistry. The sub-models were implemented in a modified version of KIVA-3V and used to predict the effects of the combustion processes on in-cylinder combustion performance of a Direct Injection Natural Gas (DING) engine. The most significant contributions of the study are the modification of Eddy Dissipation Concept (EDC) model for engine combustions the development of a chemistry and mixing model to represent the transition from ignition to turbulent combustion, and the development of soot model that includes soot transport effects. The EDC model originally proposed by Magnussen was modified for engine combustion processes and used to predict ignition delay. Using the modified EDC model, reaction rates during the ignition/turbulent combustion phases were predicted via interactions between the mixing rate (i.e. physical mixing of the fuel and oxidizer) and the chemical rate (i.e. the detailed chemical kinetics of the reaction mechanism). The results for ignition delay agreed well with experimental data, and it was found that mixing effects play an important role in predicting ignition delay, particularly at high temperatures. The transition phenomenon was well captured using the transition model. The soot model initially developed by Frenklach and co-workers was modified for use in predicting soot production in a DI engine. The model was validated by comparison with experimental data for an ethylene flame/perfectly stirred reactor system. When the soot model was implemented in engine studies, transport effects were found to play an important role in predicting soot formation in a DING engine. The soot model represents a considerable improvement in evaluating and predicting soot emissions in engine studies due to the basis on relevant physical phenomena and due to the inclusion of particle transport effects. The effects of detailed chemistry on the in-cylinder engine combustion processes were also investigated. Detailed chemistry produces similar cylinder pressures and engine power compared to modeling results using a global reaction. However, lower cylinder temperatures and changes in species profiles were obtained when detailed chemistry was used. The changes in temperature and species concentrations significantly affect the pollutant emissions and substantiate the need to include detailed chemistry. Engine performance studies were also conducted using the newly developed and validated sub-models. Retarding injection timing leads to incomplete combustion, and eventually produces low IMEP and high soot concentrations. Increasing exhaust gas recirculation (EGR) reduces NO significantly without notable changing the IMEP.

A simulation-based, systems engineering approach to the optimal design of complex, diesel-based powertrains has been addressed in this work from the perspective of maximizing vehicle fuel economy. A framework and methodology for the design of a diesel engine within a conventional and hybrid electric vehicle have been proposed, and contrasted with optima from traditional engine design and selection processes. The main contribution of this work is the development of a non-iterative technique that essentially solves the inverse of the engine design problem, in a robust and computationally inexpensive manner. That is, instead of configuring a vehicle with an engine and iterating on its design until all system objectives are met, the vehicle itself can now directly interpret the demands placed on it into the best engine design. The techniques introduced here are specific to the types of system models and optimization schemes implemented. Modifications of this methodology have been developed for other types of systems; the results from which indicate similar time savings as well as increased computational robustness.

A quasi-dimensional, multi-zone, direct-injection diesel combustion model has been developed and implemented in a full cycle simulation of a turbocharged engine. The combustion model accounts for transient fuel spray evolution, fuel-air mixing, ignition, combustion and NO and soot pollutant formation. In the model, the fuel spray is divided into a number of zones, which are treated as open systems. While mass and energy equations are solved for each zone, a simplified momentum conservation equation is used to calculate the amount of air entrained into each zone. The model is validated with experimental data obtained in a constant volume chamber and real engines. First, predictions of spray penetration and spray angle are validated against measurements in a pressurized constant volume chamber. Subsequently, predictions of heat release rate, as well as NO and soot emissions are compared with experimental data obtained from representative heavy-duty, turbocharged diesel engines. It is demonstrated that the model can predict rate of heat release with high fidelity across a wide range of operating conditions. NO and soot emissions predictions also show good agreement with experimental data. However, additional effort is required to enhance the fidelity of NO and soot emissions predictions across a wider range of operating conditions.

Models for describing ignition, combustion and pollutant formation in direct injection (DI) natural gas engines are developed in this work. These models are coupled with a multi-dimensional reactive flow code KIVA-3 to study the energy conversion process in DI engines. Emphasis is placed on gaining a fundamental understanding of each of these processes by studying them in detail individually. Ignition of natural gas injected under compression ignition conditions is simulated by using a detailed chemical kinetic mechanism for natural gas oxidation (consisting of 22 species and 104 elementary reactions) in conjunction with KIVA-3. The mechanism is chosen after a systematic study of its predictive capabilities under typical end-of-compression conditions in DI engines. It is shown that the choice of a suitable detailed kinetic mechanism is essential to account for all the thermodynamic and fuel composition related factors that affect the ignition process in DI natural gas engines. Calculations of ignition delay of different blends of natural gas in a combustion bomb compare very well with measurements. The coupled detailed kinetics-multi-dimensional flow model is also used to investigate the effect of parameters like chemical additives, injection rate and fuel temperature on ignition delay and ignition location. It is established that a specified mass of fuel burned is a more consistent criterion to detect the length of the ignition delay period than pressure rise while comparing different natural gas blends over a range of temperatures and pressures. After ignition occurs, combustion of injected fuel with air is described using a mixing controlled model. Fuel and air are assumed to burn in stoichiometric proportions at a rate proportional to the rate of mixing. Transition from detailed kinetics to the combustion model is assumed to occur when 4% of the total injected fuel has burned. Pollutant formation is modeled by implementing the extended Zeldovich kinetic mechanism for thermal nitric oxide formation in KIVA-3. Finally, calculations are done using the geometry of a medium size heavy-duty diesel engine with pure methane as fuel. It is shown that operating parameters like engine speed and load, coupled with different strategies for injection timing and level of boost, can have a significant impact on engine output. The tradeoffs between performance and emissions are discussed in terms of variation in engine operating conditions and injection strategies.

Alternative strategies for reducing hydrocarbon emissions from automotive and small utility engines were assessed using experimental and modeling studies. For the experimental study, a small utility engine test facility, and an automotive engine test facility were designed and constructed. Nineteen small utility engines, selected by the U.S. EPA, were tested at various air/fuel ratios under steady-state and transient operations. The effects of combustion chamber design, carburetor design, lean burning, and fuel composition on performance, hydrocarbon and carbon monoxide emissions were studied. A Ford 2.0L Zetec engine was also tested under different operating conditions. For the modeling study, a hydrocarbon emissions model was developed. The model accounts for unburned hydrocarbon emissions from oil film absorption and desorption, crevice flows, and post-flame and exhaust port oxidation. The correlation between engine-out hydrocarbon emissions, exhaust temperature, and engine speed was also investigated and a semi-empirical formula was derived. The model was then implemented within a thermodynamic, spark-ignition engine cycle simulation code. Model calibration and validation with experiment were carried out systematically on baseline engines. Comparisons between model predictions and experiment of small utility engines showed good agreement at high and medium loads but significant differences at idle. Nevertheless, the model predicted satisfactory the SAE J1088 A Cycle-Weighted hydrocarbon emissions. This is of practical significance since the index of cycle-weighted emissions is the sole regulatory indicator. For the automotive engines, the model yielded reasonable agreement with experiment. This is caused by the different designs between small and automotive engines. To improve model accuracy for automotive engines, some sophisticated submodels are needed. With the assistance of model prediction and experiment, alternative emissions control strategies were studied extensively. Those included lean burning, in-cylinder fuel injection, catalytic conversion, acceleration pumps, oxygenated fuels, and air injection. It was found that since small utility engines were designed to run fuel rich to obtain satisfactory cooling and since their size, weight and cost were very critical, any single emission control method could not yield satisfactory results. A compromise has to be reached between emissions, performance, durability, size, weight, and cost for a given engine design. Finally, several effective strategies were recommended.

With increasing interest in alternative fuel technology, natural gas has become an attractive fuel for reciprocating engines. In this work, a modified version of the Los Alamos KIVA3 code has been used to model gaseous injection and mixing, by establishing the necessary boundary conditions at the injector/cylinder interface. It has been shown that gaseous injection models have to be accompanied by sufficient grid refinement to capture length scales of the order of the injector diameter (Abraham, 1997). In addition, high fidelity turbulence models are needed to monitor the plume evolution away from the injector. Turbulence modeling is explored in this work by considering variants of the k-$\epsilon$ and LES models. The standard k-$\epsilon$ model (Launder and Spalding, 1972) was found to underpredict the recirculation length in the backward-facing step geometry, overpredict dissipation in the confined co-flow jet geometry and underpredict penetration histories in all three gaseous jet configurations considered. These jet configurations included downward injection of methane in a pressurized vessel (Aesoy, 1996), horizontal injection of hydrogen in a confined chamber (Tomita et al., 1997) and upward impinging transient acetylene jet in a confined box (Fujimoto et al., 1997). The implementation of the nonlinear k-$\epsilon$ model (Speziale, 1986) either provided moderate corrections in the geometries of the backward-facing step and confined co-flow jets, or produced convergence problems in engine calculations. The implementation of the RNG based k-$\epsilon$ model (Han and Reitz, 1995) provided better agreement with experiments for both the recirculating (backward-facing step) and gaseous injection cases. The implementation of the LES model was based on the formulation of Moin et al. (1991), and the evaluation of the sub-grid scale stresses proposed by Lily, (1992). It was shown that this model was significantly affected by the high numerical diffusivity of KTVA-3. Based on the better predictions for recirculating flows and closer agreement with measured penetration data, the RNG k-$\epsilon$ model was selected for engine calculations. Parametric studies of injector and chamber geometries indicated that an optimum injection strategy could lead to as much as 100% flammable mixture near top dead center.