Search Results

Two different types of mesh used for diesel combustion with the KIVA-4 code are compared. One is a well established conventional KIVA-3 type polar mesh. The other is a non-polar mesh with uniform size throughout the piston bowl so as to reduce the number of cells and to improve the quality of the cell shapes around the cylinder axis which can contain many fuel droplets that affect prediction accuracy and the computational time. This mesh is specialized for the KIVA-4 code which employs an unstructured mesh. To prevent dramatic changes in spray penetration caused by the difference in cell size between the two types of mesh, a recently developed spray model which reduces mesh dependency of the droplet behavior has been implemented. For the ignition and combustion models, the Shell model and characteristic time combustion (CTC) model are employed.

Neural networks are useful tools for optimization studies since they are very fast, so that while capturing the accuracy of multi-dimensional CFD calculations or experimental data, they can be run numerous times as required by many optimization techniques. This paper describes how a set of neural networks trained on a multi-dimensional CFD code to predict pressure, temperature, heat flux, torque and emissions, have been used by a genetic algorithm in combination with a hill-climbing type algorithm to optimize operating parameters of a diesel engine over the entire speed-torque map of the engine. The optimized parameters are mass of fuel injected per cycle, shape of the injection profile for dual split injection, start of injection, EGR level and boost pressure. These have been optimized for minimum emissions. Another set of neural networks have been trained to predict the optimized parameters, based on the speed-torque point of the engine.

Neural networks have been used for engine computations in the recent past. One reason for using neural networks is to capture the accuracy of multi-dimensional CFD calculations or experimental data while saving computational time, so that system simulations can be performed within a reasonable time frame. This paper describes three methods to improve upon neural network predictions. Improvement is demonstrated for in-cylinder pressure predictions in particular. The first method incorporates a physical combustion model within the transfer function of the neural network, so that the network predictions incorporate physical relationships as well as mathematical models to fit the data. The second method shows how partitioning the data into different regimes based on different physical processes, and training different networks for different regimes, improves the accuracy of predictions.

The present study investigated HCCI combustion in a heavy-duty diesel engine both experimentally and numerically. The engine was equipped with a hollow-cone pressure-swirl injector using gasoline direct injection. Characteristics of HCCI combustion were obtained by very early injection with a heated intake charge. Experimental results showed an increase in NOx emission and a decrease in UHC as the injection timing was retarded. It was also found that optimization can be achieved by controlling the intake temperature together with the start-of-injection timing. The experiments were modeled by using an engine CFD code with detailed chemistry. The CHEMKIN code was implemented into KIVA-3V such that the chemistry and flow solutions were coupled. The model predicted ignition timing, cylinder pressure, and heat release rates reasonably well. The NOx emissions were found to increase as the injection timing was retarded, in agreement with experimental results.

A turbulent combustion model based on the coherent flamelet model was developed in this study and applied to diesel engines. The combustion was modeled in three distinct but overlapping phases: low temperature ignition kinetics using the Shell ignition model, high temperature premixed burn using a single step Arrhenius equation, and the flamelet based diffusion burn. Two criteria for transitions based on temperature, heat release rate, and the local Damköhler number were developed for the progression of combustion between each of these phases. The model was implemented into the computational computer code KIVA-II. Previous experiments on a Caterpillar model E 300, # 1Y0540 engine, a Tacom LABECO research engine, and a single cylinder version of a Cummins N14 production engine were used to validate the cylinder averaged predictions of the model.

Multidimensional computations were carried out to simulate the in-cylinder fuel/air mixing process of a direct-injection spark-ignition engine using a modified version of the KIVA-3 code. A hollow cone spray was modeled using a Lagrangian stochastic approach with an empirical initial atomization treatment which is based on experimental data. Improved Spalding-type evaporation and drag models were used to calculate drop vaporization and drop dynamic drag. Spray/wall impingement hydrodynamics was accounted for by using a phenomenological model. Intake flows were computed using a simple approach in which a prescribed velocity profile is specified at the two intake valve openings. This allowed three intake flow patterns, namely, swirl, tumble and non-tumble, to be considered. It was shown that fuel vaporization was completed at the end of compression stroke with early injection timing under the chosen engine operating conditions.

A procedure has been developed to build system level predictive models that incorporate physical laws as well as information derived from experimental data. In particular a soot model was developed, trained and tested using experimental data. It was seen that the model could fit available experimental data given sufficient training time. Future accuracy on data points not encountered during training was estimated and seen to be good. The approach relies on the physical phenomena predicted by an existing system level phenomenological soot model coupled with ‘weights’ which use experimental data to adjust the predicted physical sub-model parameters to fit the data. This approach has developed from attempts at incorporating physical phenomena into neural networks for predicting emissions. Model training uses neural network training concepts.