Search Results

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.

Characterization of high pressure diesel sprays has been performed both experimentally and numerically. The experimental study was conducted using a fuel injection system which has a capability of producing multiple injection sprays. The fuel sprays were injected from a multi-hole nozzle into a pressurized cylindrical chamber with optical windows. In order to investigate the effects of a multiple injection strategy on spray characteristics, a double injection spray with the mass evenly distributed between the first and second sprays, and a 1 millisecond dwell between sprays was compared with a single injection spray. Both single and double injection cases had nominally the same injection pressure, injection delivery, and ambient gas density. Transient spray tip penetration lengths and spray angles were obtained from high speed photographic spray images. The spray droplet sizes were derived from the images by using a light extinction method.

An experimental investigation of the spectral characteristics of turbulent flow in a scale model of a high pressure diesel fuel injector nozzle hole has been conducted. Instantaneous velocity measurements were made in a 50X transparent model of one hole of an injector nozzle using an Aerometrics Phase/Doppler Particle Analyzer (PDPA) in the velocity mode. Turbulence spectra were calculated from the velocity data using the Lomb-Scargle method. Injector hole length to diameter ratio (L/D) values of 1.3, 2.4, 4.9, and 7.7 and inlet radius to diameter ratio (R/D) values of approximately 0 and 0.3 were investigated. Results were obtained for a steady flow average Reynolds number of 10,500, which is analogous to a fuel injection velocity of 320 m/s and a sac pressure of approximately 67 MPa (10,000 psi). Turbulence time frequency spectra were obtained for significant locations in each geometry, in order to determine how geometry affects the development of the turbulent spectra.

An experimental investigation of turbulent flow patterns in a scale model of a high pressure diesel fuel injector nozzle has been conducted. Instantaneous velocity measurements were made in a 50X transparent model of one hole of the injector nozzle using an Aerometrics Phase Doppler Particle Analyzer (PDPA) in the velocity mode. Length to diameter ratio (L/D) values of 1.3, 2.4, 4.9, and 7.7 and inlet radius to diameter ratio (R/D) values of approximately 0 and 0.3 were investigated. Two steady flow average Reynolds numbers (10,500 and 13,300), analogous to fuel injection velocities and sac pressures of approximately 320 and 405 m/s and 67 and 107 MPa (10,000 and 16,000 psi), were investigated. The axial progression of mean and root mean square (rms) axial velocities was obtained for both sharp and rounded inlet conditions and varying L/D. The discharge coefficient was also calculated for each geometry.

A three-dimensional computer code (KIVA) is being modified to include state-of-the-art submodels for diesel engine flow and combustion: spray atomization, drop breakup/coalescence, multi-component fuel vaporization, spray/wall interaction, ignition and combustion, wall heat transfer, unburned HC and NOx formation, soot and radiation and the intake flow process. Improved and/or new submodels which have been completed are: wall heat transfer with unsteadiness and compressibility, laminar-turbulent characteristic time combustion with unburned HC and Zeldo'vich NOx, and spray/wall impingement with rebounding and sliding drops.