In order to see the limit of the two-stage combustion for NOx reduction, they conducted numerical simulation of NO formation with a two zone model for multiple elements of fuels in 1995 . The result showed the effect of the two stage combustion on NOx reduction and the importance of time scales of turbulent mixing.
For further emission reduction in diffusion controlled combustion, analysis was made on general relationships between the smoke reduction index and mixing strength based on available experimental data in 1998 . This analysis was basic and provided only a first approximation of turbulent mixing phenomena, however the results indicated a general relationship of the two factors, and offered ideas for a better understanding of the phenomena occurring in combustion chambers.
To identify optimum conditions of jet impingement on the spray flames for effective soot reduction, experiments were conducted with a constant volume combustion vessel, which permitted optical observation and equipped with a jet-generating cell. Observation was made for a variety combinations of distances between spray nozzle and jet orifice and for different directions of impingement [14, 15]. Apparent soot reduction was observed with the jet impingement. However, when the jet was very close to the flame it penetrated the soot cloud and caused little mixing. There were no apparent differences in the combustion duration when the direction of impingement was varied.
2. EXPERIMENTAL APPARATUS AND SIMULATION METHOD
The basic engines used in this research were four stroke, single cylinder, naturally aspirated, direct injection diesel engines, with displacements of 694 and 1425 cc. The 694 cc engine (rated speed: 3600 rpm) was used in most of the experiments, except for the results in Fig.2, where the 1425 cc engine (rated speed: 1200 rpm) was used. A turbulence generating cell and auxiliary injection system are equipped to the base engine as shown in Fig.1. This cell is referred to as the Combustion Chamber for Disturbance (CCD), and the system and auxiliary injection to the CCD are referred to as the "CCD system" and "CCD injection" in the following. Small amounts of fuel are injected into the CCD, and the jet of this burned gas is injected into the main combustion chamber, where it generates strong turbulence. The volume of the CCD corresponds to 5.5% of the total clearance volume. The injection timing and fuel ratio of the CCD injection were 10 deg. after top dead center (ATDC) and comprised about 10% of the total injected fuel at full load.
The smoke density was measured by a Bosch Smoke Meter, and a CLD type exhaust gas analyzer was used to measure the NOx concentration. To investigate the formation and oxidation process of particulate in the cylinder, particulate was sampled from the combustion chamber during the combustion process. The apparatus consists of a needle type high speed gas sampling valve, a particulate trap filter, a gas sampling chamber, and a vacuum pump. Particulate concentrations were determined by measuring the filter weight before and after sampling and the volume of sampled gas. Details of the apparatus and procedures are given in reference .
The calculations of the NO reaction kinetics were made by modifying subroutines of the KIVA program , a 3-dimensional combustion simulation program developed at the Los Alamos National Laboratory, USA. The calculation adopts Extended Zel'dovich Reactions, in which radicals necessary for the NO calculations, OH, H, O, CO, etc., are derived from the chemical equilibrium. Octane is used as a fuel, and it burns in an one step reaction of Arrhenius form. The present simulation is zero-dimensional, and the calculations consider the changes in temperature and species in a small space (cell) in the cylinder. The cell contains fuel, air, and combustion gas, and allows fresh air to mix in the cell. The cell is subject to a cylinder pressure history from the surroundings over a range of crank angles, and the gas in the cell is compressed and expands isentropically. When combustion takes place, the cell expands at quasi-constant pressure during a short time step. The mixing rate of fresh air in the cell is given by Wiebe's function. Details of the simulation are shown in reference .
A constant volume combustion vessel was made as outlined in Fig.10 for the observations of the combustion. This vessel has a 100 mm diameter 30 mm thick test section. A small amount of fuel is injected in a turbulence generating cell (volume 2.5 cc and path diameter 3.4 mm) termed CCD. The combustion in the cell creates a jet aimed at the main flame and generating strong turbulence. Main spray nozzles are located on the cylinder wall and on the end plate, so that different settings of the main spray are possible. Initially the chamber is riffled by a pre-mixed charge of ethylene (5 vol%), nitrogen (60 vol%), and oxygen (35 vol%) at a pressure of 0.8 MPa. Then the charge is ignited by a spark, creating a hot, high pressure atmosphere for the following spray injection. At an appropriate pressure, 900 K and 3.0 MPa in the present experiments, the main spray is injected into the chamber filled with combustion gas with sufficient oxygen, and this is followed by the secondary injection from the CCD.
3. EFFECT OF TURBULENT JET ON SMOKE REDUCTION
Figure 2 shows smoke, NOx, and brake specific fuel consumption (BSFC) for different brake mean effective pressures (BMEP) with and without CCD injection. The abscissa is the engine load expressed as the excess air ratio. The engine speeds are 1200, 1000 and 800 rpm, and the injection timing of the main fuel is 10 deg. CA before top dead center (BTDC). The amount of fuel injected into the CCD is constant, and equal to 20% of the total fuel amount at a load of 0.37 MPa. The engine load is controlled by the fuel injected into the main chamber.
The figure shows that the CCD system reduces smoke very effectively. The smoke density is 0 above an excess air ratio of 2, and is below 2 Bosch Smoke Unit even at an excess air ratio of 1.2. With the CCD injection the brake specific fuel consumption becomes poorer, and total hydrocarbon (THC) increases at partial loads. This is because the ratio of the CCD fuel amount relative to the total fuel increases at partial loads as the CCD fuel quantity is fixed. At full loads the BSFC is slightly better with CCD injection, and THC is as low as without CCD injection. The decreased fuel consumption at high load was attributed to the shorter combustion clue to the enhanced mixing. The NOx emission with CCD injection is slightly lower than the standard engine, and it was approximately equal to the standard engine in most of the experiments.
This confirms that the CCD system is very effective in reducing smoke, while it maintains NOx at the level of the standard engine. Generally NOx tends to increase when smoke decreases, but the result here indicates that the system improves on the trade-off relationship between NOx and smoke.