SCAVENGING MEASUREMENTS IN A DIRECT-INJECTION TWO-STROKE ENGINE
Date
2009-07-08Author
Foundray, Hickory Zachariah
Advisor(s)
Ghandhi, Jaal B.
Metadata
Show full item recordAbstract
The scavenging process in a direct-injection two-stroke research engine was examined.
A physical model was developed to characterize scavenging based on known quantities.
An electromagnetic valve was used to sample pre-combustion gases from the running
engine. The valve proved to be very robust and was able to sample greater than 30mass
under all test conditions. The concentration of either oxygen or carbon dioxide, the mass
ow rate of the fuel, and the intake and exhaust temperatures were used to calculate the
scavenging ef ciency, trapping ef ciency, trapped mass, and concentration of combustion
products. A modi ed de nition of delivery ratio, based on a reference trapped cylinder
mass, was used for all scavenging calculations.
Good agreement was found between scavenging parameters calculated using the trapped
oxygen concentration and those calculated using the trapped carbon dioxide concentration.
Carbon monoxide was found to be a poor indicator of scavenging, because CO production
continued after combustion. Oxides of nitrogen were theoretically good indicators of scavenging,
but experimental problems with sampling were not resolved, and the NOx data were
not considered to be reliable.
Tests were performed with a variable delivery ratio and: a constant air/fuel ratio, a
constant fueling rate, or with fueling controlled by the engine control unit. Scavenging
and trapping ef ciency were shown to be strongly a function of delivery ratio. Scavenging
ef ciency increased with increasing delivery ratio, and trapping ef ciency decreased. Tests
were also run at constant delivery ratio with a variable air/fuel ratio or a variable engine
speed. Scavenging ef ciency decreased slightly with increasing air/fuel ratio, and with
decreasing speed. Trapping ef ciency did not seem to be affected by either parameter.
Scavenging and trapping ef ciency were near the non-isothermal perfect mixing model for
all test cases.