Meeting the stringent efficiency demands of next generation
direct injection engines requires not only optimization of the
injection system and combustion chamber, but also an optimal
in-cylinder swirling charge flow. This charge motion is largely
determined by the shape of the intake port arm geometry and the
valve position.
In this paper, we outline an extensible methodology implemented
in OPENFOAMĀ® for multi-objective geometry optimization based on the
continuous adjoint. The adjoint method has a large advantage over
traditional optimization approaches in that its cost is not
dependent upon the number of parameters being optimized. This
characteristic can be used to treat every cell in the computational
domain as a tunable parameter - effectively switching cells
"on" or "off" depending on whether this action
will help improve the objectives. Unlike CAD-based parameter
optimization, the adjoint approach starts from a supplied design
space and then systematically removes all elements
counter-productive to the design objectives. The final design is
then the fluid volume left over after all the counter-productive
elements have been blocked.
The adjoint system is implemented as an adjunct to a
compressible steady state flow solver with the ability to maximize
the swirl in a target volume while minimizing the pressure loss of
the system. The tool is used to optimize the shape of the intake
port arms of a combustion chamber in a static flow test
configuration. A range of results were produced at different
weightings of the pressure loss and swirl objectives and the
ability to generate a trade-off curve between the objectives is
demonstrated. At the high end, an increase in swirl of up to 250%
was observed for modest increases in pressure loss, unequivocally
proving the effectiveness of the new methodology.