An Adjoint Approach, Via Automatic Differentiation, to Aerodynamic Design Optimization

Speaker: 
Christopher Johnson
Date of Seminar: 
Fri, 2007-03-02 12:00
Semester & Year: 
Spring 2007
Seminar Location: 
Hudson Hall Room 216
Seminar Contact(s): 
Elizabeth Irish eri@duke.edu or Justin Jaworski jwj@duke.edu
Special Instructions: 
Lunch and beverages will be served
Engineering design optimization often involves the minimization of a cost function that is a nonlinear function of a set of design variables. Sensitivity derivatives (gradients) of the cost function with respect to the design variables are computed to determine the desired direction towards which to improve the design. The adjoint method is an efficient technique that can be used to compute the sensitivities of design variables in large computational fluid dynamics codes. However, application of the adjoint method involves a code transformation that is cumbersome and time costly to hand code. An automatic differentiation (AD) code will be presented which creates discrete adjoint codes with minimal human effort. An overview of the adjoint method and the abilities of this AD program will be presented; as well as an application of these tools towards the improvement of an airfoil design.