Adjoint DNS of Turbulent Channel Flow

A discrete adjoint is developed on top of a parallel-computing code for the direct numerical simulation of wall turbulence. The architecture of the direct code is designed to obtain good scalar performance and high parallel efficiency using a cluster of dual-CPU Personal Computers. The adjoint code, written in a way that mimicks the basic structure of its direct counterpart, can also benefit of the same computational and parallel efficiency. After describing how we tested the adjoint part of the code against its direct counterpart, we present results concerning the sensitivity of the wall drag to blowing/suction applied at the wall for the case of a turbulent channel flow at Reτ = 180