This article deals with prediction of turbulent flows in ducts of nonc
ircular cross sections and, in particular, assessment of the performan
ce in such flows of two very different models of turbulence. One model
is of the two-equation, eddy-viscosity type, which is used in conjunc
tion with a non-linear stress-strain relationship. The other is a comp
lete Reynolds-stress transport closure that involves the solution of a
differential transport equation for each of the sk components of the
Reynolds-stress tenser. The flows considered are characterized by the
presence of secondary motions that are largely driven by the turbulenc
e anisotropy and whose prediction remains a severe challenge to turbul
ence closures. Data from several experiments involving such flows are
used here to assess the overall performance of the two models. It is f
ound that; the two models yield very similar results that are also of
adequate engineering accuracy - an outcome that argues in favor of the
use of the nonlinear two-equation model in practical computations.