The role of surface heat flux in the evaluation of Nusselt number
 

Hello,
I’m considering a smooth duct with rectangular cross section with aspect ratio 1.5 and uniform heat flux. I should determine, by numerical tests, a Nu correlation as a function of the Re number in fully developed turbulent flow.
I decided to study the Reynolds range [1e4, 1e6]. The convergence analysis has been carried at Re=1e6. I verified that y+ near wall was below 1.

Computational domain:
Rectangular cross section a=0.09 m b=0.06 m

Boundary conditions: T_inlet=300 K, heat flux=1000 W/(m2*K), Imposed mass flowrate, outlet pressure P=0 Pa, fully developed interface,
Topolgy: periodic.

Model: SST k-omega.

The problem is as follows.
For Reynolds numbers higher than 8e5 the Nusselt number starts to decrease. For low Reynolds numbers the results are validated by Dittus-Boelter. Can the question be related to the low heat flux imposed?
When I increase the imposed heat flux, the computed Nusselt number changes and this shouldn’t happen.

Thank you in advance

Ps: For Re=1e6 the difference between the mean surface temp and the bulk temp is only 0.571 K.