Reply To: Thermal Poiseuille 2D.
by ADE Lattice it would be better to use defineRho() instead of defineRhoU(), because the velocity values are taken automatically from the NSE Lattice by the code itself.
By definition of the initial velocities you could try to set 0 at the start. At the inlet the necessary velocity should be reached gradually by using the SinusStartScale functor. Such initialization should be better for convergency. If not, then write us again and we will look for an other solution.