Hello Evert, to approximate the cracked deck with elastic material, use 3D linear elastic material as a base, and add smeared reinforcement in the direction(s) with better properties.

While it is quite common to carry multiple analyses with torsional load to see the influence of some geometry or material changes on torsional stiffness or strength, I do not understand what should be the outcome of a nonlinear analysis neglecting a specific part of the complex response? Thoughts like this can be very useful with analytical solutions, but I doubt the sense of trying this with NL FEM, where the torsional stiffness is a result of the complex response, mainly influenced by shear concrete properties.

To build a model without torsion stiffness, I guess the best way would be to only use reinforcement bars (with the corresponding cross sections), disable geometrical nonlinearity, and apply the load in a single step. Of course this approach can only make sense with elastic materials.