I was thinking to make a material model for beams and especially for the fibered sections, which is initially dependent on the strains, and then - to simulate the localized deformation behavior (crack mouth opening) - use the total deformation of the fiber.

I can somehow access the length of the element and convert yield strains for each element.

For example, the first yield point will be a strain of 0.01, and the second will deformation of 0.1mm, so I will divide 0.1mm/Length of the beam to obtain the second yield strain for each beam and save it in material status.

but I wonder if this would be the right approach to do this?

...