The failure load looks pretty good but the FE model does not deflect as much as the real beam.  Over-stiff behaviour is to be expected from displacement based finite elements if the models are not sufficiently refined.  This may not be the answer but I would ask whether you have performed a convergence study on your FE model?

Thanks Angus. What do you mean by convergence study? If you mean the convergence the results after each analysis, I would say yes.

Hi Hamed,  I mean running the problem with increasing numbers of elements to represent your beam.  You should then see convergence of your quantities of interest and should be able to find a mesh that gives you sufficient accuracy.

Hi...I try the different mesh sizes already Angus.