Abstract: At present the strut and tie models of D-regions in concrete girder bridges are often obtained qualitatively according to stress trace and integration, by which the location of members is random to some extent. To solve this problem, a multi-scale evolutionary optimization algorithm is proposed, taking post-tensioned anchorage zone as research objects. First, the equivalent ground structure is built by a multi-node pattern of micro-truss element. Then the algorithm is used to analyze the ground structure subjected to different eccentric loads in order to generate the strut and tie model of anchorage zone. Finally, the models are compared with those in AASHTO bridge design specifications. The research shows that the algorithm can identify definitely the position of struts and ties of models, and the distribution of ties are found to be in good agreement with finite element analysis.