The optimization criterion of evolutionary topology optimization algorithms is one of the key factors affecting the structural optimization results. In this paper, some deep beam models under different load and boundary conditions are taken as a numerical examples, and comparing the difference between the optimization solution and the computational efficiency of the three algorithms based on different optimization criteria. The results show that the evolutionary topology optimization algorithms based on one-way optimization criteria and deterministic optimization criteria can efficiently obtain the optimal topology for components with simple conditions such as load and boundary, and the evolutionary topology optimization algorithms based on probabilistic optimization criterion or bidirectional optimization criterion have a stronger scope of application, and it also shows a strong ability to avoid optimized distortion and conduct global optimization on components with complicated conditions such as load and boundary. At the end of this paper, a flow chart is established for the genetic bidirectional evolutionary structural optimization algorithm combining probabilistic optimization criterion and bidirectional optimization criterion. The preliminary discussion is carried out to further improve the practicability and optimization ability of evolutionary structural optimization algorithm.