Abstract: Based on the nonlinear flexural finite element model, the stability of the X-braced structure with out-of-plane support was investigated combined orthogonal test and analysis of variance (ANOVA), and nonlinear surface regression and neural network method were used for optimal design, which provided a way for the design of complex nonlinear structures. Specifically, finite element models supporting automated simulations in tabular form were constructed using ANSYS and SOLIDWORKS. The effects of geometric parameters and boundary conditions were fully considered, and factors with important influence were identified by significance orthogonal test and analysis of variance. Then, orthogonal test for optimal design was conducted followed by the structural optimal design using nonlinear regression and BP neural network. The study found that the influence of the in-plane support geometric parameters was greater than that of the out-of-plane support, where a larger diameter of the in-plane support tube led to a more stable structure. In addition, the out-of-plane support can effectively improve the structural stiffness, while the junction position will have great influence on the stability. Finally, the load ratio affected buckling and the tension increased the stability. This paper provided a data-driven optimization design approach in preliminary design.