In order to solve the problem of insufficient rigidity of the column of the disc milling column of efficient and powerful compound milling machine, the cutting force was obtained by cutting titanium alloy, the dynamic and static characteristics of the column were calculated by ABQUS finite element analysis module, and FEM is used to calculate the static and dynamic characteristics of the key structure of the column. Unit structure and frame structure are collected and optimized by variational analysis method. Several improved design schemes of the column structure are presented for the purpose of optimizing the structure’s natural frequency. BP neural network model is also put forward to find out the optimal design variable. Compared with the original one, the natural frequency of the optimized column structure increases observably. Finally, this method is applied to analyze the structure of principle prototype and the original column is optimized according to the computed results. Through machining experiments, the static and dynamic characteristics of the machine tool are improved greatly, and the correctness and feasibility of this theory are verified.