Frequency Prediction and Optimization of Truss Structure Based on BP Neural Network

Truss structure is a very classic structure form and is widely used. Frequency characteristic is one of the most important characteristics of truss structure. Generally, the finite element method can well predict the modal information of truss structure, which has little calculation and convenient modeling. However, considering the optimization problem, a few parameters can produce millions or even billions of combinations. It will take unbearable time to use the finite element method alone. Therefore, this paper introduces BP neural network for faster frequency prediction and optimization. Firstly, a four node 4-bars truss structure is established, and its characteristics are characterized as the combinations of five parameters. Secondly, the frequency characteristics of 3125 truss structures with different parameter combinations are calculated by finite element method. Thirdly, BP neural network is trained to obtain the corresponding relationship between the parameter combination and frequency characteristics of truss structure. Finally, the trained BP neural network is used to optimize the truss parameters. This method can obtain the parameter combination with the highest frequency under limited conditions, which indicates that it is an effective method.