A Prediction Model for Thin-walled Duct Structure based with RBF Neural Network

Abstract

Component structure design is based on the finite element analysis technology, and the structure optimization is obtained by the iterative process of the boundary. However, this method is time-consuming and expensive to deal with the numerical simulation or complex engineering problems. In this paper, we combined the technology of experimental design and RBF neural network model technology, established a high accuracy analysis and prediction model of the structural shape by using a small number of sample points. We instituted the intrinsic part model and found the structural parameters within the constraints in the surrogate model. Consequently, we predicted the optimize structure of the part. Firstly, this paper described the design idea and theoretical basis of the method, and then illustrated the application process of the method with the example of the structure optimization of thin walled duct. The experimental results showed that this method can improve the computational efficiency and the design quality of the structure.