Layout optimization of box girder with RBF-NNM-APSO algorithm

The layout optimization problem of complex box girder structure is solved with a new method RBF-NNM-APSO formed with the digital neural network model (NNM) of radial basis function (RBF) and adaptive particle swarm optimization (APSO) algorithm in this paper. The optimized surrogate model is proposed and applied to the configuration optimization of heavy-duty box girder of casting crane for improving the mechanical properties of the optimized object and expediting proceedings. First, the parametric command flow finite element numerical model of box girder is established. The RBF neural network is trained by constructing a mixed orthogonal experimental table of parameters, and the relationship between the design variables and the maximum stress and deformation is established. Subsequently, the trained RBF neural network design scheme is optimized by APSO algorithm. Finally, on the premise of not increasing the total mass, a new layout form of box girder is obtained.