A Method for Predicting Ultimate Bearing Capacity of Bolts Based on PSO-RBF Neural Network

Abstract

The test of bearing capacity of bolt is very important for bolt quality test. Improper parameter of radial basis function (RBF) neural network may lead to large network convergence error and worse generalization capacity. Particle swarm optimization (PSO) is used to optimize the parameters in the study of improved RBF neural network. First, eigenvector respecting the bearing capacity of bolt is chosen as input of RBF neural network. The output of RBF neural network is the ultimate bearing capacity of bolt and the number of hidden layer nodes is determined by the subtraction clustering algorithm; Second, the initial center and width of the hidden layer is determined by K-means and PSO algorithm is introduced to optimize the the hidden layer centers and widths of RBF neural network. The weight between hidden layer and output layer is obtained by the least squares method; finally, a PSO-RBF neural network model is established to predict ultimate bearing capacity of bolts. In practical project cases, PSO-RBF neural network enhances the detection precision.