Drilling Axial Force Prediction Based on PSO-RBF Neural Network

This paper focus on different parameters of casing drill on axial force in the process of Carbon Fiber Reinforced Plastic (CFRP) hole making and establish a reliable axial force prediction model, the influence law of each variable on axial force was analyzed based on response surface method with different parameters as design factors, and the Response Surface Methodology (RSM) empirical prediction model of axial force was constructed. Aiming at the problem of large error of RSM prediction model, the accuracy of axial force prediction model could be improved by optimizing Radial Basis Function Neural Network (RBF-NN) through Particle Swarm Optimization (PSO) algorithm. The experimental results show that the influence order of the main factors on the axial force is feed > spindle speed > aperture. The average relative error of the PSO-RBF Neural Network (PSO-RBF-NN) prediction model is 3.27%, which is 17.60% lower than the standard RBF-NN prediction error, and 44.67% lower than the RSM empirical model prediction error.