Abstract: When drilling carbon fiber composites (CFRP), the delamination damage caused by the bit axial force is one of the most important factors affecting the assembly quality and service life of parts. In order to improve the quality of CFRP hole making, this paper analyzes the influence of spindle speed, feed rate and edge diameter on the axial force of casing drill through orthogonal experiment, and establishes the regression prediction model of axial force based on response surface method (RSM). Aiming at the problem of model accuracy, a radial basis neural network (RBF) prediction model optimized by particle swarm optimization (PSO) algorithm was proposed and verified by experiments. The results show that the main factors influence the axial force in the order of feed rate > spindle speed > aperture. The combination of high speed, low feed rate and large blade diameter can obtain smaller axial force. The average relative error of the PSO-RBF neural network prediction model is 3.27%, which is 32.85% and 44.67% lower than that of the standard RBF neural network prediction model and RSM regression prediction model, respectively. Therefore, the PSO-RBF neural network model can predict the axial force in the process of casing drilling more effectively.

Key words: carbon fiber reinforced plastic, thrust force, electroplated diamond core drill, response surface method(RSM), PSO-RBF neural network