Process optimization of high-speed dry milling UD-CF/PEEK laminates using GA-BP neural network

High-performance carbon fiber-reinforced polyetheretherketone (CF/PEEK) is widely used in aerospace and premium-end medical fields due to its high strength-weight ratio, shock resistance, and reusability. However, its dry machining requirement is a significant limit to improving machining efficiency and machining quality using a traditional process. Addressing this issue, the high-speed dry (HSD) machining technique is imported in this paper. Multi-level mixed orthogonal experiments of dry milling unidirectional (UD) CF/PEEK laminates with the fiber orientation of 0° and 90° are designed. Aiming at quantitative characterizing surface quality, three-dimensional (3D) surface roughness Sq and 3D fractal dimension Ds are used to present surface roughness and surface defects, respectively. A characterization system for surface defects generated in milling CRF/PEEK is proposed. A prediction model of surface quality considering fiber orientation, cutting speed, feed per tooth, and cutting width is then established using the genetic algorithm optimized BP (GA-BP) neural network. The prediction results show that the model is of acceptable generalization capability with a prediction accuracy of over 90.39%. Based on the analysis of surface qualities and cutting temperatures, the HSD machining technique is verified to be feasible in milling UD-CF/PEEK, and the recommended cutting speed in the HSD milling boundary is 1500–1600 m/min. The 3D fractal dimension is verified feasible to evaluate the size of complex surface defects of the machined UD-CF/PEEK. It has a non-rigid negative correlation with Sq in general. Besides, cutting speed and fiber orientation are the key factors affecting the machined surface microstructural characteristics. The present study gives technical references for improving surface quality in HSD milling CF/PEEK.