Designing of Prediction Model for Parameter Optimization in CNC Machining Based on Artificial Neural Network

The consistent surface finishes in polishing remain a significant challenge. Variations in machining parameters often lead to inconsistent results, negatively impacting both the appearance and functionality of polished components. Despite advances in CNC technology, the selection of optimal machining parameters remains complex due to the interplay of multiple factors. This study addresses this gap by developing a prediction model to systematically determine appropriate machining parameters such as cutting speed (vc), feed rate (vf), and depth of cut (doc). Surface roughness (Ra) was used as the key metric to evaluate the surface quality of CNC end-mill products. The above machining parameters were varied according to a 33 -extended full factorial design, resulting in 108 experimental output targets of Ra. These outputs were then utilized to train an ANN prediction model based on a feed-forward backpropagation (FFBP) algorithm. The results demonstrated a strong correlation coefficient (R = 0.992) across all data sets. In the regression plot, the predicted values closely matched the actual values, indicating a high level of accuracy in the regression model. Furthermore, error evaluation using normalized root mean square error (NRMSE) revealed a low error rate of 3.79%, which is considered highly acceptable, particularly in the context of polishing.