A High-Accuracy Robotic Enveloping Grinding Method of Sharp-Edged Parts Based on Mechanism and Small Sample Data-Driven Model

Sharp-edged parts such as leading and trailing edges (LTEs) of blades and blisks have difficulties in guaranteeing the profile accuracy during grinding, where “overgrinding” often occurs. To address this issue, the robotic enveloping grinding technology is developed in this paper, which can relieve the stress concentration between the abrasive belt and the workpiece. Furthermore, to address the “modeling residuals” problem of simplified mechanism methods and the poor generalization of data-driven methods for small sample data, a novel mechanism-data hybrid-driven method is proposed. Combining large amounts of the simulated data generated by the mechanism model with small amounts of the experimental data, a mechanism-data adaptive domain adversarial neural network (ADANN) is established to realize the accurate prediction of the material removal depth (MRD) under various grinding parameter combinations. In order to realize the stable and consistent MRD, the grinding force is planned using the Bayesian theorem combined with the proposed prediction model. The experiment results show that the proposed prediction method has higher prediction accuracy compared with the mechanism model and the ADANN-2R model. Furthermore, grinding experiments on LTEs of blades show that based on the proposed grinding parameters planning method, the stable and consistent material removal is achieved compared with the existing method.