Intelligent recognition method for material removal mode during high-quality ground surface of RB-SiC ceramics based on YOLOv8-Slim-Neck-Ca model

Reaction-bonded silicon carbide (RB-SiC) is a typical brittle material. Surface removal modes such as brittle fracture and ductile groove will directly influence the performance of RB-SiC. This study proposes an improved YOLOv8 intelligent recognition method to enhance the accuracy and efficiency of recognising material removal mode on the surface of RB-SiC. The model employs the lightweight YOLOv8n architecture with a Slim-neck structure to reduce network parameters and accelerate detection speed, integrates the Coordinate Attention (CA) module for enhanced feature extraction, and utilises the Wise-IoU loss function to improve loss calculation. The experimental results showed that the original YOLOv8 model achieved a mean Average Precision (mAP) of 84.7% and the proposed model achieved an mAP of 88.6%, outperforming the original by 3.9%. Meanwhile, the mapping relationship between the material removal mode and the grinding parameters on the surface of RB-SiC ceramics was established. Based on the material removal mechanism, advanced approaches for evaluating the quality of the grinding surface were explored. A new method for detecting material removal modes on RB-SiC surfaces using improved YOLOv8 is proposed.The slim-neck structure and coordinate attention module optimise the model for lightweight, efficient, robust performance.The model was validated by comparison and ablation tests employing the dataset obtained from RB-SiC precision grinding.Mapping the relationship between grinding parameters and the material removal modes of RB-SiC surface reveals strategies for reducing brittle fracture damage. A new method for detecting material removal modes on RB-SiC surfaces using improved YOLOv8 is proposed. The slim-neck structure and coordinate attention module optimise the model for lightweight, efficient, robust performance. The model was validated by comparison and ablation tests employing the dataset obtained from RB-SiC precision grinding. Mapping the relationship between grinding parameters and the material removal modes of RB-SiC surface reveals strategies for reducing brittle fracture damage.