Comparative analysis of surface morphology characteristics of abrasive belt wear in conventional and creep-feed grindings based on the Yolov8 algorithm

Creep-feed belts grinding (CFBG) can improve processing efficiency while ensuring the surface quality of TC4 titanium alloy materials. However, the difficult-to-machine characteristics of titanium alloys and the single-layer abrasive nature of the abrasive belts accelerate belt wear. Meanwhile, grinding creates micron-level scratches on the surface, which can lead to localized stress concentration on the titanium alloy surface and significantly reduce the service performance of workpiece. Due to the randomness of the grain and the complexity of grinding process, which makes it a challenging problem to investigate the regularity and mechanism of the surface morphology characteristics with belt wear under different grinding forms. Therefore, this research proposes a surface topography evaluation method of titanium alloy based on YOLOv8 segmentation algorithm, and explores the regularity of surface topography after creep-feed/conventional abrasive belt (CBG) wear. Firstly, the YOLOv8 algorithm was used to segment the collected titanium alloy surface topography images. Accuracy of the algorithm is 0.85, Precision, Recall and F1 values are 0.85, 0.72, F1 and 0.78 respectively, which indicates that the algorithm performs relatively well. Secondly, the surface morphology characterization regularity was quantitatively characterized; with belt wear, the area ratio, perimeter ratio and depth of scratches in CBG gradually decreased, and the aspect ratio, fractal dimension and radius of curvature gradually increased; while in CFBG, the parameters all changed rapidly (decreasing/increasing) and the value then fluctuated within a range. Finally, a model of titanium alloy surface morphology molding in the whole life cycle of abrasive belts under different grinding methods was established. • The Yolov8 algorithm achieved accurate segmentation of micro-scratches. • First proposed 6 parameters for comprehensive scratch characterization. • Image gray scale equivalent characterization of depth of scratches is achieved. • Attritional wear leads to a continuous tendency of morphological parameters. • Fracture wear leads to similarity (fluctuations) in the characteristic parameters.