A new method for predicting workpiece fatigue life based on segmentation features of titanium alloy surface morphology

Wear of abrasive belts directly changes the depth and morphological characteristics of abrasive scratches on titanium surfaces. These abrasive scratches on titanium surfaces are stress-concentrated areas directly influencing the fatigue life of the workpiece. For the purpose of accurate fatigue life prediction with different abrasive belt wear, a new method of abrasive belt wear-surface morphology-fatigue life is proposed. The Mask RCNN network is used to segment the grinding scratch, which results in segmentation accuracy of up to 0.9 with a loss value of 0.5. The depth of grinding scratches is equivalently characterized using the gray-scale information of the image, and calculating the evolution of scratches depth with belt wear (decreasing from 21 to 1 µm). Meanwhile, the regularity statistics of the length and width of the grinding scratches with the abrasive belt wear were also carried out. In addition, a prediction model for grinding scratch depth, morphological characteristics (length and width) and fatigue life is established. A high prediction accuracy of the model is demonstrated by the fact that the model can predict up to 34.08% before the final stage of abrasive belt wear.