SIFT-NRBO-VMD-Transformer: A Vision-Based Data-Driven Interface Morphology Prediction Framework for Intelligent Wear Diagnosis of Wet Friction Components

Wet friction components are critical to power transmission in petroleum drilling machinery, where their reliability directly affects system stability. Surface defects, such as scratches and plowing grooves, can significantly degrade transmission performance, highlighting the importance of interface morphology prediction for intelligent wear diagnosis. In this study, interface morphology data under different conditions are acquired using a UMT-Tribolab test platform and a white light interferometer. The Scale-Invariant Feature Transform (SIFT) algorithm is employed to achieve precise localization of microscopic regions before and after testing. Based on this, an NRBO-VMD-Transformer model is developed to predict the interface morphology of wet friction components under varying conditions. The results demonstrate that SIFT enables accurate localization of microscopic regions, while the proposed model achieves high-precision prediction of interface morphology evolution. These findings provide a reliable basis for interface morphology prediction and wear evolution analysis of wet friction components.