Vision-based quantification of WC loss during Ni-WC MMC overlay deposition across various GMAW transfer modes

Ni–WC metal matrix composite (MMC) overlays deposited by gas metal arc welding (GMAW) are widely used for wear-resistant hardfacing, but conventional measurements of retained-WC cannot distinguish physical particle loss during deposition from WC dissolution within the molten pool. This study develops a vision-based material-balance workflow for quantifying physical WC loss and partitioning the feed-wire WC fraction into retained, physically lost, and dissolved fractions. High-speed videos covering cold metal transfer (CMT), short-circuiting GMAW (GMAW-S), pulsed GMAW(GMAW-P), and globular GMAW (GMAW-G) transfer modes were analyzed using DeepLabV3+ based semantic segmentation, particle extraction, and trajectory tracking. A probabilistic correction based on kernel density estimation and molten-pool geometry was developed to account for ejected particles that subsequently re-enter the molten pool. The corrected image-based particle-loss estimates were then combined with retained-WC measurements through material-balance analysis, whereby the fraction of WC dissolution is evaluated. The results show that physical WC loss and WC dissolution exhibit different process dependencies. The dissolved WC fraction generally increased with average instantaneous power (AIP) in the lower-power regime, whereas the physically lost WC fraction showed no clear monotonic relationship with AIP and was more strongly associated with transfer-mode characteristics. CMT exhibited the highest physical-loss fraction but the lowest dissolved fraction, while GMAW-S showed the lowest physical-loss fraction and one of the highest retained-WC fractions. The proposed framework enables quantitative identification and distinction of competing mechanisms responsible for WC non-retention and provides a practical tool for optimizing Ni–WC GMAW hardfacing processes.