Nickel-based superalloys are indispensable for high-temperature engineering applications, yet their additive manufacturing (AM) is plagued by significant cracking defects. This review investigates crack failure mechanisms in ... | Find, read and cite all the research you need on Tech Science Press
28 April 2025 in Journal of Thermal Spray Technology
To overcome the bottleneck in automated, high-throughput image analysis for materials characterization, this study establishes a novel quantitative microstructural analysis paradigm that integrates expert validation with convolutional neural networks (CNNs). This approach substantially reduces manual annotation costs, achieving 94.49% segmentation accuracy for γ′ precipitates in Ni-based superalloys with a sparse training corpus of only 20 samples. Furthermore, by coupling this deep learning framework with the multicomponent diffusion-multiple (MCDM) technique, isothermal section phase diagrams of the Ni–Co–Mo(W) system were systematically constructed across multiple temperatures. The findings reveal that Co exerts a non-linear thermodynamic modulation on the precipitation of detrimental topologically close-packed (TCP) phases, with the zero-phase-fraction (ZPF) boundary exhibiting a minimum at approximately 17 wt.% Co. Concurrently, elevated temperatures significantly diminish TCP phase stability. Increasing the Co content substantially broadens the continuous γ single-phase domain, thereby defining a precise compositional window for complete supersolvus heat treatments at 1180°C. Ultimately, the methodology established herein provides a versatile pathway for constructing comprehensive "composition–processing–microstructure–property" databases, offering significant potential to accelerate the research and development cycle of advanced multiphase alloys.