基于深度学习的金相智能分析可视化系统开发  

The development of an intelligent metallographic analysis and visualization system based on deep learning

作　　者：张方飞园 张跃飞[1] 张泽[1] Zhang Fangfeiyuan;Zhang Yuefei;Zhang Ze(The Institute of High Temperature Alloys of the School of Materials Science and Engineering,Zhejiang University Hangzhou 310000 Zhejiang Province)

机构地区：[1]浙江大学材料科学与工程学院高温合金研究所,杭州310000

出　　处：《冶金标准化与质量》2025年第1期17-21,共5页Metallurgical Standardization & Quality

摘　　要：本研究开发了一种基于深度学习的金相智能分析系统。传统金相分析依赖人工操作,效率低且主观性强,难以满足复杂材料系统的需求。本系统基于U-Net模型,通过实验确定了适用于铁素体相分割与晶界提取的最优模型组合,集成了图像采集、预处理、分割分析、几何测量及可视化展示功能。实验表明,该系统在分割精度和边界细节捕捉方面表现出色,可为材料微观结构分析提供有效支持,有助于材料失效机制研究等应用。This study develops an intelligent metallographic analysis system based on deep learning.Traditional metallographic analysis relies on manual operations,which are inefficient and highly subjective,making it difficult to meet the needs of complex material systems.The proposed system,built on the U-Net model,identifies the optimal model combination for ferrite phase segmentation and grain boundary extraction through experiments.It integrates image acquisition,preprocessing,segmentation analysis,geometric measurement,and visualization features.Experimental results demonstrate that the system excels in segmentation accuracy and boundary detail capture,providing effective support for the analysis of material microstructures.This approach is beneficial for applications such as the study of material failure mechanisms.

关 键 词：深度学习 金相分析 U-Net模型 晶界提取 

分 类 号：TG115.21[金属学及工艺—物理冶金]