摘要
增材制造(Additive manufacturing,AM)技术的迅速发展为复杂结构部件的构建与功能梯度材料的实现提供了新的可能性。但是其工艺过程涉及复杂的多物理场耦合与动态演化机制,易引发孔隙、翘曲和裂纹等微观缺陷,进而对成形质量与结构性能构成严峻挑战。机器学习(Machine learning,ML)技术凭借数据驱动优势,逐步应用于增材制造的设计优化、工艺参数调整、质量预测与缺陷检测等领域。本文综述了增材制造主要工艺及其控制特性,分析了传统物理建模在复杂耦合过程下的局限性,并系统探讨了监督学习、无监督学习、深度学习与强化学习等算法在3D打印工艺参数优化、过程异常检测、内部质量评判等关键任务中的应用,最后展望了智能化AM系统在实时感知、自适应控制与反馈优化、多模态融合与物理引导建模等方向的发展潜力。
The rapid advancement of additive manufacturing(AM)technologies has opened new frontiers for fabricating complex structural components and enabling functionally graded materials.However,the intrinsic nature of AM processes-characterized by strongly coupled multi-physical interactions and dynamic evolution-renders them highly susceptible to the formation of microstructural defects such as porosity,warping,and cracking,thereby posing substantial challenges to build quality and structural integrity.In response,machine learning(ML),with its data-driven modeling capabilities,has emerged as a powerful tool for addressing these complexities.It is increasingly being applied to design optimization,process parameter tuning,quality prediction,and defect detection in AM workflows.This review provides a comprehensive overview of the principal AM techniques and their process control characteristics,critically evaluates the limitations of conventional physics-based modeling in handling complex coupled phenomena,and systematically explores the integration of supervised learning,unsupervised learning,deep learning,and reinforcement learning in key tasks such as process optimization,anomaly detection,and internal quality evaluation in 3D printing.Finally,we outline future directions for intelligent AM systems,highlighting opportunities in real-time sensing,adaptive control,feedback-driven optimization,multimodal data fusion,and physics-informed learning frameworks.
作者
王鑫旭
闫承琳
李晓旭
王琦
崔浦
WANG Xinxu;YAN Chenglin;LI Xiaoxu;WANG Qi;CUI Pu(Research Institute of Wood Industry,Chinese Academy of Forestry,Beijing 100091,China)
出处
《材料导报》
北大核心
2025年第S2期564-576,共13页
Materials Reports
基金
国家重点研发计划(2024YFD2200700)。
关键词
机器学习
人工智能
3D打印
原位监测
工艺优化
machine learning
artificial intelligence
3D printing
in-situ monitoring
process optimization
