机器学习在功能梯度材料设计-制备中的应用综述

A Review:Applications of Machine Learning in Design-Fabrication of Functionally Graded Materials

功能梯度材料作为一类新型复合材料在航空航天、生物医疗等多个先进领域中存在巨大的应用价值,但其在设计与制备的过程中所存在的多维复杂问题制约着其快速发展。大数据驱动的人工智能技术的快速崛起促使传统研发模式向数字化研发转型,数字化研发技术可应对功能梯度材料设计与制备中的复杂问题、不确定性问题,并可大幅提升产品质量、生产效率和降低成本,快速推动着功能梯度材料在多个先进领域中的发展。本文聚焦近年来机器学习技术应用于功能梯度材料领域中的研究现状,并总结了融合不同优化方法解决在功能梯度材料设计与制备中的复杂问题,包含对不同维度变化的材料组分信息的准确反演;对功能梯度材料零件的固有属性、微观结构、材料特征、服役性能进行预测与评价;根据单一或多目标确定与优化材料组分信息以及制备工艺过程参数;以及基于数据驱动方法建立带解释的智能数据库,为获取具有更优良性能的功能梯度材料提供新的设计思路。最后,本文总结了机器学习技术在功能梯度材料设计与制备领域中所存在的主要挑战与发展机遇。

Functionally graded materials as a novel type of composite material have great application value in many advanced fields such as aerospace,biomedicine,etc.However,the high-dimensional and complex problems existing in the process of design and fabrication restrict their rapid development.The rapid rise of big data-driven artificial intelligence technologies is driving the transformation of traditional research and development(R&D)models to the digital R&D.Digital R&D technologies can address the complexities and uncertainties in the design and fabrication of functionally graded materials,and it can significantly improve product quality,production efficiency and reduce costs,and it is rapidly driving the development of functionally graded materials in a number of advanced fields.This paper focused on recent research in the field of functionally graded materials using machine learning techniques,and summarized the integration of different optimization methods to solve complex problems in the design and preparation of functionally graded materials.It contains the accurate inversion of material components information with different dimensional variations;the prediction and evaluation of natural properties,microstructure,material characteristics and service performance of functionally graded material parts;the determination and optimization of material component information and preparation process parameters based on single or multiple objectives;establishment of intelligent databases with explanation based on the data-driven method,which provide a new design idea for obtaining functionally graded materials with superior properties.Moreover,the main challenges and opportunities of machine learning techniques in the field of functionally graded materials were summarized in this paper.