文章信息/Info

Title:

Research Progress of Machine Learning Aided Titanium Alloys Design

文章編號:

1674-3962(2025)04-0319-11

作者:

張 閆; 薛德禎; 辛社偉; 王 曉; 周 偉; 潘 曦; 李星吾; 張冰潔; 郝夢園

1. 西北有色金屬研究院，陜西 西安 710016 2. 西安交通大學金屬材料強度國家重點實驗室，陜西 西安 710049

Author(s):

ZHANG Yan; XUE Dezhen; XIN Shewei; WANG Xiao; ZHOU Wei; PAN Xi;  LI Xingwu; ZHANG Bingjie; HAO Mengyuan

1. Northwest Institute for Nonferrous Metal Research，Xi’an 710016，China 2.State Key Laboratory for Mechanical behavior of Materials，Xi’an Jiaotong University，Xi’an 710049，China

關鍵詞:

鈦合金; 機器學習; 合金設計; 特征工程; 數據驅動

Keywords:

titanium alloys;  machine learning;  alloy design;  feature engineering;  data driven

分類號:

TP181;TG146.23

DOI:

10.7502/j.issn.1674-3962.202501004

文獻標志碼:

A

摘要:

鈦合金以其優良的力學性能、生物相容性、耐蝕性及耐熱性等特點已成為高性能結構件的首選材料，被廣泛應用在醫療器械、化工、航天航空、艦船等領域。隨著鈦合金中合金化元素種類的進一步增加，鈦合金成分、工藝與性能間的映射機制關系愈加復雜，以鉬當量、d�搽娮雍辖鹄碚摗�價電子濃度等為代表的傳統鈦合金設計方法很難準確捕捉到合金元素間復雜的交互作用及其對組織和性能的影響規律。近年來，機器學習技術有望從材料數據中通過算法挖掘材料成分、工藝、組織、性能之間的隱藏關系，實現實驗過程優化，突破研究人員基于經驗和“試錯法”高成本、低效率的材料設計瓶頸，為鈦合金智能設計開辟了新的思路。以機器學習輔助鈦合金設計研究的流程為主線，介紹了機器學習輔助鈦合金設計研發中的數據來源與預處理、特征工程、機器學習建模預測和優化設計等技術，綜述了數據驅動的智能化研發范式在鈦合金設計中的研究進展。最后，分析了這一新型研發范式在鈦合金領域面臨的問題并展望了其發展前景。

Abstract:

Titanium alloys, known for their excellent mechanical properties,biocompatibility, corrosion resistance, and heat resistance, have become the material of choice for high-performance structural components. They are widely used in fields such as medical devices, chemical engineering, aerospace, and naval ships. As the variety of alloying elements in titanium alloys continues to increase, the mapping relationship between composition, processing, and performance becomes increasingly complex. Traditional design methods for titanium alloys, such as molybdenum equivalence, d-electron alloy theory, and valence electron concentration, struggle to accurately capture the complex interactions between alloying elements and their impact on the microstructure and performance. In recent years, machine learning technologies have shown promise in uncovering hidden relationships between material composition, processing, microstructure, and performance by mining material data through algorithms. This offers the potential to optimize experimental processes and overcome the high cost and inefficiency of trial-and-error methods in material design, opening new avenues for intelligent design of titanium alloys. This paper presents an overview of the machine learning-assisted design process for titanium alloys, including data sourcing and preprocessing, feature engineering, machine learning modeling and prediction, and optimization design. It reviews the research progress of data�瞕riven intelligent design paradigms in titanium alloy development. Finally, the paper analyzes the challenges faced by this new research paradigm in the titanium alloy field and discusses its future prospects.