文章信息/Info

Title:

Research Progress of Machine Learning Aided Titanium Alloys Design

文章編號(hào):

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(mèng)園

1. 西北有色金屬研究院，陜西 西安 710016 2. 西安交通大學(xué)金屬材料強(qiáng)度國(guó)家重點(diǎn)實(shí)驗(yàn)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(guān)鍵詞:

鈦合金; 機(jī)器學(xué)習(xí); 合金設(shè)計(jì); 特征工程; 數(shù)據(jù)驅(qū)動(dòng)

Keywords:

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

分類(lèi)號(hào):

TP181;TG146.23

DOI:

10.7502/j.issn.1674-3962.202501004

文獻(xiàn)標(biāo)志碼:

A

摘要:

鈦合金以其優(yōu)良的力學(xué)性能、生物相容性、耐蝕性及耐熱性等特點(diǎn)已成為高性能結(jié)構(gòu)件的首選材料，被廣泛應(yīng)用在醫(yī)療器械、化工、航天航空、艦船等領(lǐng)域。隨著鈦合金中合金化元素種類(lèi)的進(jìn)一步增加，鈦合金成分、工藝與性能間的映射機(jī)制關(guān)系愈加復(fù)雜，以鉬當(dāng)量、d�搽娮雍辖鹄碚�、價(jià)電子濃度等為代表的傳統(tǒng)鈦合金設(shè)計(jì)方法很難準(zhǔn)確捕捉到合金元素間復(fù)雜的交互作用及其對(duì)組織和性能的影響規(guī)律。近年來(lái)，機(jī)器學(xué)習(xí)技術(shù)有望從材料數(shù)據(jù)中通過(guò)算法挖掘材料成分、工藝、組織、性能之間的隱藏關(guān)系，實(shí)現(xiàn)實(shí)驗(yàn)過(guò)程優(yōu)化，突破研究人員基于經(jīng)驗(yàn)和“試錯(cuò)法”高成本、低效率的材料設(shè)計(jì)瓶頸，為鈦合金智能設(shè)計(jì)開(kāi)辟了新的思路。以機(jī)器學(xué)習(xí)輔助鈦合金設(shè)計(jì)研究的流程為主線(xiàn)，介紹了機(jī)器學(xué)習(xí)輔助鈦合金設(shè)計(jì)研發(fā)中的數(shù)據(jù)來(lái)源與預(yù)處理、特征工程、機(jī)器學(xué)習(xí)建模預(yù)測(cè)和優(yōu)化設(shè)計(jì)等技術(shù)，綜述了數(shù)據(jù)驅(qū)動(dòng)的智能化研發(fā)范式在鈦合金設(shè)計(jì)中的研究進(jìn)展。最后，分析了這一新型研發(fā)范式在鈦合金領(lǐng)域面臨的問(wèn)題并展望了其發(fā)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.