서브메뉴
검색
Machine Learning in Material Property Prediction
Machine Learning in Material Property Prediction
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20250211152922
- ISBN
- 9798383701102
- DDC
- 004
- 저자명
- Schultz, Lane E.
- 서명/저자
- Machine Learning in Material Property Prediction
- 발행사항
- [Sl] : The University of Wisconsin - Madison, 2024
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2024
- 형태사항
- 231 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 86-02, Section: B.
- 주기사항
- Advisor: Morgan, Dane.
- 학위논문주기
- Thesis (Ph.D.)--The University of Wisconsin - Madison, 2024.
- 초록/해제
- 요약This thesis explores the application of machine learning techniques to predict properties of metallic glasses and assesses the applicability domain of machine learning models. The work is primarily based on three papers focusing on predicting glass forming ability through experimental data, molecular dynamics simulations, and other computational methods. A fourth paper proposes a general approach for determining the applicability domain of machine learning models.Chapter 2 introduces computational and machine learning approaches to predict the glass forming ability of metallic alloys. The first study investigates the use of characteristic temperatures to predict critical casting diameters using various machine learning models. The second study employs molecular dynamics to calculate characteristic temperatures for metal alloys, which were then used as features in glass forming ability models. The third study explores the prediction of critical cooling rates using a combination of elemental properties and simulated features. Simulated features utilize machine-learned interatomic potentials. This work demonstrates the potential of combining machine learning with physics-based simulations for accelerating the discovery of metallic glasses.Chapter 3 focuses on developing a general approach for assessing the applicability domain of machine learning models. The proposed method uses kernel density estimation to measure training data density at feature values for inference points. Points from low-density regions have large dissimilarities (i.e., the features come from differing spaces), while those from high-density regions have small dissimilarities. High dissimilarity measures are associated with poor model performance and unreliable uncertainty estimation.In summary, this thesis makes two significant contributions: first, it enhances the understanding and creation of metallic glass forming ability models through computational techniques; second, it addresses the challenge of determining the applicability domains of models. The findings highlight the potential of machine learning in guiding the discovery of new metallic glasses while emphasizing the need for careful consideration of model limitations.
- 일반주제명
- Computer science
- 일반주제명
- Materials science
- 키워드
- Domain
- 키워드
- Machine learning
- 키워드
- Metallic glasses
- 키워드
- Metal alloys
- 기타저자
- The University of Wisconsin - Madison Materials Science and Engineering
- 기본자료저록
- Dissertations Abstracts International. 86-02B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008250123s2024 us c eng d■001000017164200
■00520250211152922
■006m o d
■007cr#unu||||||||
■020 ▼a9798383701102
■035 ▼a(MiAaPQ)AAI31558299
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a004
■1001 ▼aSchultz, Lane E.
■24510▼aMachine Learning in Material Property Prediction
■260 ▼a[Sl]▼bThe University of Wisconsin - Madison▼c2024
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2024
■300 ▼a231 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 86-02, Section: B.
■500 ▼aAdvisor: Morgan, Dane.
■5021 ▼aThesis (Ph.D.)--The University of Wisconsin - Madison, 2024.
■520 ▼aThis thesis explores the application of machine learning techniques to predict properties of metallic glasses and assesses the applicability domain of machine learning models. The work is primarily based on three papers focusing on predicting glass forming ability through experimental data, molecular dynamics simulations, and other computational methods. A fourth paper proposes a general approach for determining the applicability domain of machine learning models.Chapter 2 introduces computational and machine learning approaches to predict the glass forming ability of metallic alloys. The first study investigates the use of characteristic temperatures to predict critical casting diameters using various machine learning models. The second study employs molecular dynamics to calculate characteristic temperatures for metal alloys, which were then used as features in glass forming ability models. The third study explores the prediction of critical cooling rates using a combination of elemental properties and simulated features. Simulated features utilize machine-learned interatomic potentials. This work demonstrates the potential of combining machine learning with physics-based simulations for accelerating the discovery of metallic glasses.Chapter 3 focuses on developing a general approach for assessing the applicability domain of machine learning models. The proposed method uses kernel density estimation to measure training data density at feature values for inference points. Points from low-density regions have large dissimilarities (i.e., the features come from differing spaces), while those from high-density regions have small dissimilarities. High dissimilarity measures are associated with poor model performance and unreliable uncertainty estimation.In summary, this thesis makes two significant contributions: first, it enhances the understanding and creation of metallic glass forming ability models through computational techniques; second, it addresses the challenge of determining the applicability domains of models. The findings highlight the potential of machine learning in guiding the discovery of new metallic glasses while emphasizing the need for careful consideration of model limitations.
■590 ▼aSchool code: 0262.
■650 4▼aComputer science
■650 4▼aMaterials science
■653 ▼aDomain
■653 ▼aMachine learning
■653 ▼aMetallic glasses
■653 ▼aMolecular dynamics
■653 ▼aMetal alloys
■690 ▼a0794
■690 ▼a0984
■690 ▼a0800
■71020▼aThe University of Wisconsin - Madison▼bMaterials Science and Engineering.
■7730 ▼tDissertations Abstracts International▼g86-02B.
■790 ▼a0262
■791 ▼aPh.D.
■792 ▼a2024
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17164200▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.


