서브메뉴
검색
Understanding Microstructure-ultrasound Relationships in Additively Manufactured Materials: Experiments and Numerical Simulations
Understanding Microstructure-ultrasound Relationships in Additively Manufactured Materials: Experiments and Numerical Simulations
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20260202105319
- ISBN
- 9798297663886
- DDC
- 610
- 서명/저자
- Understanding Microstructure-ultrasound Relationships in Additively Manufactured Materials: Experiments and Numerical Simulations
- 발행사항
- [Sl] : The Pennsylvania State University, 2025
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2025
- 형태사항
- 206 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 87-04, Section: B.
- 주기사항
- Advisor: Lear, Matthew H.;Shokouhi, Parisa.
- 학위논문주기
- Thesis (Ph.D.)--The Pennsylvania State University, 2025.
- 초록/해제
- 요약Due to their sensitivities to microstructural evolution during processes including heat treatment and plastic deformation, ultrasonic parameters including the acoustic nonlinearity parameter (β) and resonance frequency are useful nondestructive evaluation (NDE) tools for monitoring microstructural changes in metals, ceramics, and other engineered materials relevant to the growing utilization of additive manufacturing (AM). Despite their diagnostic capabilities, the competing contributions of different microstructural features to the measured ultrasonic response render the interpretation of received ultrasonic signals a nontrivial task. The primary objective of this dissertation is to utilize experimental and modeling-based approaches to advance the capabilities of ultrasonic testing for the characterization and quality assurance of metallic and ceramic components, emphasizing AM parts. Two sample sets are investigated: 316L stainless steel (SS) plates manufactured by laser powder bed fusion, and alumina cylinders with complex internal geometries manufactured by direct material jetting.For metallic samples, two studies are described: An experimental investigation of the microstructure-ultrasound-property relationships in AM 316L stainless steel focused on Second Harmonic Generation (SHG) testing to measure β, and a numerical study where 3D Dislocation Dynamics (DD) simulations are developed to elucidate the relationship between β and specific dislocation characteristics (e.g., length, density, and orientation). In the former, ultrasonic testing, microstructural characterization, and mechanical testing are combined to create a more complete picture of microstructure-ultrasound-property relationships in AM 316L SS as they evolve with heat treatment. The results indicate that measurable variations in β from SHG testing scale with heat treatment-induced microstructure changes, such as dislocation density and the volume fraction of oxide inclusions. These observations are considered alongside the trends in tensile strength with heat treatment, where an inverse relationship between β and these strength properties is measured that coincides with an increase in the presence of detrimental oxide inclusions. The numerical modeling investigation focuses on DD simulations of many differently-oriented dislocations within one domain - an advancement from prior analytical and numerical models concerning a single dislocation. The multi-dislocation framework is tested on several parametric studies of dislocation length and density, and compared extensively to canonical analytical modeling results. The DD predictions indicate a more complex relationship between dislocation morphology, applied stress, and β than the prior models, particularly for cases involving edge dislocations. The competing effects of edge and screw dislocations on β are investigated within the same crystal for the first time, and unique behaviors are predicted that are unrepresented in existing models. The DD simulations are used to recreate SHG results in fatigued copper using published microstructural data, and to aid in the validation of a recent, contradictory experimental observation in 316L SS, where an anomalous decrease in β was measured despite increasing plastic deformation.Finally, an application-focused study is presented, in which machine learning and finite element analysis are combined to evaluate subtle geometric variations in AM ceramic cylinders using their resonant ultrasonic signatures. NDE studies are often constrained by small sample sizes - this limitation is even more prohibitive in the context of training machine learning models to interpret the data. To address this data scarcity, high-fidelity finite element simulations are employed to generate synthetic training data, which is used to train a Random Forest model for binary classification of the experimental samples. Notable efficacy of the synthetic data is observed: A model trained on 50 simulated and three experimental frequency spectra has a testing accuracy ∼20% higher than a model trained on the same three experiments alone. Further improvements in accuracy are achievable with a careful and interpretable feature selection process, informed by simulation results and feature importance rankings from the Random Forest model. Classification accuracies greater than 75% are obtained using the described transfer learning approach. While this specific study involves Resonant Ultrasound Spectroscopy and finite element analysis, the concept of using synthetic data to improve machine learning model accuracy for real samples is broadly applicable to the NDE field.
- 일반주제명
- Tomography
- 일반주제명
- Heat treating
- 일반주제명
- Spectrum analysis
- 일반주제명
- Grain size
- 일반주제명
- Deformation
- 일반주제명
- Acoustics
- 일반주제명
- Radiation
- 일반주제명
- Materials fatigue
- 일반주제명
- Analytical chemistry
- 일반주제명
- Industrial engineering
- 일반주제명
- Medical imaging
- 일반주제명
- Optics
- 일반주제명
- Thermodynamics
- 기본자료저록
- Dissertations Abstracts International. 87-04B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008260126s2025 us c eng d■001000017360193
■00520260202105319
■006m o d
■007cr#unu||||||||
■020 ▼a9798297663886
■035 ▼a(MiAaPQ)AAI32289473
■035 ▼a(MiAaPQ)PennState29628clw5756
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a610
■1001 ▼aWilliams, Colin L.
■24510▼aUnderstanding Microstructure-ultrasound Relationships in Additively Manufactured Materials: Experiments and Numerical Simulations
■260 ▼a[Sl]▼bThe Pennsylvania State University▼c2025
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2025
■300 ▼a206 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 87-04, Section: B.
■500 ▼aAdvisor: Lear, Matthew H.;Shokouhi, Parisa.
■5021 ▼aThesis (Ph.D.)--The Pennsylvania State University, 2025.
■520 ▼aDue to their sensitivities to microstructural evolution during processes including heat treatment and plastic deformation, ultrasonic parameters including the acoustic nonlinearity parameter (β) and resonance frequency are useful nondestructive evaluation (NDE) tools for monitoring microstructural changes in metals, ceramics, and other engineered materials relevant to the growing utilization of additive manufacturing (AM). Despite their diagnostic capabilities, the competing contributions of different microstructural features to the measured ultrasonic response render the interpretation of received ultrasonic signals a nontrivial task. The primary objective of this dissertation is to utilize experimental and modeling-based approaches to advance the capabilities of ultrasonic testing for the characterization and quality assurance of metallic and ceramic components, emphasizing AM parts. Two sample sets are investigated: 316L stainless steel (SS) plates manufactured by laser powder bed fusion, and alumina cylinders with complex internal geometries manufactured by direct material jetting.For metallic samples, two studies are described: An experimental investigation of the microstructure-ultrasound-property relationships in AM 316L stainless steel focused on Second Harmonic Generation (SHG) testing to measure β, and a numerical study where 3D Dislocation Dynamics (DD) simulations are developed to elucidate the relationship between β and specific dislocation characteristics (e.g., length, density, and orientation). In the former, ultrasonic testing, microstructural characterization, and mechanical testing are combined to create a more complete picture of microstructure-ultrasound-property relationships in AM 316L SS as they evolve with heat treatment. The results indicate that measurable variations in β from SHG testing scale with heat treatment-induced microstructure changes, such as dislocation density and the volume fraction of oxide inclusions. These observations are considered alongside the trends in tensile strength with heat treatment, where an inverse relationship between β and these strength properties is measured that coincides with an increase in the presence of detrimental oxide inclusions. The numerical modeling investigation focuses on DD simulations of many differently-oriented dislocations within one domain - an advancement from prior analytical and numerical models concerning a single dislocation. The multi-dislocation framework is tested on several parametric studies of dislocation length and density, and compared extensively to canonical analytical modeling results. The DD predictions indicate a more complex relationship between dislocation morphology, applied stress, and β than the prior models, particularly for cases involving edge dislocations. The competing effects of edge and screw dislocations on β are investigated within the same crystal for the first time, and unique behaviors are predicted that are unrepresented in existing models. The DD simulations are used to recreate SHG results in fatigued copper using published microstructural data, and to aid in the validation of a recent, contradictory experimental observation in 316L SS, where an anomalous decrease in β was measured despite increasing plastic deformation.Finally, an application-focused study is presented, in which machine learning and finite element analysis are combined to evaluate subtle geometric variations in AM ceramic cylinders using their resonant ultrasonic signatures. NDE studies are often constrained by small sample sizes - this limitation is even more prohibitive in the context of training machine learning models to interpret the data. To address this data scarcity, high-fidelity finite element simulations are employed to generate synthetic training data, which is used to train a Random Forest model for binary classification of the experimental samples. Notable efficacy of the synthetic data is observed: A model trained on 50 simulated and three experimental frequency spectra has a testing accuracy ∼20% higher than a model trained on the same three experiments alone. Further improvements in accuracy are achievable with a careful and interpretable feature selection process, informed by simulation results and feature importance rankings from the Random Forest model. Classification accuracies greater than 75% are obtained using the described transfer learning approach. While this specific study involves Resonant Ultrasound Spectroscopy and finite element analysis, the concept of using synthetic data to improve machine learning model accuracy for real samples is broadly applicable to the NDE field.
■590 ▼aSchool code: 0176.
■650 4▼aTomography
■650 4▼aHeat treating
■650 4▼aSpectrum analysis
■650 4▼aGrain size
■650 4▼aDeformation
■650 4▼aAcoustics
■650 4▼aRadiation
■650 4▼aMaterials fatigue
■650 4▼aAnalytical chemistry
■650 4▼aIndustrial engineering
■650 4▼aMedical imaging
■650 4▼aOptics
■650 4▼aThermodynamics
■690 ▼a0986
■690 ▼a0486
■690 ▼a0546
■690 ▼a0574
■690 ▼a0752
■690 ▼a0348
■71020▼aThe Pennsylvania State University.
■7730 ▼tDissertations Abstracts International▼g87-04B.
■790 ▼a0176
■791 ▼aPh.D.
■792 ▼a2025
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17360193▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.


