서브메뉴
검색
Modeling Wind Loading on Low-Rise Buildings in Urban Environments: Leveraging Large-Eddy Simulations and Deep Neural Networks to Quantify Interference Effects
Modeling Wind Loading on Low-Rise Buildings in Urban Environments: Leveraging Large-Eddy Simulations and Deep Neural Networks to Quantify Interference Effects
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20260202104739
- ISBN
- 9798290649573
- DDC
- 551.63
- 서명/저자
- Modeling Wind Loading on Low-Rise Buildings in Urban Environments: Leveraging Large-Eddy Simulations and Deep Neural Networks to Quantify Interference Effects
- 발행사항
- [Sl] : Stanford University, 2024
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2024
- 형태사항
- 120 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 87-03, Section: A.
- 주기사항
- Advisor: Gorle, Catherine.
- 학위논문주기
- Thesis (Ph.D.)--Stanford University, 2024.
- 초록/해제
- 요약In the United States, over 90% of housing units are low- and mid-rise buildings, making them highly susceptible to wind-induced interference effects. These effects pose critical challenges to structural safety and urban sustainability, as evidenced by at least seven billion-dollar wind-induced disasters recorded in 2024. Investigating these effects in site-specific contexts, where environmental and architectural factors interact, is essential to mitigate risks.Traditionally, wind tunnel experiments have been the primary method for evaluating wind effects on buildings. While effective, these experiments have notable limitations, including high costs, limited spatial resolution, and challenges in maintaining Reynolds similarity. These constraints hinder the understanding of wind-induced interference effects, particularly in urban areas with complex building configurations.To address these limitations, this study leverages Computational Fluid Dynamics (CFD), specifically Large-Eddy Simulations (LES), as a powerful alternative for evaluating wind loads on low-rise buildings. LES provides high-resolution predictions of flow and pressure fields, enabling the identification of wind-prone areas and improving safety assessments. This study's first objective is to establish trust in LES as a reliable design tool for wind engineering. An LES-based framework is proposed to predict pressure loads on isolated low-rise buildings, validated against two wind tunnel datasets. The framework is then applied to a realistic urban area, with further validation against experimental data.Despite their potential, CFD simulations are computationally intensive, limiting their routine application. To overcome this, data-driven approaches using deep neural networks (DNNs) are explored. DNNs, trained on large CFD datasets, enable fast and accurate predictions of wind patterns and pressure loads. The second objective of this study is to make fast predictions of the flowfield and the pressure loads on buildings in urban areas. To do so, a DNN is utilized that can predict the flowfield in urban areas by taking as inputs 2D planes of the city layout at different heights. Additionally, for wind loading predictions, another DNN is used in combination with a multi-fidelity framework to predict the pressure loads on buildings at an affordable cost. This DNN architecture extracts features from the flowfield using autoencoders.
- 일반주제명
- Skewness
- 일반주제명
- Kurtosis
- 일반주제명
- Urban areas
- 일반주제명
- Buildings
- 일반주제명
- Visualization
- 일반주제명
- Neural networks
- 일반주제명
- Urban planning
- 일반주제명
- Sustainability
- 기타저자
- Stanford University.
- 기본자료저록
- Dissertations Abstracts International. 87-03A.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008260126s2024 us c eng d■001000017358697
■00520260202104739
■006m o d
■007cr#unu||||||||
■020 ▼a9798290649573
■035 ▼a(MiAaPQ)AAI32149683
■035 ▼a(MiAaPQ)Stanfordmh272yc9965
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a551.63
■1001 ▼aVargiemezis, Themistoklis.
■24510▼aModeling Wind Loading on Low-Rise Buildings in Urban Environments: Leveraging Large-Eddy Simulations and Deep Neural Networks to Quantify Interference Effects
■260 ▼a[Sl]▼bStanford University▼c2024
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2024
■300 ▼a120 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 87-03, Section: A.
■500 ▼aAdvisor: Gorle, Catherine.
■5021 ▼aThesis (Ph.D.)--Stanford University, 2024.
■520 ▼aIn the United States, over 90% of housing units are low- and mid-rise buildings, making them highly susceptible to wind-induced interference effects. These effects pose critical challenges to structural safety and urban sustainability, as evidenced by at least seven billion-dollar wind-induced disasters recorded in 2024. Investigating these effects in site-specific contexts, where environmental and architectural factors interact, is essential to mitigate risks.Traditionally, wind tunnel experiments have been the primary method for evaluating wind effects on buildings. While effective, these experiments have notable limitations, including high costs, limited spatial resolution, and challenges in maintaining Reynolds similarity. These constraints hinder the understanding of wind-induced interference effects, particularly in urban areas with complex building configurations.To address these limitations, this study leverages Computational Fluid Dynamics (CFD), specifically Large-Eddy Simulations (LES), as a powerful alternative for evaluating wind loads on low-rise buildings. LES provides high-resolution predictions of flow and pressure fields, enabling the identification of wind-prone areas and improving safety assessments. This study's first objective is to establish trust in LES as a reliable design tool for wind engineering. An LES-based framework is proposed to predict pressure loads on isolated low-rise buildings, validated against two wind tunnel datasets. The framework is then applied to a realistic urban area, with further validation against experimental data.Despite their potential, CFD simulations are computationally intensive, limiting their routine application. To overcome this, data-driven approaches using deep neural networks (DNNs) are explored. DNNs, trained on large CFD datasets, enable fast and accurate predictions of wind patterns and pressure loads. The second objective of this study is to make fast predictions of the flowfield and the pressure loads on buildings in urban areas. To do so, a DNN is utilized that can predict the flowfield in urban areas by taking as inputs 2D planes of the city layout at different heights. Additionally, for wind loading predictions, another DNN is used in combination with a multi-fidelity framework to predict the pressure loads on buildings at an affordable cost. This DNN architecture extracts features from the flowfield using autoencoders.
■590 ▼aSchool code: 0212.
■650 4▼aSkewness
■650 4▼aKurtosis
■650 4▼aUrban areas
■650 4▼aAtmospheric boundary layer
■650 4▼aBuildings
■650 4▼aVisualization
■650 4▼aNeural networks
■650 4▼aUrban planning
■650 4▼aSustainability
■653 ▼aComputational Fluid Dynamics
■653 ▼aUrban sustainability
■690 ▼a0640
■690 ▼a0999
■71020▼aStanford University.
■7730 ▼tDissertations Abstracts International▼g87-03A.
■790 ▼a0212
■791 ▼aPh.D.
■792 ▼a2024
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17358697▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.
