발표와 학술 활동

본 연구는 국부적인 횡방향 벽면 속도가 존재하는 3차원 난류 경계층 유동에 대한 RANS 난류 모델의 예측 성능을 비교 평가하는 것을 목적으로 한다. 이를 위해 유동 영역의 입구 조건으로 2차원 영압력구배 난류 경계층(2D ZPG TBL)을 설정하고, k-ω SST, BSL EARSM, SSG RSM 등 총 3종의 RANS 난류 모델을 적용하여 수치해석을 수행하였다. 도출된 해석 결과는 고해상도 LES 데이터와의 비교를 통해 검증하였으며, 주요 유동 변수인 형상계수, 횡방향 벽면 전단응력, 압력 및 난류운동에너지 유동장을 바탕으로 각 모델의 예측 정확도를 평가하였다. 이를 통해 난류 모델이 정교해짐에 따른 예측 성능의 향상 정도를 파악하고, 복잡한 횡방향 전단 유동 영역에서 각 모델의 예측 한계와 오차 발생 경향을 확인하였다. 이를 통해 3차원 난류 경계층 유동에 대한 기존 RANS 모델들의 예측 특성을 정량적으로 파악하고, 복잡한 전단 흐름의 정확한 모사를 위한 난류 모델의 적용 방향을 제시하고자 한다.

벤투리관 내 공동현상은 국부 압력 저하로 인해 공동의 형성과 붕괴가 반복되는 비정상 다상 유동 거동을 특징으로 한다. 실제 유체시스템에서는 펌프, 밸브 및 배관 진동 등으로 인해 유동에 시간적 섭동이 발생할 수 있으며, 이는 공동현상의 발달과 압력 맥동 특성에 영향을 미칠 수 있다. 본 연구는 상류 압력 섭동이 벤투리관 내 공동현상의 거동 및 압력 맥동 특성에 미치는 영향을 수치해석적으로 분석하는 것을 목적으로 한다. 이를 위해 기존 연구의 벤투리 유로 형상과 경계조건을 재현하고, URANS 기반 다상 유동 해석을 수행하였다. 공동현상 모델은 Rayleigh-Plesset 계열의 Schnerr-Sauer 모델을 사용하였다. 먼저 섭동이 없는 기준 조건에서 공동현상의 주기적 거동을 분석한 후, 입구 압력에 시간적 섭동을 부여하여 공동 구조, 붕괴 위치 및 압력 맥동 변화를 비교하였다. 해석 결과는 시간에 따른 압력 신호, 증기 체적분율의 공간-시간 선도 및 FFT 분석을 통해 평가하였으며, 이를 바탕으로 공동의 주기적 거동과 압력 맥동의 주요 주파수 특성을 도출하였다. 이를 통해 입구 압력 섭동 조건에서 벤투리관 내 공동현상의 동적 응답 특성을 정량적으로 제시하고자 한다.

수소 연료전지 자동차는 고압 수소 저장 시스템과 스택 탑재 등으로 인해 차체 중량이 증가하여 섀시 및 현가장치에 작용하는 동적 하중 부담이 크다. 이에 따라 오리피스를 통과하는 작동유의 유동 저항을 정밀하게 제어하는 댐퍼 감쇠력 특성의 최적 설계가 핵심 과제로 대두되고 있다. 본 연구에서는 실제 댐퍼 제조 현장의 기술 수요를 반영하여, 선택식 오리피스 구조를 대상으로 홀 형상 및 유량 변화에 대한 매개변수 연구를 효율적으로 수행하기 위한 3차원 격자 자동 생성 절차와 Agentic AI 기반의 자동화 해석 파이프라인을 구축하였다. 형상 모델링과 격자 생성, 격자 품질 검증은 물론 해석 케이스 설정, 솔버 구동 및 후처리 전 과정을 Agentic AI가 자율적으로 연계 수행하도록 설계하였다. 유동 해석은 OpenFOAM을 이용해 3차원 정상 층류 조건에서 진행되었으며, 해석 조건으로는 홀 직경 3종과 유량 5종의 조합 조건 및 테이퍼(taper) 홀의 정·역방향 유동 조건을 적용하여 압력강하와 국부 저압 특성을 분석하였다. 해석 결과, 압력강하는 유량이 증가함에 따라 비선형적으로 증가하였고 홀 직경이 작을수록 급격히 상승하였다. 특히 테이퍼 홀의 경우 고유량 영역에서 유동 방향에 따른 총 압력강하의 차이는 미미하였으나, 국부 최소 압력 분포에서는 뚜렷한 차이를 보여 향후 캐비테이션 억제 설계를 위해 압력강하와 국부 저압을 독립적으로 평가해야 함을 규명하였다. 본 연구에서 확립한 자동화 프레임워크는 다양한 오리피스 형상의 유동 특성을 신속하게 도출하는 데 유용하며, 향후 캐비테이션 다상 유동 해석 및 댐퍼 밸브계 힘 평형 모델과 연계한 감쇠력 정밀 예측으로 확장할 예정이다.

친환경 자동차는 내연기관 차량에 비해 중량이 무거워 현가장치에 가해지는 하중 부담이 크다. 자동차용 댐퍼 오리피스 부근에서는 국부적인 저압 형성으로 인해 캐비테이션이 발생할 수 있으며, 이로 인해 생성된 기포는 소음과 진동을 유발하고 댐퍼의 감쇠 특성에 부정적인 영향을 미칠 수 있다. 본 연구에서는 내경 1 mm 오리피스를 대상으로 오일-증기 2상 유동 해석 모델을 구축하고, 하류 압력 조건에 따른 증기 발생량과 압력강하 특성을 단상 유동 해석 결과와 비교 분석하였다. 2상 유동 해석은 OpenFOAM의 상변화 솔버인 interPhaseChangeFoam과 Schnerr-Sauer 캐비테이션 모델을 적용하여 수행하였다. 작동유로는 40°C의 Mobil DTE 25 유압유 물성을 적용하였다. 유입 유량을 고정한 상태에서 하류 압력을 두 가지 조건으로 설정하였다. 본 연구(발표)에서는 하류 압력 조건별 증기 체적률, 압력 분포 및 시간에 따른 압력강하 거동을 단상 해석 결과와 비교하여 캐비테이션이 오리피스 유동에 미치는 영향을 고찰하였다. 본 연구 결과는 차량용 댐퍼의 오리피스 설계 단계에서 캐비테이션 발생 위험을 예측하고 성능을 검토하기 위한 기초자료로 활용될 수 있을 것으로 기대된다.

공기용으로 설계된 원심압축기에 암모니아를 작동유체로 적용하면 기체상수와 비열비, 점성의 차이로 인해 압축 성능과 내부 유동이 달라질 수 있다. 본 연구에서는 형상과 실험 자료가 공개된 NASA CC3 원심압축기를 대상으로, 형상을 바꾸지 않고 작동유체를 공기에서 암모니아로 바꿀 때의 공력 성능과 내부 유동 특성 변화를 수치해석으로 평가하고자 한다. 해석은 Ansys CFX를 이용한 3차원 정상 압축성 RANS 해석으로 수행하며, 난류모델은 SST k-ω 모델을 사용한다. 암모니아의 회전수는 공기와 주속 마하수가 같아지도록 정하고, 두 작동유체의 운전점은 유량계수를 기준으로 대응시킨다. 해석의 신뢰성은 공기 조건의 성능곡선을 기존 실험 자료와 비교하여 확인한다. 두 작동유체에 대해 압력비와 효율, 일계수를 비교하고, 블레이드 사이 유동장에서 충격파와 박리, 팁 누설 구조의 차이를 분석한다. 이를 통해 공기용 압축기를 암모니아 압축에 활용할 때 필요한 상사 조건과 그 한계를 검토한다. 본 연구의 결과는 암모니아용 원심압축기의 형상 개선 방향을 설정하는 기초자료로 활용될 수 있을 것으로 기대된다.

공동현상 유동의 수치해석에서 상변화 모형과 모형 매개변수의 설정은 공동의 발생과 성장뿐만 아니라 압력 및 유량의 예측 결과에도 영향을 미칠 수 있다. 본 연구에서는 벤투리관 내부의 물-증기 공동유동을 대상으로 공동현상 모형의 매개변수 변화가 RANS 해석 결과에 미치는 영향을 정량적으로 분석하고자 한다. 정상상태 RANS 해석에는 k-ω 난류모델과 Rayleigh-Plesset 방정식 기반의 Schnerr-Sauer 캐비테이션 모델을 적용하며, 핵생성 및 상변화와 관련된 주요 모형 매개변수인 핵생성 수밀도, 핵 직경, 증발 계수 및 응축 계수를 변화시켜 민감도 분석을 수행한다. 각 조건에서 공동 길이, 증기 체적분율, 압력강하 및 국부 정압을 주요 평가 지표로 선정하고, 매개변수 변화에 따른 각 지표의 응답 특성을 비교한다. 또한 일민감도 분석을 통해 주요 영향 인자를 선별하고, 이후 Latin Hypercube Sampling을 이용하여 파라미터 간 상호작용과 출력 변수의 변동 범위를 분석한다. 이를 통해 공동현상 모형의 매개변수가 벤투리관 공동유동의 RANS 예측에 미치는 영향을 파악하고, 단일 매개변수 조건을 사용한 해석 결과의 민감성과 불확실성을 평가하기 위한 기초자료를 제시하고자 한다.

액체수소 극저온 밸브 열응력 해석 시 관행적으로 적용되는 원관 상관식 기반 상수 대류열전달계수는 '크기 오차'와 '공간 분포 무시 오차'를 동시에 내포한다. 본 연구는 6 인치 글로브 밸브를 대상으로 3 차원 RANS 유동해석과 열-구조 연성해석을 수행하고, CFD 국부 분포·그와 면적가중 평균이 같은 균일 조건·상관식 균일 조건의 세 가지를 동일한 고체 격자에서 비교하여 두 오차 기여를 분리·정량화하였다. 해석 결과 등가응력 400 MPa 초과 체적비는 계수의 크기 오차만으로 22.3 %가 줄어든 반면, 공간 분포 오차는 반대 방향으로 7.5 %에 그쳐 크기 오차가 지배적이었다. 반면 외부 자연대류 경계조건에서는 외부 열저항이 병목이 되어 두 오차 모두 0.2 % 이내로 소멸하였다. 따라서 밸브 열응력 해석 시 공간 분포 재현보다는 형상에 부합하는 대표 열전달계수 크기 산출이 우선되며, 이는 젖은면 결합이 병목인 조건에서 유효하다.

구름 공동현상은 공동의 형성과 붕괴가 주기적을 발생하는 비정상 유동 거동을 특징으로 한다. 이러한 과도 응답의 해석은 주기적 유동이 갖는 비선형적 특성을 이해하는데 중요한 의의를 가진다. 본 연구는 벤투리에서 발생하는 비정상 구름 공동현상의 시간 전개 특성을 정량적으로 제시하는 것을 목적으로 한다. 기존 연구[1]의 유로 형상과 경계조건을 재현하고, 난류모델로 SST k-ω를 적용한 URANS기반 다상 해석을 수행하였다. 공동현상의 수치모델은 Rayleigh-Plesset 계열의 Schnerr-Sauer모델을 사용하였으며, 표면장력 항을 모델에 포함하였다. 해석 결과의 검증은 벤투리 중심선의 속도와 정압 분포 및 선도를 문헌 실험치와 대조하여 정합성을 확인하였다. 초기 과도 구간이 경과한 후, 비정상 해석에서 형성되는 주기적 유동 구간을 대상으로, 시간 변화에 따른 유동장과 공간-시간 선도를 통해 구름 공동현상의 거동을 한 주기 내에서 기술하였다. 제시한 시간 분해 분석은 주기적 유동 구간의 변화 양상을 통합적으로 드러내며, 시간 변화에 따른 공동현상 특성을 명확히 보여주었다.

터보기계 내부에서는 정지벽과 움직이는 벽이 국소적으로 공존하여, 벽 근처 유동이 횡방향 벽면운동을 받는 3차원 난류경계층의 성격을 갖는다. 난류전단응력 벡터가 평균변형률 벡터를 지연하여 따라가므로[1] 선형 와점성모형은 구조적 한계를 가지며 레이놀즈응력모형이 대안이 되지만, 고전적 모형의 상당수는 응력 텐서를 직접 풀면서도 난류 시간척도를 소산율 ε로 닫는다. 점성저층을 해상하는 y+ < 1 격자에서는 벽에서 k → 0이 되어 ε/k 항이 수치적으로 경직해진다. 본 연구는 예측 오차가 와점성 가정뿐 아니라 근벽 시간척도 처리에도 좌우됨을 보이고자 한다. 압축성 축대칭 형상이었던 선행 연구[2]와 달리, 비압축성 평판 위에 횡방향 이동벽(48 ≤ x/δin ≤ 68, Ww/U∞ = 1)을 둔 3차원 난류경계층을 대상으로 한다. 기준해는 디지털 필터[3] 기반 합성난류를 입구에 부여한 직접수치모사이며, 이동벽과 정지벽의 차분으로 벽면 운동의 기여를 분리한다. 난류모델 비교는 OpenFOAM으로 수행한다. 표준 k–ε, k–ω, SST와 LRR, Launder–Gibson 모형에 더해 근벽 시간척도를 β*ω로 대체한 ω 기반 모형과 ω/ε 혼합 모형을 동일 격자에서 비교하고, 평균속도·레이놀즈응력·벽전단응력의 이동벽 증분을 직접수치모사와 대조하여 후속 모델링 방향을 제시한다.

실린더, 토출 밸브와 배출 유로를 포함하는 용적형 수소 압축기를 LBM으로 모사하고 난류·소음 특성을 분석합니다.

동일한 총 압력 강하에서 글로브 밸브의 개도율을 바꾸어 유량, 박리와 난류 운동에너지를 비교합니다.

RANS와 LBM의 수소 엘보 유동 예측을 DNS와 비교하고 곡률 반경에 따른 유동 특성을 조사합니다.

벤트 스택의 기체 수소 유동, 강재의 전도 열전달과 외부 수소 농도 분포를 결합해 분석합니다.

공개 압축기 형상을 바탕으로 실린더·토출 밸브·엘보를 모델링하고 LBM 유동·소음 해석용 복셀을 생성합니다.

NASA Rotor 67의 상류에 경계층에 따른 불균일 동압 분포를 적용하고 블레이드 하중과 유동 특성을 균일 조건과 비교합니다.

CMP 장치의 복잡한 패드·헤드 형상과 회전 운동을 반영하는 비정상 VOF 해석 프로세스를 개발하고, 격자 크기에 따른 슬러리 분포와 압력을 비교합니다.

엘보와 와류기의 상대 위치에 따른 내부 유동과 유동 저항을 RANS와 SPH로 조사합니다.

GIS 지형 정보와 VOF·DPM 기법으로 경사지의 빗물 유동을 모델링하고 강우량 및 지면 흡수율의 영향을 비교합니다.

SST 기반 URANS와 DDES를 비교하여 글로브 밸브의 유동 박리, 유량과 압력 강하 예측 및 계산 비용의 차이를 조사합니다.

액화 수소 유동과 밸브 고체의 전도 열전달을 결합하는 해석 프로세스를 구축하고 격자 크기에 따른 결과를 비교합니다.

수위와 유속에 따른 인체 모형 주변의 수면과 난류 유동을 다상유동 RANS로 해석합니다.

DNS를 이용한 난류 구조 분석에서 터보기계의 하이브리드 RANS/LES와 공력소음 해석까지 CFD의 활용 경험을 소개합니다.

가스터빈 엔진 터빈 블레이드의 파단 메커니즘을 재료 분석, 열응력·진동 모드 해석과 위험도 평가를 통해 조사한 연구입니다.

Accurate prediction of thermal field in high pressure turbines is a critical aspect of aerodynamic and durability design. This is particularly true when the flow at turbine inlet exhibits large gradients in temperature, both radially and circumferentially. In other words, in the presence of hot streaks from the combustor. In the numerical study presented in this paper, coupled highfidelity eddy-resolving simulations of a combustor and a turbine are used to study the differences in the temperature profile at the exit of the first vane and the heat flux on the first blade, resulting from different positioning, or clocking, between the combustor fuel nozzles and turbine vanes. The resolved unsteadiness and turbulence from the combustor impacts mixing and secondary flow in the high pressure turbine. Temperature profiles from both actual combustor CFD simulations, as well as and modulated profiles with more pronounced variation, or pattern factor, are used at the turbine inlet. A threshold of the pattern factor that brings the benefit of clocking is identified. Clocking positioning between the combustor and vanes was studied for the most benefit.

In this paper, wall-resolved LES computations for a compressor cascade from Ecole Centrale de Lyon [1] are presented. A computational grid containing about 600 million computational cells was used in these simulations. This grid resolves the details of tripping strips used in the experiments, located near the leading edge of the blade on both suction and pressure sides. Endwall turbulent boundary layer at cascade inlet was measured to be at a momentum thickness based Reynolds number of about 7000 to 8000, with quite a bit of variation in the pitchwise direction. In order to avoid the cost of simulating the entire duct upstream of the cascade, and any auxiliary flat plate boundary layer simulations, the inlet fluctuations for LES computations were generated using digital filtering method for synthetic turbulence generation [27]. Turbulence statistics from a database of high fidelity eddy simulations of flat plate boundary layers (at similar Reynolds numbers) from KTH Royal Institute of Technology in Stockholm [28] were used to fully define the properties of the cascade inlet boundary layer. In this paper, time-averaged results from three LES computations for this configuration are presented – one with no inlet fluctuations at the cascade endwall at the domain inlet, and then two computations with inlet fluctuations and boundary layers at Re  of 7000 and 8183. These provide a sensitivity of LES predictions of corner separation in the cascade to the boundary layer thickness at cascade inlet. A comparison of these simulations with prior DDES (and RANS) simulations from UTRC [26], as well as existing LES results from Ecole Centrale de Lyon [12], allows to further the understanding of critical elements of the endwall flow physics. More specifically, it provides more insight into 1 1 Corresponding author medicg@utrc.utc.com which phenomena need to be sufficiently resolved (e.g. horseshoe vortex) in order to capture both the average behavior of the corner separation, as well as its unsteady dynamics. In addition, it provides new information which will help define best practice guidelines for the use of eddy simulations to resolve endwall features in compressors at off-design conditions.

Development of an axisymmetric turbulent boundary layer (ATBL) over a stationaryrotating-stationary cylinder was investigated using large-eddy simulation (LES) for a turbomachinery-relevant configuration. In this case, the rotating speed was set to twice the axial free-stream velocity, aΩ = 2U∞ , where a is the cylinder radius, and the inlet Reθ to 4060. The rotating cylinder creates significant turbulence production through a Taylor-Couette-type flow regime and its coupling with the boundary layer. The nearwall transverse flow propagates to the outer region and contributes to a transverse mean flow. As a result, the upstream two-dimensional (2D) boundary layer turns into a threedimensional (3D) ATBL. The LES results show that the strong smaller-scale turbulence near the rotating wall propagates toward the outer region of the downstream stationary region. Finally, flow simulation of the same configuration using the standard k-ω model was also conducted for comparison.

Transitional boundary layers feature long coherent motions of streamwise velocity fluctuation, u0 , in both the laminar and turbulent regions. In the former, Klebanoff streaks amplify and become seats for breakdown to turbulence. In the fully turbulent region, large-scale motions contribute appreciably to the turbulence energy and shear stresses. Direct numerical simulation (DNS) of boundary-layer bypass transition over a flat plate with a leading edge is performed. Instantaneous realizations of spatially and temporally resolved fields are stored in a database. Structure identification techniques are used to identify these coherent flow structures [Lee, Sung & Zaki, J. Fluid Mech. 819, 165-187 (2017)]. The inception rate, lifetime and amplification rate of Klebanoff streaks are evaluated in the laminar region, and conditional averaging is used to examine the early stages of streak formation. Structure identification and tracking is also used to study the inception of large-scale coherent motion in the nascent turbulent spots and fully turbulent boundary layer downstream.

It has been widely investigated that there is a substantial intermittency between high and low drag states in wallbounded shear flows. Recent experimental and computational studies in a turbulent channel flow have identified low-drag time intervals based on wall shear stress measurements. These intervals are a weak turbulence state characterized by low-speed streaks and weak streamwise vortices. In this study, the spatiotemporaldynamics of low-drag intervals in a turbulent boundary layer is investigated using experiments and simulations. The low-drag intervals are monitored based on the wall shear stress measurement. We show that near the wall conditionally-sampled mean velocity profiles during low-drag intervals closely approach that of a low-drag nonlinear traveling wave solution as well as that of the so-called maximum drag reduction asymptote. This observation is consistent with the channel flow studies. Interestingly, the large spatial stretching of the streak is very evident in the wall-normal direction during low-drag intervals. Lastly, a possible connection between the mean velocity profile during the low-drag intervals and the Blasius profile will be discussed.

The turbulent/non-turbulent (T/NT) interface in a transitional boundary layer is examined using direct numerical simulation (DNS) flow field. The interface is identified using the local vorticity, and its statistical properties are reported. Similarities between the turbulence spots and fully turbulent region are evaluated using conditional sampling. In addition, the modulation of the interface by largescale motion (LSM) is investigated with the aid of a streak detection algorithm.

The vorticity response to small-amplitude surface undulations in a viscoelastic Couette flow is examined. An integral measure is defined to quantify the depth of vorticity penetration into the bulk flow. Three regimes of penetration are identified, and each occupies a sector of a phase diagram parameterized by the ratios of the channel depth and of the critical-layer (CL) height to the lengthscale of the surface roughness. The CL emerges due to the propagation of vorticity waves along the tensioned mean-flow streamlines. The most intriguing behavior, the trans-critical regime, occurs when the CL lies within one wavelength from the wall. In trans-critical flows a kinematic mechanism associated with the relative realignment of the polymer by the mean shear amplifies the vorticity in ‘sheets’ around the CL. The effect is dominant at this location since the vorticity waves appear steady to an observer advecting with the mean flow.

Characteristics of the boundary separating the laminar and turbulent regions in a transitional boundary layer are examined using a time series of three-dimensional flow fields extracted from direct numerical simulations (DNS). In order to accurately mimic boundary-layer experiments perturbed by grid turbulence, the current simulation includes the leading edge of the flat plate and the incoming homogeneous isotropic turbulence. The Reynolds number based on the momentum thickness reaches up to 1400, and high-resolution three-dimensional flow fields of the DNS data will be publicly accessible via the Johns Hopkins Turbulence Database (JHTDB). The laminar-turbulence discrimination algorithm isolates the turbulence spots within the transition zone and the bounding surface of the fully-turbulent flow. Attention is placed on the cross-stream surface between the transition zone and fully-turbulent boundary layer. The shape of this interface is dictated by a balance between downstream advection, destabilization of upstream flow and merging of turbulence spots. Conditionally sampled statistics are examined across the interface, and are also compared to the downstream equilibrium turbulent boundary layer.

Direct numerical simulations (DNSs) of a fully developed turbulent channel flow with the slip and no-slip boundary conditions were performed to investigate the spatial characteristics of large-scale structures in the dragreducing flow and the effect of the slip wall on those structures. The slip boundary condition was used in the streamwise direction and the overall mean skin-friction coefficient was reduced by 35% relative to the channel flow with the no-slip condition at the same bulk Reynolds number of 10,333. The wall-normal variations of average streamwise and spanwise length scales of the channel flow with the slip condition are relatively greater than those with the no-slip condition and the difference is significant in the outer region (y/δ > 0.2). The populations of low- and highmomentum structures are mainly due to the large populations of low-momentum structures in the channel flow with the slip condition. Based on the conditional twopoint correlations, the average streamwise length scales of negative-u structures in the flow with the slip condition are larger than those with the no-slip condition in the outer layer. Conditionally averaged velocity fields present that the lowmomentum structures in the flow with the slip condition have the weak congregative motions and ejections, by which the near-wall streamwise velocity fluctuations are less affected.

A direct numerical simulation of turbulent pipe flow is performed at Re τ = 3008 with a long streamwise domain length (30R) to investigate wall scaling laws in the physical and wavenumber spaces. The power law follows the streamwise mean velocity in the overlap region, y + = 90 – 300, evaluated by the power law indicator function. The scale-separated Reynolds shear stress shows that the largescale motions (λ x + > 3000) are more responsible to construct the constant-stress layer than the small-scale motions (λ x + < 3000), and thus it is proposed that the large-scale motions more contribute to the growth of the mean velocity in the overlap region than the small-scale motions. In the premultiplied energy spectra of the streamwise velocity fluctuations, the k x -1 region ssociates with the attached eddies appeared in λ x /R = 2 – 5 at y + = 90 – 300 with the bimodal distribution. Linear growth of small-scale energies to large-scale energies helps to appear the k x –1 region at high Reynolds number.

The spatial and temporal formation of very-large-scale motions (VLSMs) is investigated using data from direct numerical simulation (DNS) of turbulent channel flow at a friction Reynolds number Re τ = 930. The physical structures of negative and positive streamwise velocity disturbances (u) are examined. The population and area fraction of the structures reveal that negative-u regions are organized into long streamwise extent as compared with positive-u regions whose streamwise length is dominantly less than 3δ in the log layer. The conditional correlations show that the footprint of negative-u structure is relatively narrower but longer than that of positive counterpart. This difference is due to the opposite spanwise motions in the vicinity of their footprints which are the congregative (negative-u) and dispersive (positive-u) motions in the near-wall region induced by the ejection and sweep events associated with the outer large-scale structures. Instantaneous perturbations fields show that near-wall streaks beneath the merging event undergo a spanwise congregative migration. The time series of conditionally-averaged velocity fields demonstrate that this congregative behavior is generated by a roll motion in the outer region, and enhances the magnitude of the nearwall low-speed streaks. This reduces downstream structure convection speed thus promoting the merging process and the formation of VLSMs. The present study confirms that the top-down and bottom-up interactions are the origin of the local difference in convection velocities of outer large-scale structures, and hence their merging into VLSMs.

The slip over a super-hydrophobic surface has an effect on the drag reduction in wall-bounded turbulent flows. The slip condition mimics a super-hydrophobic surface, which leads to drag reduction via weakened mean-shear rate on the surface. Direct numerical simulations (DNSs) of turbulent channel flows with the slip condition were performed to see the changes of large- and very-large-scale motions (LSMs and VLSMs) over the surface. For comparison, a DNS of turbulent channel flow with the no-slip condition was also carried out. The slip condition was chosen such that the drag reduction rate decreased by 35%. The Reynolds number Re H (≡ U CL H/ν) was 15,500, where U CL , H and ν are the laminar centerline velocity, the channel half-height and the kinematic viscosity, respectively. The drag reduction over a slip surface was examined numerically by Min & Kim (2004) [1]. Recently, Park et al. (2014) experimentally showed that the drag reduction was obtained up to 75% [2]. Although a little information on the slip flow structure was revealed by the pioneers, much research has focused on the turbulence statistics. Because the drag reduction is directly affected by the change of turbulence structures [3], the investigation of drag-reduced turbulent structures is of great importance in the field of the flow control. The objective of the present study, therefore, is to understand turbulent structures including LSMs and VLSMs and the mechanism of the drag reduction over the slip surface, by comparing with the turbulent channel flows with the no-slip condition. A very-long computational domain, which is larger than 30 times of H, should be adopted to describe the LSMs and VLSMs. The computation domain was 10πH × 2H × 3πH in the streamwise, wall-normal, and spanwise directions, respectively.

The relationship between large-scale motions (LSMs) and the shape of the turbulent/non-turbulent interface (TNTI) is examined using data from direct numerical simulation (DNS) of turbulent boundary layer (TBL) flow. The Reynolds number based on the momentum thickness and the free-stream velocity ranges from 1180 to 3500. Feature extraction techniques are used to identify cores of the large-scale motions in the perturbation fields. Since turbulence kinetic energy and enstrophy are different inside low- and high-speed LSMs, the wall-normal elevation of TNTI is correlated with the streamwise momentum of these structures. The large-scale crests and troughs of TNTI are matched to the locations of the wall-detached LSMs of low- and high-speed streaks, respectively. In addition, abrupt changes in turbulence statistics across the TNTI reported in previous studies are associated with population trends of the wall-detached LSMs near the TNTI.

Direct numerical simulation (DNS) of turbulent pipe flow at Re τ = 3008 with a very long streamwise domain length (Lx = 30R, R is a pipe radius) was performed to explore the wall scaling laws in the overlap region. The high-Re turbulent pipe flow was found not to follow a log law, but rather to follow a power law. By contrast, the high-Re turbulent channel flow (Re τ ≥ 2006) followed a log law. A mesolayer was observed in both the pipe and channel flows, in agreement with the power law. The retarded log law in the turbulent pipe flow was attributed to the presence of large-scale structures in the outer region of the pipe flow. These large-scale structures were more dominant in the turbulent channel flows than in the turbulent pipe flows. As the Reynolds number increased, the fluids transitioned from a power law to a log law. The development of large-scale structures in the pipe flow was slower than the corresponding development in the channel flow The proportion of largescale and very-large-scale motions (LSMs and VLSMs) was obtained in comparison with the low-Reynolds-number pipe flow at Re τ = 934.

Direct numerical simulation (DNS) data of a turbulent channel flow at Re τ =930 was scrutinized to investigate the formation of very large-scale motions (VLSMs) by merging of two large-scale motions (LSMs), aligned in the streamwise direction. We mainly focused on the supportive motions by the near-wall streaks during the merging of the outer LSMs. From visualization of the instantaneous flow fields, several low-speed streaks in the near-wall region were collected in the spanwise direction, when LSMs were concatenated in the outer region. The magnitude of the streamwise velocity fluctuations in the streaks was intensified during the spanwise merging of the near-wall streaks. Conditionally-averaged velocity fields around the merging of the outer LSMs showed that the intensified near-wall motions were induced by the outer LSMs and extended over the near-wall regions. The intense near-wall motions influence the formation of the outer low-speed regions as well as the reduction of the convection velocity of the downstream LSMs. The interaction between the near-wall and the outer motions is the essential origin of the different convection velocities of the upstream and downstream LSMs for the formation process of VLSMs by merging.

The hydrodynamic patterns in the wake of swimming jellyfish are based on the bell morphology. Jellyfish with a prolate bell morphology form a clear jet structure in the wake. A three-dimensional computational model was used to analyze the hydrodynamic patterns. The Froude propulsion efficiency, defined by the ratio of the value of the energy required to deform the elastic bell to the value of the average center velocity multiplied by the thrust, was compared with different forms of the elastic bell deformation. The immersed boundary method was adopted to consider the interaction between the swimming jellyfish and surrounding fluid. Due to the effect of the momentum transferred to the surrounding fluid by the bell deformation, the rotational fluid mass was formed, called vortices. The vortex structures in the wake of prolate swimming jellyfish were elucidated in detail in both quantitative and qualitative ways. A dimensionless temporal parameter was employed to investigate the vortex formation process quantitatively. The starting/stopping vortex structures were generated during the contraction/relaxation phase. During the early stage of the contraction, the vortex structures were mainly generated by the stroke, then the ejected fluid was entrained into the vortex structures.

Turbulent pipe and channel flows have been generally accepted as having a similar turbulent flow structure. However, turbulence statistics in the core region of pipe flow are different from that of channel flow owing to the difference in the averaged spanwise dimension of the low-speed structures. In particular, wall-normal and spanwise stresses of channel flow are smaller than those of pipe flow. In the present study, DNS dataset of turbulent channel and pipe flows with the friction Reynolds number Re=934 have been compared to elucidate the difference of statistics in terms of the populations of large and very-large scales in the low-speed region. To this end, large and very-large scales were extracted by a low-pass filtered streak detection algorithm. We found that the population density of large scales of pipe flow is more increased in the core region than that of channel flow. Although the density of very-large scales of pipe flow decreases, the area of low-speed region increases due to the large number of large scales. Further comparison of pipe and channel flows showed that the higher turbulence intensity of pipe flow is caused by the interference of large scales with the azimuthal distance.

Direct numerical simulation was carried out to investigate the spatial features of very large-scale motions (VLSMs) in a turbulent channel flow with Re τ ≈ 930. Using the streak detection algorithm and the population trend, statistics of the spatial extents for both large-scale motions (LSMs) and VLSMs were obtained. This result showed that, at least one VLSM, on average, can be found in the wall-parallel 10δ × δ plane (here δ is the channel half height), which provides a statistical evidence of the presence of VLSMs. Moreover, to clarify one of previous hypotheses regarding the formation of VLSMs, the upstream tracking of individual streaks was utilized for temporal analysis. The present study statistically supported that the connection of the upstream LSMs is mainly contributed to the formation of VLSMs; that is, the concatenation of the small structures to form VLSMs. The downstream tracking of VLSMs indicated that over than 30% of VLSMs survive for 9 wall-unit time. Finally, the spatial organization of LSMs and VLSMs was examined by the conditional averaging.

Direct numerical simulations (DNS) of turbulent boundary layers over isothermally-heated walls were performed to investigate the effect of viscosity stratification on the turbulent and thermal boundary layer flows. An empirical relation of temperature-dependent viscosity for water was adopted. Based on the free-stream temperature (30 ◦ C), two wall temperatures (70 ◦ C and 99 ◦ C) were selected. In the heated flows, the turbulence energy diminishes in the buffer layer, but increases near the wall. The reduction in turbulence kinetic energy in the buffer layer is accompanied by smaller levels of Reynolds shear stresses and, hence, weaker turbulence production. The enhanced turbulence energy near the wall is attributed to enhanced transfer of energy via additional diffusion-like terms due to the viscosity stratification. Wall heating also results in increased scalar flux in the sublayer. Large wall-normal gradients of U and Θ lead to increased production in the scalar flux budget.

Direct numerical simulation (DNS) database of a zero pressure-gradient turbulent boundary layer was scrutinized to investigate the spatial distribution of vortices around a streamwise-elongated low-speed structure. The turbulent flow field has been numerically produced by preceding the bypass transition simulation using the isotropic free-stream turbulence, in which the Reynolds number reaches up to Reθ = 3280. In the present study, the low-speed structures and vortices have been identified and tracked in order to obtain spatial properties of the large-scale structures. The information about the inclination angle, distance and population of individual vortices on the long streamwise structure has been investigated. Finally, the present study provides statistical evidence on the formation of largescale packet-like structure and its variation along the downstream.

난류 채널 유동의 스펙트럴 해석과 병렬 계산을 소개하는 발표입니다.

Direct numerical simulations of turbulent flows over isothermally-heated walls were performed to investigate the influence of viscosity stratification on boundary-layer turbulence and drag. The adopted model for temperature-dependent viscosity was typical of water. The free-stream temperature was set to 30◦ C, and two wall temperatures, 70◦ C and 99◦ C, were simulated. In the heated flows, the mean shear-rate is enhanced near the wall and reduced in the buffer region, which induces a reduction in turbulence production. On the other hand, the turbulence dissipation is enhanced near the wall, despite the the reduction in fluid viscosity. The higher dissipation is attributed to a decrease in the smallest length scales and near-wall fine-scale motions. The combined effect of the reduced production and enhanced dissipation leads to lower Reynolds shear stresses and, as a result, reduction of the skin-friction coefficient.

Direct numerical simulations of turbulent boundary layers with temperature-dependent viscosity were performed to investigate the influence of wall-heating on skin friction. The Arrhenius type viscosity model was chosen to represent water at atmospheric pressure. Based on the free-stream temperature of T∞=30C, two wall temperatures (Tw=70C and 99C) were selected. These wall temperatures correspond to viscosity ratios (w∞) of approximately 50% and 35% at the wall. When the wall is heated, a low-viscosity layer near the wall is formed. The reduced viscosity leads to a decrease in the Kolmogorov length-scale () of flow in the near-wall region. The smallest scale of motion near the wall enhances the dissipation range of the turbulent kinetic energy (TKE) spectrum (figure 2), which results in augmented dissipation in the TKE budget. Furthermore, the viscosity drop due to wall-heating also gives rise to increased mean shear rate in the near-wall region and decreased mean shear rate in the buffer region. Reduction of the shear rate in the buffer region induces a decreased production of TKE which can weaken vortical structures as shown in figure 3. Subsequently, the combined increase of the dissipation and decrease of the production lead to a reduction of the Reynolds shear stress, which is the largest contribution to the skin friction coefficient. The skin-friction coefficient and its reduction rate are reported in figure 4. The drag reduction mechanism will be discussed with focus on turbulent structures and their role.

원뿔형 디퓨저의 난류 구조와 역압력 구배의 영향을 수치해석으로 조사합니다.

벽 가열에 따른 점성 변화가 물의 난류 경계층 구조와 운동량 수송에 미치는 영향을 분석합니다.

Direct numerical simulation (DNS) of open-channel flow in the presence of an air-water interface was performed to examine the effects of interface deformation on the turbulence structures. In the water-driven turbulence, flows characterized by either of two Froude numbers (Fr=0.2 and 0.8) were examined and compared. A coupled level-set and volume-offluid (CLSVOF) method was employed to track the interface. The air that contacted the water was entrained into the turbulent flow. At high Fr, all turbulent normal stresses on the air side of the interface were high near the sidewall. Moreover, all terms of the Reynolds shear stress were intensified at the mixed-boundary corner on the air side of the interface. Twopoint correlations between the streamwise vortex and the velocity fluctuations provided structural information about the near-wall streaky structures and the inner secondary flows in the cross-stream plane. Linear stochastic estimates of the conditionally averaged flow field showed that the inner secondary flow consisted of not only in-plane velocity components but also streamwise velocity components.

Direct numerical simulation (DNS) of open-channel flow in the presence of an air-water interface was performed to examine the effects of interface deformation on the turbulence structures. In the waterdriven turbulence, flows characterized by either of two Froude numbers (Fr=0.2 and 0.8) were examined and compared. A coupled level-set and volume-of-fluid (CLSVOF) method was employed to track the interface. At high Fr, all turbulent normal stresses on the air side of the interface were high near the sidewall. Two-point correlations between the streamwise vortex and the velocity fluctuations provided structural information about the near-wall streaky structures and the inner secondary flows in the cross-stream plane. Linear stochastic estimates of the conditionally averaged flow field showed that the inner secondary flow consisted of not only in-plane velocity components but also streamwise velocity components.

Coherent wall-normal structures in turbulent boundary layer (TBL) have been investigated by scrutinizing the direct numerical simulation (DNS) dataset with Re θ =2500. The spatial signatures of hairpin vortex legs were frequently observed in the vector fields of the wall-normal/spanwise (yz ) plane and it was found that groups of such hairpin legs induce wall-normal-aligned elongated structures with a large Reynolds shear stress; this result strongly supports the typical hairpin packet model. The two-point correlation of the velocity fluctuations showed that the wall-normal length scales vary linearly with the distance from the wall and the two-point correlations between the signed swirling motions showed that interactions between counter-rotating vortices are predominant throughout the boundary layer and especially frequent in the near-wall region.

역압력 구배가 있는 난류 경계층의 DNS 데이터를 분석하여 횡방향 와류 분포와 대규모 유동 구조의 관계를 조사합니다.

Turbulent structures in spanwise/wall-normal plane of the turbulent boundary layers (TBLs) subjected to adverse pressure gradients (APGs) were investigated by analyzing the DNS database of Lee & Sung (2009). Probability density functions of the strength of the vortex cores normalized by their r.m.s. values displayed that strong swirling motions are frequently observed on the APG TBLs than zero pressure gradient TBLs. Influence of APGs on the population trends of spanwise vortex cores showed that those have a local maximum at the outer region for APG TBLs which might be due to the maximum Reynolds shear stress. Moreover, two-point correlations and linear stochastic estimations were scrutinized to provide statistical evidence for hairpin packet motions in the vertical plane of the TBLs with APGs. We found that wall-normal extent of the contours is elongated vertically owing to the strong swirling motion of the individual vortex located in the wake region or wake-type detached structures.