The relationship between body weight (BW) and respiratory minute volume (V(m)) was reviewed by co... more The relationship between body weight (BW) and respiratory minute volume (V(m)) was reviewed by collecting a database from the literature. The data were separated into anaesthetized and non-anaesthetized groups. Only young adult terrestrial mammals were included in the final data set. This database is the largest to be reported to date, is the first to separate the anaesthetized and non-anaesthetized groups and is matched to the target population of young, fit adult humans. The data set of non-anaesthetized animals contained 142 studies representing 2616 animals and 18 species from mice at 12 g body weight to horses and a giraffe at ca. 500 kg body weight. Analysis of the data indicated a power law (allometric) relationship between the minute volume and body weight. The resulting allometric equations for the empirical relationship between minute volume and body weight are: log(10)V()(m)= -0.302 + 0.809 log(10)BW and V(m) = 0.499 BW(0.809)where V(m) is the minute volume (l min(-1)) and BW is the body weight (kg). From these equations, a minute volume of 15.5 lmin(-1)was obtained for a 70 kg human in the same physiological and/or emotional state as the animals. The results of the analyses were compared to other empirical studies in the literature, the more recent of which also indicated a scaling factor of 0.8. The relationship between minute volume and body weight is recommended for use in estimating the inhalation toxicity to young adult humans (military personnel), because this is the first study to use a large database focused exclusively upon non-anaesthetized young adult terrestrial mammals.
The properties of the c-response function c (w) arising from a very extensive class of conductivi... more The properties of the c-response function c (w) arising from a very extensive class of conductivity distributions (class s') are studied. A generalized Helmholtz equation for the complex electric field, that involves a formal differential operator incorporating the slowness function as the Lebesgue-Stieltjes measure; is obtained and its equivalent formulation in terms of a Lebesgue-Stieltjes integral equation is developed. Two fundamental solutions w,(z, w) and wz(z, w) to the generalized Helmholtz equation are obtained and their relationship to the c-response function is derived. The analytic properties of c (w) are deduced from those of wl(z, w) and w2(z, w). The relationship between c E (o) and E(z,, w) is investigated, where cE(w) is the c-response function observed at the Earth's surface, for a conductivity profile u E 3' supported on the interval [0, zm] and terminated in the basement half-space at z = z,,, by a arbitrary conductivity distribution with c-response function given by E(zm, w). The two response functions, namely cE(w) and C(zm, w) , are shown to be related through a Mobius transformation. The analytic properties of c (w) are utilized to develop two representations for the response function. First, an infinite product expansion based on the Weierstrass factor theorem is obtained and second, an integral representation based on Cauer's representation theorem is developed. This integral representation is the same as that obtained by Weidelt for conductivity distributions of class Co+. Furthermore, an integral representation of exponential form is derived for the apparent resistivity function. Finally, the properties of the Mobius transformation which relate cE(w) to E(z,, w) are exploited to provide a quantitative characterization of the critical depth below which the conductivity distribution may be arbitrarily specified without affecting the surface response function. This corresponds to the region of indeterminancy first studied by Parker. The result of this analysis is applied to the Larsen response data set.
Journal Of Geophysical Research: Atmospheres, Sep 24, 2014
The Cooperative Research Centre for Greenhouse Gas Technologies (CO2CRC) Otway Project is Austral... more The Cooperative Research Centre for Greenhouse Gas Technologies (CO2CRC) Otway Project is Australia's first demonstration of the geological storage of carbon dioxide (CO 2), where about 65,000 metric tons of fluid consisting of 92% CO 2 and 8% methane (CH 4) by mass have been injected underground. As part of the project objective of developing methodologies to detect, locate, and quantify potential leakage of the stored fluid into the atmosphere, we formulate an inverse atmospheric model based on a Bayesian probabilistic framework coupled to a state-of-the-art backward Lagrangian particle dispersion model. A Markov chain Monte Carlo method is used for efficiently sampling the posterior probability distribution of the source parameters. Controlled experiments used to test the model involved releases of the injected fluid from one of the nearby wells and were staggered over 1 month. Atmospheric measurements of CO 2 and CH 4 concentrations were taken at two stations installed in an upwind-downwind configuration. Modeling both the emission rate and the source location using the concentration measurements from only two stations is difficult, but the fact that the emission rate was constant, which is not an unrealistic scenario for potential geological leakage, allows us to compute both parameters. The modeled source parameters compare reasonably well with the actual values, with the CH 4 tracer constraining the source better than CO 2 , largely as a result of its 6 times higher signal-to-noise ratio. The results lend confidence in the ability of atmospheric techniques to quantify potential leakage from CO 2 storage as well as other source types. LUHAR ET AL.
Adopting a purely systems-theoretic viewpoint, it is shown how a small set of well-established ge... more Adopting a purely systems-theoretic viewpoint, it is shown how a small set of well-established general physical principles such as causality, stability, and passivity, applied to the earth system, determines the analytical properties of the associated impedance tensor in the complex frequency plane. These physical principles and their consequences are utilized to obtain an integral representation for the impedance tensor based upon the Cauer representation for positive functions. This representation is developed primarily by the imposition of the passivity and symmetry requirement on the impedance tensor. As an application, the representation is employed to develop a set of inequality constraints (necessary conditions) that must be verified by the impedance tensor. Since linearity and passivity imply causality, the Cauer integral representation for the impedance tensor is then utilized to derive a subtracted dispersion relation that connects the real and imaginary Hermitian parts of the tensor on the real frequency axis. Furthermore, it is shown how subtracted dispersion relations for the impedance tensor may be developed directly from the causality principle without recourse to the Cauer representation. In this fashion, the dispersion relations for the scalar response function, developed by Weidelt and Fischer & Schnegg in the context of 1-D and 2-D earth systems, respectively, have been extended to the tensor response function corresponding to general 3-D earth systems. Although the focus in this paper is limited to the explicit development of dispersion relations for the impedance tensor, the applicability and scope of the methods utilized are of a more general nature and can be employed to develop integral relationships for other Earth response functions (e.g. tipper responses). For the general 3-0 earth, the dispersion relations have direct applications in the formulation of consistency tests, the construction of consistent estimates of the impedance tensor, and the identification of non-linearities.
The dispersion of windborne material released near the earth`s surface is strongly influenced by ... more The dispersion of windborne material released near the earth`s surface is strongly influenced by this impenetrable boundary, which inhibits downward mixing and creates sharp vertical gradients in wind, temperature, turbulence. These strong gradients and the continuous creation of turbulence at the surface cause a rapid evolution of the vertical concentration structure for material released into the atmospheric surface layer (ASL). Recent developments in fast-response instrumentation and an increased realization of potential hazards from the release of common industrial chemicals into the ASL have led to a series of tripartite (US, UK, Canada) field experiments at the US Army Dugway Proving Ground, Utah. This paper contains a preliminary analysis of the data from the most recent follow-on experiments, which included measurements of the vertical profiles of mean and peak concentrations.
Submitted for the DFD13 Meeting of The American Physical Society On the interaction of two scalar... more Submitted for the DFD13 Meeting of The American Physical Society On the interaction of two scalar plumes in a turbulent flow BING-CHEN WANG, SHAHIN OSKOUIE, Univ. of Manitoba, EUGENE YEE, Defence R&D Canada-Suffield-Direct numerical simulation is used to study the interaction of two plumes released by two point sources in the context of a turbulent open channel flow. This study is inspired by the classical experiment in quantum physics, Young's double-slit interference experiment on light and energy. The results of the first-order concentration statistics show that in the convective range, the two ground plumes mix faster in the spanwise direction, however, in the turbulent diffusion range, the elevated plumes spread and mix faster. It is observed that streamwise evolution of the second-order correlation function at the midpoint between the two plumes exhibits four distinct mixing stages for both ground and elevated sources. The second-order correlation function demonstrates that the degree of mixing is minimum at the midpoint between the two plumes and is maximum at plume fringes. In general, the elevated plumes exhibit higher degree of mixing in comparison with the ground plumes due to the meandering effects. The scatterplots prove that the higher order statistics of the concentration can be predicted by the knowledge of the first and second-order statistics.
International Journal of Heat and Fluid Flow, 2015
The dispersion and mixing of passive scalars released from two near-ground point sources into an ... more The dispersion and mixing of passive scalars released from two near-ground point sources into an open-channel flow are studied using direct numerical simulation. A comparative study based on eight test cases has been conducted to investigate the effects of Reynolds number and source separation distance on the dispersion and interference of the two plumes. In order to determine the nonlinear relationship between the variance of concentration fluctuations of the total plume and those produced by each of the two plumes, the covariance of the two concentration fields is studied in both physical and spectral spaces. The results show that at the source height, the streamwise evolution of the cross correlation between the fluctuating components of the two concentration fields can be classified into four stages, which feature zero, destructive and constructive interferences and a complete mixing state. The characteristics of these four stages of plume mixing are further confirmed through an analysis of the pre-multiplied co-spectra and coherency spectra. From the coherency spectrum, it is observed that there exists a range of 'leading scales', which are several times larger than the Kolmogorov scale but are smaller than or comparable to the scale of the most energetic eddies of turbulence. At the leading scales, the mixing between the two interfering plumes is the fastest and the coherency spectrum associated with these scales can quickly approach its asymptotic value of unity.
The geometrical properties of isoconcentration surfaces in a plume dispersing in the atmospheric ... more The geometrical properties of isoconcentration surfaces in a plume dispersing in the atmospheric surface layer are studied using a generalized box-counting method applied to a limited random point set. This method yields the hierarchy of generalized dimensions Dq that can be used to characterize the fractal nature of the plume concentration level sets. The dimension spectra for the concentration level sets are computed from one-dimensional cuts of the concentration field. The concentration level sets are found to be monofractals that can be characterized by one scaling exponent or fractal dimension. The fractal dimension of the level sets is independent of the concentration threshold over a wide range of threshold values. The evolution of the fractal dimension of plume concentration level sets with distance x downwind from the source, cross-wind distance y from the lateral mean-plume centerline, and vertical height z above the ground is examined. At a fixed plume height, the fractal...
A modified k-model is used for the simulation of the mean wind speed and turbulence for a neutral... more A modified k-model is used for the simulation of the mean wind speed and turbulence for a neutrally-stratified flow through and over a building array, where the array is treated as a porous medium with the drag on the unresolved buildings in the array represented by a distributed momentum sink. More specifically, this model is based on time averaging the spatially averaged Navier-Stokes equation, in which the effects of the obstacle-atmosphere interaction are included through the introduction of a distributed mean-momentum sink (representing drag on the unresolved buildings in the array). In addition, closure of the timeaveraged, spatially averaged Navier-Stokes equation requires two additional prognostic equations, namely one for the time-averaged resolved-scale kinetic energy of turbulence, j, and another for its dissipation rate,. The performance of the proposed model and some simplified versions derived from it is compared with the spatially averaged, time-mean velocity and various spatially averaged Reynolds stresses diagnosed from a high-resolution computational fluid dynamics (CFD) simulation of the flow within and over an aligned array of sharp-edged cubes with a plan area density of 0.25. Four different methods for diagnosis of the drag coefficient C D for the aligned cube array, required for the volumetric drag force representation of the cubes, are investigated here. We found that the model predictions for mean wind speed and turbulence in the building array were not sensitive to the differing treatments of the source and sink terms in the j and equations (e.g., inclusion of only the 'zeroth-order' approximation for the source/ sink terms compared with inclusion of a higher-order approximation for the source/sink terms in the j and equations), implying that the higher-order approximations of these source/sink terms did not offer any predictive advantage. A possible explanation for this is the utilization of the Boussinesq linear stress-strain constitutive relation within the k-modelling framework, whose implicit omission of any anisotropic eddy-viscosity effects renders it incapable of predicting any strong anisotropy of the turbulence field that might exist in the building array.
International Journal of Computational Fluid Dynamics, Aug 9, 2022
Several machine learning frameworks for augmenting turbulence closure models have been recently p... more Several machine learning frameworks for augmenting turbulence closure models have been recently proposed. However, the generalizability of an augmented turbulence model remains an open question. We investigate this question by systematically varying the training and test sets of several models. An optimal three-term tensor basis expansion is used to develop a model-agnostic data-driven turbulence closure approximation. Then, hyperparameter optimization was performed for a random forest, a neural network, and an eXtreme Gradient Boosting (XGBoost) model. We recommend XGBoost for data-driven turbulence closure modelling owing to its low-tuning cost and good performance. We also find that machine learning models generalize well to new parametric variations of flows seen in the training dataset, but lack generalizability to new flow types. This generalizability gap suggests that machine learning methods are most suited for developing specialized models for a given flow type, a problem often encountered in industrial applications.
International Journal of Heat and Fluid Flow, Dec 1, 2007
The present study is focused on large eddy simulations (LES) that use a statistical (RANS) turbul... more The present study is focused on large eddy simulations (LES) that use a statistical (RANS) turbulence model near solid walls, and on the artificial buffer layer that is formed at the interface between these two modeling regions. Additional forcing is used to trigger resolved motions in the LES region more quickly, and leads to improved results in several ways. The study investigates the artificial buffer layer and how it changes with the use of forcing in an in-depth manner, with the purpose of increased understanding of the increasingly popular hybrid LES/RANS group of methods. The artificial buffer layer is shown to extend from below the modeling interface to well above it, in fact up to 20% of the boundary layer thickness for the cases studied here. The artificial buffer layer is found to be similar to the true buffer layer in many aspects, including a high correlation between the streamwise and wall normal velocity components in the 'superstreaks'. This indicates that while the superstreaks are highly anisotropic and have unphysical length scales, they still contribute to the resolved shear stress. The forcing does not remove the artificial buffer layer, but it does reduce its extent and increases the resolved shear stress. This increase is mainly associated with increased fluctuations of the wall normal velocity. A simple, low-dimensional forcing model is proposed and tested, with favorable results. The model is simple to implement and easily generalized to more complex geometries.
A study of the neutrally-stratified flow within and over an array of three-dimensional buildings ... more A study of the neutrally-stratified flow within and over an array of three-dimensional buildings (cubes) was undertaken using simple Reynolds-averaged Navier-Stokes (RANS) flow models. These models consist of a general solution of the ensemble-averaged, steady-state, threedimensional Navier-Stokes equations, where the k-turbulence model (k is turbulence kinetic energy and is viscous dissipation rate) has been used to close the system of equations. Two turbulence closure models were tested, namely, the standard and Kato-Launder k-models. The latter model is a modified k-model designed specifically to overcome the stagnation point anomaly in flows past a bluff body where the standard k-model overpredicts the production of turbulence kinetic energy near the stagnation point. Results of a detailed comparison between a wind-tunnel experiment and the RANS flow model predictions are presented. More specifically, vertical profiles of the predicted mean streamwise velocity, mean vertical velocity, and turbulence kinetic energy at a number of streamwise locations that extend from the impingement zone upstream of the array, through the array interior, to the exit region downstream of the array are presented and compared to those measured in the wind-tunnel experiment. Generally, the numerical predictions show good agreement for the mean flow velocities. The turbulence kinetic energy was underestimated by the two different closure models. After validation, the results of the high-resolution RANS flow model predictions were used to diagnose the dispersive stress, within and above the building array. The importance of dispersive stresses, which arise from point-to-point variations in the mean flow field, relative to the spatially-averaged Reynolds stresses are assessed for the building array.
The relationship between body weight (BW) and respiratory minute volume (V(m)) was reviewed by co... more The relationship between body weight (BW) and respiratory minute volume (V(m)) was reviewed by collecting a database from the literature. The data were separated into anaesthetized and non-anaesthetized groups. Only young adult terrestrial mammals were included in the final data set. This database is the largest to be reported to date, is the first to separate the anaesthetized and non-anaesthetized groups and is matched to the target population of young, fit adult humans. The data set of non-anaesthetized animals contained 142 studies representing 2616 animals and 18 species from mice at 12 g body weight to horses and a giraffe at ca. 500 kg body weight. Analysis of the data indicated a power law (allometric) relationship between the minute volume and body weight. The resulting allometric equations for the empirical relationship between minute volume and body weight are: log(10)V()(m)= -0.302 + 0.809 log(10)BW and V(m) = 0.499 BW(0.809)where V(m) is the minute volume (l min(-1)) and BW is the body weight (kg). From these equations, a minute volume of 15.5 lmin(-1)was obtained for a 70 kg human in the same physiological and/or emotional state as the animals. The results of the analyses were compared to other empirical studies in the literature, the more recent of which also indicated a scaling factor of 0.8. The relationship between minute volume and body weight is recommended for use in estimating the inhalation toxicity to young adult humans (military personnel), because this is the first study to use a large database focused exclusively upon non-anaesthetized young adult terrestrial mammals.
The properties of the c-response function c (w) arising from a very extensive class of conductivi... more The properties of the c-response function c (w) arising from a very extensive class of conductivity distributions (class s') are studied. A generalized Helmholtz equation for the complex electric field, that involves a formal differential operator incorporating the slowness function as the Lebesgue-Stieltjes measure; is obtained and its equivalent formulation in terms of a Lebesgue-Stieltjes integral equation is developed. Two fundamental solutions w,(z, w) and wz(z, w) to the generalized Helmholtz equation are obtained and their relationship to the c-response function is derived. The analytic properties of c (w) are deduced from those of wl(z, w) and w2(z, w). The relationship between c E (o) and E(z,, w) is investigated, where cE(w) is the c-response function observed at the Earth's surface, for a conductivity profile u E 3' supported on the interval [0, zm] and terminated in the basement half-space at z = z,,, by a arbitrary conductivity distribution with c-response function given by E(zm, w). The two response functions, namely cE(w) and C(zm, w) , are shown to be related through a Mobius transformation. The analytic properties of c (w) are utilized to develop two representations for the response function. First, an infinite product expansion based on the Weierstrass factor theorem is obtained and second, an integral representation based on Cauer's representation theorem is developed. This integral representation is the same as that obtained by Weidelt for conductivity distributions of class Co+. Furthermore, an integral representation of exponential form is derived for the apparent resistivity function. Finally, the properties of the Mobius transformation which relate cE(w) to E(z,, w) are exploited to provide a quantitative characterization of the critical depth below which the conductivity distribution may be arbitrarily specified without affecting the surface response function. This corresponds to the region of indeterminancy first studied by Parker. The result of this analysis is applied to the Larsen response data set.
Journal Of Geophysical Research: Atmospheres, Sep 24, 2014
The Cooperative Research Centre for Greenhouse Gas Technologies (CO2CRC) Otway Project is Austral... more The Cooperative Research Centre for Greenhouse Gas Technologies (CO2CRC) Otway Project is Australia's first demonstration of the geological storage of carbon dioxide (CO 2), where about 65,000 metric tons of fluid consisting of 92% CO 2 and 8% methane (CH 4) by mass have been injected underground. As part of the project objective of developing methodologies to detect, locate, and quantify potential leakage of the stored fluid into the atmosphere, we formulate an inverse atmospheric model based on a Bayesian probabilistic framework coupled to a state-of-the-art backward Lagrangian particle dispersion model. A Markov chain Monte Carlo method is used for efficiently sampling the posterior probability distribution of the source parameters. Controlled experiments used to test the model involved releases of the injected fluid from one of the nearby wells and were staggered over 1 month. Atmospheric measurements of CO 2 and CH 4 concentrations were taken at two stations installed in an upwind-downwind configuration. Modeling both the emission rate and the source location using the concentration measurements from only two stations is difficult, but the fact that the emission rate was constant, which is not an unrealistic scenario for potential geological leakage, allows us to compute both parameters. The modeled source parameters compare reasonably well with the actual values, with the CH 4 tracer constraining the source better than CO 2 , largely as a result of its 6 times higher signal-to-noise ratio. The results lend confidence in the ability of atmospheric techniques to quantify potential leakage from CO 2 storage as well as other source types. LUHAR ET AL.
Adopting a purely systems-theoretic viewpoint, it is shown how a small set of well-established ge... more Adopting a purely systems-theoretic viewpoint, it is shown how a small set of well-established general physical principles such as causality, stability, and passivity, applied to the earth system, determines the analytical properties of the associated impedance tensor in the complex frequency plane. These physical principles and their consequences are utilized to obtain an integral representation for the impedance tensor based upon the Cauer representation for positive functions. This representation is developed primarily by the imposition of the passivity and symmetry requirement on the impedance tensor. As an application, the representation is employed to develop a set of inequality constraints (necessary conditions) that must be verified by the impedance tensor. Since linearity and passivity imply causality, the Cauer integral representation for the impedance tensor is then utilized to derive a subtracted dispersion relation that connects the real and imaginary Hermitian parts of the tensor on the real frequency axis. Furthermore, it is shown how subtracted dispersion relations for the impedance tensor may be developed directly from the causality principle without recourse to the Cauer representation. In this fashion, the dispersion relations for the scalar response function, developed by Weidelt and Fischer & Schnegg in the context of 1-D and 2-D earth systems, respectively, have been extended to the tensor response function corresponding to general 3-D earth systems. Although the focus in this paper is limited to the explicit development of dispersion relations for the impedance tensor, the applicability and scope of the methods utilized are of a more general nature and can be employed to develop integral relationships for other Earth response functions (e.g. tipper responses). For the general 3-0 earth, the dispersion relations have direct applications in the formulation of consistency tests, the construction of consistent estimates of the impedance tensor, and the identification of non-linearities.
The dispersion of windborne material released near the earth`s surface is strongly influenced by ... more The dispersion of windborne material released near the earth`s surface is strongly influenced by this impenetrable boundary, which inhibits downward mixing and creates sharp vertical gradients in wind, temperature, turbulence. These strong gradients and the continuous creation of turbulence at the surface cause a rapid evolution of the vertical concentration structure for material released into the atmospheric surface layer (ASL). Recent developments in fast-response instrumentation and an increased realization of potential hazards from the release of common industrial chemicals into the ASL have led to a series of tripartite (US, UK, Canada) field experiments at the US Army Dugway Proving Ground, Utah. This paper contains a preliminary analysis of the data from the most recent follow-on experiments, which included measurements of the vertical profiles of mean and peak concentrations.
Submitted for the DFD13 Meeting of The American Physical Society On the interaction of two scalar... more Submitted for the DFD13 Meeting of The American Physical Society On the interaction of two scalar plumes in a turbulent flow BING-CHEN WANG, SHAHIN OSKOUIE, Univ. of Manitoba, EUGENE YEE, Defence R&D Canada-Suffield-Direct numerical simulation is used to study the interaction of two plumes released by two point sources in the context of a turbulent open channel flow. This study is inspired by the classical experiment in quantum physics, Young's double-slit interference experiment on light and energy. The results of the first-order concentration statistics show that in the convective range, the two ground plumes mix faster in the spanwise direction, however, in the turbulent diffusion range, the elevated plumes spread and mix faster. It is observed that streamwise evolution of the second-order correlation function at the midpoint between the two plumes exhibits four distinct mixing stages for both ground and elevated sources. The second-order correlation function demonstrates that the degree of mixing is minimum at the midpoint between the two plumes and is maximum at plume fringes. In general, the elevated plumes exhibit higher degree of mixing in comparison with the ground plumes due to the meandering effects. The scatterplots prove that the higher order statistics of the concentration can be predicted by the knowledge of the first and second-order statistics.
International Journal of Heat and Fluid Flow, 2015
The dispersion and mixing of passive scalars released from two near-ground point sources into an ... more The dispersion and mixing of passive scalars released from two near-ground point sources into an open-channel flow are studied using direct numerical simulation. A comparative study based on eight test cases has been conducted to investigate the effects of Reynolds number and source separation distance on the dispersion and interference of the two plumes. In order to determine the nonlinear relationship between the variance of concentration fluctuations of the total plume and those produced by each of the two plumes, the covariance of the two concentration fields is studied in both physical and spectral spaces. The results show that at the source height, the streamwise evolution of the cross correlation between the fluctuating components of the two concentration fields can be classified into four stages, which feature zero, destructive and constructive interferences and a complete mixing state. The characteristics of these four stages of plume mixing are further confirmed through an analysis of the pre-multiplied co-spectra and coherency spectra. From the coherency spectrum, it is observed that there exists a range of 'leading scales', which are several times larger than the Kolmogorov scale but are smaller than or comparable to the scale of the most energetic eddies of turbulence. At the leading scales, the mixing between the two interfering plumes is the fastest and the coherency spectrum associated with these scales can quickly approach its asymptotic value of unity.
The geometrical properties of isoconcentration surfaces in a plume dispersing in the atmospheric ... more The geometrical properties of isoconcentration surfaces in a plume dispersing in the atmospheric surface layer are studied using a generalized box-counting method applied to a limited random point set. This method yields the hierarchy of generalized dimensions Dq that can be used to characterize the fractal nature of the plume concentration level sets. The dimension spectra for the concentration level sets are computed from one-dimensional cuts of the concentration field. The concentration level sets are found to be monofractals that can be characterized by one scaling exponent or fractal dimension. The fractal dimension of the level sets is independent of the concentration threshold over a wide range of threshold values. The evolution of the fractal dimension of plume concentration level sets with distance x downwind from the source, cross-wind distance y from the lateral mean-plume centerline, and vertical height z above the ground is examined. At a fixed plume height, the fractal...
A modified k-model is used for the simulation of the mean wind speed and turbulence for a neutral... more A modified k-model is used for the simulation of the mean wind speed and turbulence for a neutrally-stratified flow through and over a building array, where the array is treated as a porous medium with the drag on the unresolved buildings in the array represented by a distributed momentum sink. More specifically, this model is based on time averaging the spatially averaged Navier-Stokes equation, in which the effects of the obstacle-atmosphere interaction are included through the introduction of a distributed mean-momentum sink (representing drag on the unresolved buildings in the array). In addition, closure of the timeaveraged, spatially averaged Navier-Stokes equation requires two additional prognostic equations, namely one for the time-averaged resolved-scale kinetic energy of turbulence, j, and another for its dissipation rate,. The performance of the proposed model and some simplified versions derived from it is compared with the spatially averaged, time-mean velocity and various spatially averaged Reynolds stresses diagnosed from a high-resolution computational fluid dynamics (CFD) simulation of the flow within and over an aligned array of sharp-edged cubes with a plan area density of 0.25. Four different methods for diagnosis of the drag coefficient C D for the aligned cube array, required for the volumetric drag force representation of the cubes, are investigated here. We found that the model predictions for mean wind speed and turbulence in the building array were not sensitive to the differing treatments of the source and sink terms in the j and equations (e.g., inclusion of only the 'zeroth-order' approximation for the source/ sink terms compared with inclusion of a higher-order approximation for the source/sink terms in the j and equations), implying that the higher-order approximations of these source/sink terms did not offer any predictive advantage. A possible explanation for this is the utilization of the Boussinesq linear stress-strain constitutive relation within the k-modelling framework, whose implicit omission of any anisotropic eddy-viscosity effects renders it incapable of predicting any strong anisotropy of the turbulence field that might exist in the building array.
International Journal of Computational Fluid Dynamics, Aug 9, 2022
Several machine learning frameworks for augmenting turbulence closure models have been recently p... more Several machine learning frameworks for augmenting turbulence closure models have been recently proposed. However, the generalizability of an augmented turbulence model remains an open question. We investigate this question by systematically varying the training and test sets of several models. An optimal three-term tensor basis expansion is used to develop a model-agnostic data-driven turbulence closure approximation. Then, hyperparameter optimization was performed for a random forest, a neural network, and an eXtreme Gradient Boosting (XGBoost) model. We recommend XGBoost for data-driven turbulence closure modelling owing to its low-tuning cost and good performance. We also find that machine learning models generalize well to new parametric variations of flows seen in the training dataset, but lack generalizability to new flow types. This generalizability gap suggests that machine learning methods are most suited for developing specialized models for a given flow type, a problem often encountered in industrial applications.
International Journal of Heat and Fluid Flow, Dec 1, 2007
The present study is focused on large eddy simulations (LES) that use a statistical (RANS) turbul... more The present study is focused on large eddy simulations (LES) that use a statistical (RANS) turbulence model near solid walls, and on the artificial buffer layer that is formed at the interface between these two modeling regions. Additional forcing is used to trigger resolved motions in the LES region more quickly, and leads to improved results in several ways. The study investigates the artificial buffer layer and how it changes with the use of forcing in an in-depth manner, with the purpose of increased understanding of the increasingly popular hybrid LES/RANS group of methods. The artificial buffer layer is shown to extend from below the modeling interface to well above it, in fact up to 20% of the boundary layer thickness for the cases studied here. The artificial buffer layer is found to be similar to the true buffer layer in many aspects, including a high correlation between the streamwise and wall normal velocity components in the 'superstreaks'. This indicates that while the superstreaks are highly anisotropic and have unphysical length scales, they still contribute to the resolved shear stress. The forcing does not remove the artificial buffer layer, but it does reduce its extent and increases the resolved shear stress. This increase is mainly associated with increased fluctuations of the wall normal velocity. A simple, low-dimensional forcing model is proposed and tested, with favorable results. The model is simple to implement and easily generalized to more complex geometries.
A study of the neutrally-stratified flow within and over an array of three-dimensional buildings ... more A study of the neutrally-stratified flow within and over an array of three-dimensional buildings (cubes) was undertaken using simple Reynolds-averaged Navier-Stokes (RANS) flow models. These models consist of a general solution of the ensemble-averaged, steady-state, threedimensional Navier-Stokes equations, where the k-turbulence model (k is turbulence kinetic energy and is viscous dissipation rate) has been used to close the system of equations. Two turbulence closure models were tested, namely, the standard and Kato-Launder k-models. The latter model is a modified k-model designed specifically to overcome the stagnation point anomaly in flows past a bluff body where the standard k-model overpredicts the production of turbulence kinetic energy near the stagnation point. Results of a detailed comparison between a wind-tunnel experiment and the RANS flow model predictions are presented. More specifically, vertical profiles of the predicted mean streamwise velocity, mean vertical velocity, and turbulence kinetic energy at a number of streamwise locations that extend from the impingement zone upstream of the array, through the array interior, to the exit region downstream of the array are presented and compared to those measured in the wind-tunnel experiment. Generally, the numerical predictions show good agreement for the mean flow velocities. The turbulence kinetic energy was underestimated by the two different closure models. After validation, the results of the high-resolution RANS flow model predictions were used to diagnose the dispersive stress, within and above the building array. The importance of dispersive stresses, which arise from point-to-point variations in the mean flow field, relative to the spatially-averaged Reynolds stresses are assessed for the building array.
Uploads
Papers by Eugene Yee