Eugene Yee

A new adaptive mesh refinement data structure with an application to detonation

A new Cell-based Structured Adaptive Mesh Refinement (CSAMR) data structure is developed. In our CSAMR data structure, Cartesian-like indices are used to identify each cell. With these stored indices, the information on the parent, children and neighbors of a given cell can be accessed simply and efficiently. Owing to the usage of these indices, the computer memory required for storage of the proposed AMR data structure is only 58 word per cell, in contrast to the conventional oct-tree [P. MacNeice, K.M. Olson, C. Mobary, R. deFainchtein, C. Packer, PARAMESH: a parallel adaptive mesh refinement community toolkit, Comput. Phys. Commun. 330 (2000) 126] and the fully threaded tree (FTT) [A.M. Khokhlov, Fully threaded tree algorithms for adaptive mesh fluid dynamics simulations, J. Comput. Phys. 143 (1998) 519] data structures which require, respectively, 19 and 238 words per cell for storage of the connectivity information. Because the connectivity information (e.g., parent, children and neighbors) of a cell in our proposed AMR data structure can be accessed using only the cell indices, a tree structure which was required in previous approaches for the organization of the AMR data is no longer needed for this new data structure. Instead, a much simpler hash table structure is used to maintain the AMR data, with the entry keys in the hash table obtained directly from the explicitly stored cell indices. The proposed AMR data structure simplifies the implementation and parallelization of an AMR code. Two three-dimensional test cases are used to illustrate and evaluate the computational performance of the new CSAMR data structure.

New dynamic subgrid-scale heat flux models for large-eddy simulation of thermal convection based on the general gradient diffusion hypothesis

Three new dynamic tensor thermal diffusivity subgrid-scale (SGS) heat flux (HF) models are proposed for large-eddy simulation of thermal convection. The constitutive relations for the proposed modelling approaches represent the most general explicit algebraic formulations possible for the family of SGS HF models constructed using the resolved temperature gradient and SGS stress tensor. As a result, these three new models include a number of previously proposed dynamic SGS HF models as special cases. In contrast to the classical dynamic eddy thermal diffusivity SGS HF model, which strictly requires the SGS heat flux be aligned with the negative of the resolved temperature gradient, the three new models proposed here admit more degrees of freedom, and consequently provide a more realistic geometrical and physical representation of the SGS HF vector. To validate the proposed models, numerical simulations have been performed based on two benchmark test cases of neutrally and unstably stratified horizontal channel flows.

Progress and challenges in the development of physically-based numerical models for prediction of flow and contaminant dispersion in the urban environment

The release of chemical, biological, radiological, or nuclear (CBRN) agents by terrorists or rogue states in a North American city (densely populated urban centre) and the subsequent exposure, deposition and contamination are emerging threats in an uncertain world. The modeling of the transport, dispersion, deposition and fate of a CBRN agent released in an urban environment is an extremely complex problem that encompasses potentially multiple space and time scales. The availability of high-fidelity, time-dependent models for the prediction of a CBRN agents movement and fate in a complex urban environment can provide the strongest technical and scientific foundation for support of Canadas more broadly based effort at advancing counter-terrorism planning and operational capabilities. The objective of this paper is to report the progress of developing and validating an integrated, state-of-the-art, high-fidelity multi-scale, multi-physics modeling system for the accurate and efficient prediction of urban flow and dispersion of CBRN (and other toxic) materials discharged into these flows. Development of this proposed multi-scale modeling system will provide the real-time modeling and simulation tool required to predict injuries, casualties and contamination and to make relevant decisions (based on the strongest technical and scientific foundations) in order to minimize the consequences of a CBRN incident in a populated centre.

Turbulence topologies predicted using large eddy simulations

In this paper, turbulence topologies related to the invariants of the resolved velocity gradient and strain rate tensors are studied based on large eddy simulation. The numerical results presented in the paper were obtained using two dynamic models, namely, the conventional dynamic model of Lilly and a recently developed dynamic nonlinear subgrid scale (SGS) model. In contrast to most of the previous research investigations which have mainly focused on isotropic turbulence, the present study examines the influence of near-wall anisotropy on the flow topologies. The SGS effect on the so-called SGS dissipation of the discriminant is examined and it is shown that the SGS stress contributes to the deviation of the flow topology of real turbulence from that of the ideal restricted Euler flow. The turbulence kinetic energy (TKE) transfer between the resolved and subgrid scales of motion is studied, and the forward and backward scatters of TKE are quantified in the invariant phase plane. Some interesting phenomenological results have also been obtained, including a wing-shaped contour pattern for the density of the resolved enstrophy generation and the near-wall dissipation shift of the peak location (mode) in the joint probability density function of the invariants of the resolved strain rate tensor. The newly observed turbulence phenomenologies are believed to be important and an effort has been made to explain them on an analytical basis.

A General Dynamic Linear Tensor-Diffusivity Subgrid-Scale Heat Flux Model for Large-Eddy Simulation of Turbulent Thermal Flows

In this article, a general dynamic linear tensor diffusivity model is proposed for representing the subgrid-scale (SGS) heat flux (HF). The tensor diffusivity for the model is an inhomogeneous linear function of the resolved strain and rotation rate tensors, and includes three conventional dynamic SGS HF modeling approaches as special cases. In contrast to the dynamic SGS eddy diffusivity modeling approach, the proposed model admits more degrees of freedom for representing the SGS thermal diffusivity, allows for nonalignment between the SGS HF and resolved temperature gradient, and consequently provides a more realistic geometric representation of the SGS heat flux. To validate the proposed modeling approach, numerical simulations have been performed based on a combined forced- and natural-convention flow in a vertical channel with a Reynolds number and a Grashof number Gr = 9.6 × 105. In comparison with the reported direct numerical simulation data and the results obtained using the conventional dynamic SGS eddy diffusivity model, it is shown that the proposed model is able to provide good predictions of various flow quantities at the resolved scale and, more important, offer new insights into near-wall flow physics at the subgrid scale.

Feedback-controlled forcing in hybrid LES/RANS

The computational cost of large eddy simulation (LES) increases rapidly with the Reynolds number when applied to attached boundary layers. This problem can be avoided by use of a Reynolds-averaged Navier–Stokes (RANS) model in the inner part of the boundary layer, which reduces the computational cost drastically. Such hybrid LES/RANS methods yield accurate results in general, but suffer from an artificial buffer layer and a shift in the velocity profile around the modeling interface. This velocity shift can be removed by use of additional forcing, but the results are very sensitive to the forcing amplitude. The present paper proposes a feedback algorithm which efficiently finds the appropriate amplitude and thus yields accurate flow statistics. The feedback algorithm is relatively robust, both in that it is insensitive to the values of the parameters involved and that it yields accurate results with different forcing fields and for different Reynolds numbers. It is argued that the feedback algorithm is consistent with the underlying assumptions of hybrid LES/RANS and that it does not introduce additional empiricism into the method.

Quasi-Periodicity of the Drag Coefficient and Nusselt Number Induced by Taylor-Görtler Vortices

In this article, we investigate the impact of Taylor-Görtler vortices on the drag coefficient and Nusselt number in a heated rotating channel flow using the method of large-eddy simulation (LES). We report the observation of quasi-periodicity of the drag coefficient and Nusselt number in the spanwise direction induced by Taylor-Görtler vortices. The physical conditions under which the extrema of the drag coefficient and Nusselt number occur are investigated. The turbulent flow field is characterized by a Reynolds number Re τ = 150 and various rotation numbers Ro τ ranging from 0 to 7.5. Numerical simulations are performed using two advanced dynamic subgrid-scale stress and heat flux models; namely, the dynamic nonlinear model (DNM) for closure of the filtered momentum equation and the dynamic full linear tensor thermal diffusivity model (DFLTDM) for closure of the filtered thermal energy equation.

Ensemble Nonlinear Autoregressive Exogenous Artificial Neural Networks for Short-Term Wind Speed and Power Forecasting

Short-term wind speed and wind power forecasts (for a 72 h period) are obtained using a nonlinear autoregressive exogenous artificial neural network (ANN) methodology which incorporates either numerical weather prediction or high-resolution computational fluid dynamics wind field information as an exogenous input. An ensemble approach is used to combine the predictions from many candidate ANNs in order to provide improved forecasts for wind speed and power, along with the associated uncertainties in these forecasts. More specifically, the ensemble ANN is used to quantify the uncertainties arising from the network weight initialization and from the unknown structure of the ANN. All members forming the ensemble of neural networks were trained using an efficient particle swarm optimization algorithm. The results of the proposed methodology are validated using wind speed and wind power data obtained from an operational wind farm located in Northern China. The assessment demonstrates that this methodology for wind speed and power forecasting generally provides an improvement in predictive skills when compared to the practice of using an “optimal” weight vector from a single ANN while providing additional information in the form of prediction uncertainty bounds.

Prediction of second-order concentration statistics for dispersing plumes in obstacle arrays.

In this paper, the turbulent dispersion of a passive scalar released from a continuous ground-level point-source in two obstacle arrays is simulated using the Reynolds-averaged Navier-Stokes (RANS) method. An explicit algebraic nonlinear turbulent stress model and a tensor-diffusivity scalar-flux model are used for the closure of the ensemble-averaged momentum and scalar transport equations, respectively. The principal physical mechanisms governing the transport of the second-order concentration statistics of a passive scalar are modelled following Yee et al. (2009). The results of the simulation for the flow and concentration statistics are compared against two sets of high-quality water-channel measurement data obtained using laser induced fluorescence (for concentration) and laser Doppler anemometry (for velocity).

Numerical Study of Dual-Plume Interference in a Turbulent Boundary Layer

Direct numerical simulation is used to investigate the interference arising from the dispersion of passive scalar plumes released from a pair of point sources in a fully-developed wall-bounded shear flow. Four different lateral separations of the two sources for both near ground-level and elevated releases are considered. The downwind evolution of the correlation between the plume concentrations along the centreline between the two sources and the behaviour of the lateral profiles of the correlation at various locations downwind of the two sources are examined in detail. Differences in the exceedance probability over a high concentration level for a single plume and the total plume are highlighted and studied, and the effects of destructive and constructive interferences on the exceedance probabilities for the total plume are used to explain these differences. One significant result is that all higher-order (third-order and above) moments of the total concentration can be inferred from the application of a clipped-gamma distribution using the information embodied in only the first- and second-order concentration moments of each single plume, and in the cross-correlation coefficient of the instantaneous concentration of the two plumes.