Filtered velocity was used to decompose the two-point velocity correlation and the kinetic energy in scale space in order to better understand and predict inhomogeneous turbulent flows. Filter functions introduced involve scale s which has a dimension of the square of length. Three types of filtered velocity were defined to propose a new expression for the energy density in scale space of inhomogeneous turbulence. As a first step, DNS data of homogeneous isotropic turbulence was used to examine profiles of turbulent energy and two-point correlation of filtered velocity and their dependence on scale s.

We investigate the effects of the breakage of mirror symmetry on the Lagrangian response function through the Lagrangian renormalized approximation. For the conventional Lagrangian response function, the non-mirror-symmetric part of the Lagrangian response function must be zero. Furthermore, the mirror-symmetric part is not affected by the breakage of mirror symmetry. For the alternative Lagrangian response function, the effects of the breakage of mirror symmetry explicitly remain. Under the scale-similar spectra and time scale, however, we show that the effects of the breakage of mirror symmetry vanish.

Turbulent helicity is known to be generated in the boundary layer when the system is rotated perpendicular to the wall. In order to consider the production and transport mechanism, we used LES of wall-normal rotating turbulent channel flow to investigate the contribution of each term in the budgets of the turbulent helicity and its three parts h_xx, h_yy and h_zz. We found that the pressure-diffusion term had the most contribution to the generation of the turbulent helicity throughout the whole region when the rotation parameter was as small as 0.01-0.02. However, if we decompose the turbulent helicity into three terms, the contribution of the production term is the largest in the budgets of h_xx and h_zz, which are superior to h_yy.

The k － ε model, which is typical of being employed in fluid engineering, is an eddy-viscosity model in which the eddy viscosity coefficient is expressed in terms of turbulent kinetic energy k and its dissipation rate ε for calculating mean variables. However, the k － ε model can only reproduce simple flow fields where one component of shear stress is dominant and the other normal components do not contribute to mean flows. To make the k － ε model applicable to more complex flows, there is a method to use the turbulent helicity, which is a correlated quantity of velocity and vorticity fluctuations. The helicity exists in the Ekman boundary layer, where system rotation is added to the flow field. In this study, we first predict the helicity by its modeled transport equation, and then quantitatively calculate the correction terms appearing in the model of the Reynolds stress, and finally check their effectiveness. In analysis mentioned above, we take advantage of the DNS database of the Ekman boundary layer.

Plumes in a convective flow, whose flow structure is localised in space and time, are considered to be relevant to the turbulent transport in convection. The effective mass, momentum, and heat transports in the convective turbulence are investigated in the framework of time–space double averaging, where a field quantity is decomposed into three parts: the spatiotemporal mean (spatial average of the time average), the dispersion fluctuation (deviation from the spatiotemporal mean), and the chaotic (incoherent) fluctuation. With this double-averaging framework, turbulent correlations such as the Reynolds stress, turbulent mass flux, turbulent internal-energy flux, etc., in the mean-field equations are divided into the dispersion or coherent correlation part and the chaotic or incoherent correlation part. Evolution equations of these two parts of correlations show what is responsible for the interaction between the coherent and incoherent fluctuations. By reckoning the plumes as the coherent fluctuations, a transport model for the convective turbulence is constructed.

As the use of emergency generators for buildings increases, the near-field dispersion of high-temperature and highbuoyancy exhaust gas is becoming a concern. We conducted computational fluid dynamics (CFD) simulations using steady Reynolds-averaged Navier-Stokes (RANS) model and large eddy simulation (LES) on the near-field dispersion of high-buoyancy exhaust gas emitted from a building’s wake and validated using a wind tunnel experiment. Realizable k-ε 2-layer and wall-adapting local eddy viscosity (WALE) models were used in RANS and LES, respectively. The density difference between the light gas and ambient air in the experiment was reproduced as the temperature difference by an incompressible ideal gas model. RANS model exhibited good agreement with the experimental flow field values for the time-averaged velocity but underestimated the turbulent kinetic energy. LES accurately predicted both the timeaveraged velocity and turbulent kinetic energy. For the concentration field, RANS model predicted the region of high time-averaged concentration near the exhaust port but overestimated ground-level values. Meanwhile, LES adequately predicted time-averaged and fluctuating concentrations.

Respiratory diseases can be spread via droplets and droplet nuclei ejected from the infected person due to breathing, speaking, coughing, and sneezing. The droplet nuclei may remain suspended for a long time in the air causing an airborne infection. The present study analyzed the dispersion characteristics of droplets and droplet nuclei and examined the influence of indoor ventilation on their removal. The results from CFD analysis of the spreading path of droplets and droplet nuclei are expected to be utilized for establishing a proper ventilation system and its optimal operating conditions, in order to reduce the risk of respiratory infection.

MEC ocean model was coupled with submodels of fish cage drag and aquaculture waste organic matter load to investigate the diffusion and sinking characteristics of residual food and fish fecal matter of silver salmon farming in Miho bay. The results showed that the fish cage had an influence on the current flow. We also predicted the diffusion and flux of residual food and fish fecal matter, and obtained the relationship between the flux of residual food and fish fecal matter directly below the fish cage, water depth, and current velocity. In the future, the benthic model should be improved to evaluate the sediment environment.

This paper presents an identification of the dynamic system for Kármán-type vortex shedding behind a rectangular prism building model. First, proper orthogonal decomposition was performed on the turbulent velocity data. Magnitudesquared coherence was then applied to detect the phase consistency between the modes. Within the first 4 most energetic modes, mode 2 was found to be controlled by the inflow fluctuation, and modes 1, 3, and 4 depicted the main vortex shedding on the wake of the building. Subsequently, a third-order polynomial regression model was used to fit the dynamics system of modes 1, 3, and 4. This showed an average trend of the state trajectory. The two limit cycles of the regression model depicted the two rotation directions of the Kármán-type vortex, which were the final states if the system is free of perturbations. Additionally, two characteristic periods were recognised from the trajectory generated from the regression model, which indicated short-cycle and long-cycle fluctuation of the wake vortex.

In this study, neural network (NN） is implemented to predict non-isothermal indoor airflow and temperature distributions. Various data preprocessing methods are utilized and results are compared to reveal the impact of data preprocessing on NN performance. The results show that for most cases, different preprocessing methods can lead to similar NN performances with a prediction error of less than 5% for the mean value. Without data preprocessing for output, error submergent is likely to occur, and the gradient descent algorithm may fail to reduce errors of variables with smaller orders of magnitude during the training process.

This research proposed a source term estimation method embedding the unsteady adjoint equation, simulated by largeeddy simulation (LES), into the Bayesian inference frame. The performance of the proposed method was evaluated by the dispersion measurements in the wind tunnel experiment for a regular block-arrayed building group model. The difference between the measurements and the adjoint concentrations was formulated according to Bayesian theorem to estimate the source parameters stochastically. The estimations results were compared with that of existing method, which applies the steady simulation of adjoint equation based on mean flow field of LES. According to the results, the proposed method significantly improved the modeling accuracy of the adjoint equation and estimations of source parameters.

We apply a four-dimensional variational approach to estimate the unsteady three-dimensional flow field around a cylinder in a duct based on single-plane measurement downstream of the cylinder. It is shown that both the initial velocity field and the spatio-temporal distribution of a body force are simultaneously optimized so as to minimize the error between the measurements and their estimates. Consequently, the velocity field on the measurement plane is predicted fairly well even outside the measurement region. On the other hand, there remain discrepancies between the true and estimated velocity fields beyond the measurement plane, which could be overcome by assimilating measurements in a longer time period.