Marina A Zasimova

The paper presents the thoughtful experience of developing and applying both classical and advanced machine learning (ML) methods to the regression task aiming at the prediction of heat transfer intensity of air-cooled heat exchangers in the form of horizontal single-row finned tube bundles streamlined by laminar free convective airflow. The initial data for training ML models were tabular datasets obtained in numerical simulations on the problem of changing the transversal tube pitch and the temperature difference between the base tube and the ambient air. The possibilities, quality, and performance of a number of classical and advanced ML methods were studied to forecast with acceptable precision the mean Nusselt number for the bundles with two varying parameters. An overview of the current state of experimental and numerical studies on the problem of tube bundles of a specified type and operating conditions, as well as the application of ML methods in the fields of fluid mechanics and heat transfer, with an emphasis on the physical problem under consideration, is made. The physical prototype and problem setup, as well as the mathematical model and computational aspects of CFD simulations, are described; examples of typical velocity and temperature fields and local heat transfer characteristics are provided. A brief description of ML models, evaluation metrics of quality, computing resources, and program instruments used in the research is given. The results part includes comparison and assessment of the quality of models, presentation, and analysis of predictions of the Nusselt number for the two datasets involved. Novel correlations constructed by the models on the basis of the CFD data are proposed.

The results of numerical simulation of the formation and motion of turbulent vortex clouds (puffs) resulting from the blowing of pulsed jets with various initial velocities and durations are given. A model of axisymmetric turbulent flow described by the time-dependent Reynolds equations is adopted. It is shown that, regardless of the initial conditions, a puff which has the shape similar to a sphere develops after the same dimensionless time interval from the instant of jet outflow. In the rest of the space the vortex-induced flow is close to a potential flow. It is found that in the vortices the velocity profiles in the axial and transverse directions are close to self-similar profiles and similar each other for various conditions of outflow of pulsed jets. The time dependences of the geometric and kinematic characteristics of puffs, namely, the location of the cloud center (points with the maximum velocity) and the radius of a sphere equivalent in volume to the puff, as well as the maximum and mean velocities, are given and analyzed. For the jet outflow conditions considered, the characteristics of puffs turn out to be similar.

The results of numerical simulation of the formation and motion of turbulent puffs resulting from the blowing of pulsed jets with different initial velocities and durations are presented. A model of an axisymmetric turbulent flow described by non-stationary Reynolds equations is adopted. It is shown that, regardless of the initial conditions, after the same dimensionless time interval from the instant the jet outflow begins, a vortex cloud appears, which has a spherical shape of vortex. The vortex-induced flow in the rest of the space is close to potential. It has been established that the velocity profiles in vortices in the axial and transverse directions are close to self-similar and are similar for different conditions of the outflow of pulsed jets. Time dependences of the geometric and kinematic characteristics of puffs are presented and analyzed: the position of the cloud center (points with maximum velocity) and the radius of a sphere equivalent in volume to a puff, as well as maximum and average velocities. For the studied jet outflow conditions, the characteristics of puffs turn out to be similar.

The contribution presents the results of numerical simulation of 3D turbulent airflow and heat transfer around a seated thermal manikin in a model room with mixing ventilation. The main goal of the study is to evaluate the uncertainties caused by a CFD model. The room and the detailed thermal manikin geometry correspond to the experiment of P.V. Nielsen et al. (2007). The 3D computational model based on the RANS approach with the standard k-ε turbulence model is applied. The radiation effects are taken into account with the surface-to-surface radiation model. The study considers (1) the effect of the supply air flow rate variation and (2) the effect of the thermal manikin geometrical shape. For a particular parameters set, the results of numerical simulation are validated using the experimental data available (P.V. Nielsen et al.). The contribution is focused also on studying the effects of various heat transfer mechanisms on the airflow pattern and thermal comfort parameters: the baseline mixing ventilation scenario is compared with the buoyancy-dominated case.

The objective of the paper is to demonstrate the importance of the unsteady Computational Fluid Dynamics (CFD) simulations and long-term measurements for the reliable assessment of thermal comfort indoors, for proper categorization of the indoor thermal environment and for identifying the reasons for complaints due to draught discomfort. Numerical simulations and experimental measurements were applied in combination to study ventilation in a field laboratory, a university classroom with a controlled indoor environment. Strong unsteadiness of the airflow was registered both in the unsteady RANS results and the real-scale long-term velocity data measured with thermo anemometer. Low-frequency high-amplitude velocity fluctuations observed lead to substantial time variation of the draught rate. In case of categorization of a thermal environment, the point measurements or steady-state RANS computations would lead to wrong conclusions as well as they cannot be used for identification of the reasons for people’s complaints due to draught discomfort if strong unsteadiness of the airflow exists. It is demonstrated that the length of the time interval for draught rate (DR) assessment may not be universal if low-frequency high-amplitude pulsations are present in the room airflow.

The paper presents Large Eddy Simulation (LES) data for the ventilation airflow in an isothermal room corresponding to the Nielsen et al. (1978) test. Based on a simplified formulation with the periodicity boundary conditions in the transverse direction, detailed grid sensitivity study was performed. Using LES data on velocity components and mean speed extracted from a periodic domain-size-independent solution, the paper assesses velocity-to-speed conversion procedures available in the literature for the Reynolds-Averaged Navier-Stokes ventilation airflow data processing.

The goal of the study is the validation of Wall-Modeled Large Eddy Simulation (WMLES) of ventilation in an isothermal room at conditions of experiments by Nielsen et al. Two values of the inlet slot width were considered. Calculations were carried out using the ANSYS Fluent 16.2 software with an algebraic WMLES subgrid-scale model. The uniform computational meshes up to 58 million cells were used. The focus in the results discussion is on the accuracy of mean and rms velocity fields prediction both in the jet zone and the low-velocity occupied zone. It is concluded that to predict the jet spread the meshes used are sufficient. However, there is visible difference between the computed and measured data in the occupied zone especially in the essentially 3D case with narrow inlet slot.

The goal of the study is to validate Large Eddy Simulation (LES) data on mixing ventilation in an isothermal room at conditions of benchmark experiments by Hurnik et al. (2015). The focus is on the accuracy of the mean and rms velocity fields prediction in the quasi-free jet zone of the room with 3D jet supplied from a sidewall rectangular diffuser. Calculations were carried out using the ANSYS Fluent 16.2 software with an algebraic wall-modeled LES subgrid-scale model. CFD results on the mean velocity vector are compared with the Laser Doppler Anemometry data. The difference between the mean velocity vector and the mean air speed in the jet zone, both LES-computed, is presented and discussed.