Expedited constrained multi-objective aerodynamic shape optimization by means of physics-based surrogates

Low-cost EM-simulation-driven Multi-objective Optimization of Antennas

Computational fluid dynamics (CFD) models are powerful computational tools to simulate urban-landscape scale atmospheric dispersion events. They are proven to be very useful for security management and emergency response. Essential inputs to CFD models include landscape characteristics, which are often captured by various GIS data layers. While it is logical to couple GIS and CFD models to take advantage of available GIS data and the visualization and cartographic rendering capabilities of GIS, the integration of the two tools have been minimal. In this paper, we took the first step to evaluate the use of GIS data in CFD modeling. Specifically, we explore how efficient is to use GIS data in CFD models and how sensitive the CFD results are to different GIS data formats. Using campus topography and building data, and the FEFLO-URBAN CFD model, we performed atmospheric release simulations using topographic data in contour and raster formats. We found that using raster format was quite efficient and contour data required significant effort. Though the simulation outputs from the two data formats were not identical, their overall outcomes were similar and did not post alarming discrepancies. We concluded that using GIS data have tremendous potential for CFD modeling.

In this paper, a multi-objective optimization of a transonic axial fan to enhance aerodynamic stability has been conducted using three-dimensional Reynolds-Averaged Navier-Stokes equations, surrogate modeling and multi-objective genetic algorithm (MOGA). Hub radius and first rotor chord length of the axial fan were chosen as design variables for the optimization. Peak adiabatic efficiency of the axial fan and stall margin at 60% of the designed rotational speed, were used as objective functions. Latin Hypercube Sampling (LHS) method was used to select design points in the design space. The objective functions were formulated using the response surface approximation (RSA) model. Three LHS samples with different distributions of twelve design points were tested to study their effects on prediction accuracy of the RSA model and optimization results. MOGA with the RSA models based on the best LHS sample, was used to obtain the Pareto-optimal front. As a result of optimization, an improvement of 17.2% in the stall margin at 60% of the designed rotational speed and 2.96% in peak adiabatic efficiency were obtained compared to the reference design. It was also found that distribution of the design points generated by LHS affects the effectiveness of the surrogate-based optimization.

Computational fluid dynamics (CFD) model has been widely used for prototyping data centers. Evolving it to high-fidelity {\em digital twin} is desirable for the management and operations of large-scale data centers. Manually calibrating CFD model parameters to achieve twin-class fidelity by specially trained domain expert is tedious and labor-intensive. To reduce manual efforts, existing automatic calibration approaches developed for various computational models apply heuristics to search model configurations within an empirically defined parameter bound. However, in the context of CFD, each search step requires long-lasting CFD model's iterated solving, rendering these approaches impractical with increased model complexity. This paper presents Kalibre, a knowledge-based neural surrogate approach that performs CFD model calibration by iterating four key steps of i) training a neural surrogate model based on CFD-generated data, ii) finding the optimal parameters at the moment through neural surrogate retraining based on sensor-measured data, iii) configuring the found parameters back to the CFD model, and iv) validating the CFD model using sensor-measured data as the ground truth. Thus, the parameter search is offloaded to the neural surrogate which is ultra-faster than CFD model's iterated solving. To speed up the convergence of Kalibre, we integrate prior knowledge of the twinned data center's thermophysics into the neural surrogate design to improve its learning efficiency. With about five hours computation on a 32-core processor, Kalibre achieves mean absolute errors (MAEs) of $0.81^o$C and $0.75^o$C in calibrating two CFD models for two production data halls hosting thousands of servers each while requires fewer CFD solving processes than existing baseline approaches.

Pulverized coal fired combustion is a major contributor to global emissions of nitric oxide (NO). Release of NO into the atmosphere causes toxic NO, reduction of ozone, formation of smog, and reduced visibility. Computational fluid dynamic (CFD) modelling of NO emissions from pulverized coal fired boilers offers several benefits including optimisation of boiler operating conditions to minimize NO emissions and development of low NO technology. Formation of NO within a pulverized coal fired boiler is a complex process, yet simple mechanisms are used to approximate the formation process. CFD modelling of NO formation requires prediction of the 1. nitrogen release rate from coal during devolatilization and char combustion, 2. possibly the type of nitrogen compound released during devolatilization and char combustion, 3. gas phase conversion of nitrogen compounds to NO (including possible NO reduction mechanisms), 4. contributions of thermal and prompt NO, and 5. conversion of char nitrogen to NO. This thesis assessed currently used models for processes 1-5 for accuracy and reliability in simple systems and CFD modelling. Models for processes 1-5 were improved, if required. CFD models currently assume the rate of nitrogen release equals a times the carbon release rate, where a equals 1. Laboratory and utility boiler measurements showed that a varies between 0.8 and 1.2. The sensitivity of a to NO emissions should be tested in CFD models for a variety of cases. If NO emissions are sensitive to a, a value of 1.07 is recommended for utility boilers as estimated from utility boiler measurements. The type of nitrogen compound released during devolatilization and char combustion may affect the reaction kinetics converting nitrogen compounds to NO. Previous laboratory measurements concluded that HCN and NH3 are the major nitrogen compounds released during devolatilization. The most likely compound released during char combustion is NO. The conversion of HCN and NH3 to NO was found to vary in fuel-rich conditions at temperatures between 1200 K and 2000 K. Therefore a NO prediction model should include conversion of HCN and NH3 to NO. Currently used global kinetic mechanisms were found to inadequately predict NO formation. Commonly used thermal NO mechanisms significantly over-predict NO formation at temperatures greater than 2200 K and long residence times. This trend is not critical in coal-fired systems because maximum temperatures are less than 2000 K. The de Soete mechanism is commonly used to predict the conversion of HCN to NO. This mechanism was found to under-predict NO formation at temperatures less than 1700 K. The de Soete mechanism was adjusted to improve predictions at low temperatures. An investigation into the use of reduced reaction mechanisms found it was not theoretically possible to reduce a comprehensive reaction mechanism to the extent it could be incorporated into a CFD model, and solve within the same simulation time as a CFD model using global kinetics. An empirical equation was developed that fits the profile of NO formation and HCN destruction. The equations fit NO and HCN profiles across a wide range of conditions. Values of NO and HCN at conditions other than the simulated ones are calculated by interpolation. This mechanism offers an alternative approach to prediction of NO formation by global kinetic mechanisms. Currently available char nitrogen to NO equations are simple, and do not consider parameters such as char size, temperature and char surface area. Empirical equations were developed to calculate the conversion of char nitrogen to NO given the char size, temperature, O2 concentration, NO concentration, char surface area and intrinsic reactivity. The accuracy and reliability of NO predictions determined by CFD models was assessed by comparison of model predictions and utility boiler measurements. CFD models were assessed by four criteria including how accurately NO emissions were predicted, how accurately NO profiles and profiles of species important to NO formation within the burner region were predicted, how accurately temperature and O2 concentrations were predicted and how accurately the model predicted trends in NO emissions as combustion conditions changed. The CFD model investigated, Furnace, could not satisfy the set criteria and therefore modelling results should be interpreted carefully.n

A numerical investigation was performed to study the variation in axial and lateral velocity profiles occurring downstream of the inlet nozzle of a typical Westinghouse 17×17 PWR fuel assembly. A Computational Fluid Dynamic (CFD) model was developed with commercial CFD software. The model comprised the lower region of the fuel assembly, including: the Debris Filter Bottom Nozzle (DFBN), P-grid, Bottom Inconel grid, one and half grid span, as well as the lower core plate hole. The purpose of the study was to obtain insight into the flow redistribution resulting from the interaction of the jet arising from the lower core plate hole and the fuel assembly structure. In particular the axial and lateral velocities before and after the nozzle were studied. The results, axial and lateral velocity contours, streamlines and maximum axial and lateral velocity distributions at various elevations are presented and discussed in relation to the potential risk of high turbulent excitation over the rod and the resulting rod-to-grid fretting-wear damage. The CFD model results indicated that the large jet flows from the lower core plate are effectively dissipated by DFBN nozzle and the grids components of the fuel assembly. The breakup of the large jets in the DFBN and the lower grids helps to reduce the steep velocity gradients and thus the rod vibration and fretting-wear risk in the lower part of the fuel assembly. The presented CFD model is one step towards developing advanced tools that can be used to confirm and evaluate the effect of complex PWR structures on flow distribution. In the future the presented model could be integrated in a larger CFD model involving several fuel assemblies for evaluating the lateral velocities generated due to the non-uniform inlet conditions into the various fuel assemblies.

At present, fossil-fuel power plants supply the majority of the world’s demand of electric energy. The environmental and health threats from anthropogenic emissions of greenhouse gases (GHG) of these power plants have been considered as one of the main reasons for global climate change. In order to make the existing fossil fuel energy sources more environmentally friendly, a new generation of advanced power plant with increased output and environmental performance is required to be built. Oxy-fuel combustion technique can potentially provide significant opportunities for near-zero emissions from the existing and new-build power plants by capturing and storing the carbon dioxide emissions from burning fossil fuels and biomass. To obtain a better understanding of the overall oxy-fuel combustion process, it is important to investigate the effects of oxy-fuel-fired conditions on the combustion characteristics and heat transfer performance as innovations in the energy used. As a first step in developing and validating a computational tool to numerically model combustion of gaseous fuel, a computational fluid dynamics (CFD) modelling study has been carried out involving propane combustion with associated chemical reactions, radiative heat transfer, and turbulence. Three different combustion environments that were adopted experimentally in a 100 kW drop tube firing unit, were examined. One air-fired (reference case) and two oxy-fuel-fired cases (21 vol.% O2 for one combustion case (OF21) and 27 vol.% O2 for the other combustion case (OF27)) were investigated. The irreversible singlestep and reversible multi-step reaction mechanisms were considered. The results obtained with the multi-step chemistry mechanism showed improved agreement, particularly in the flame zone. The unburnt fuel in air-fired and OF27 cases was less than that of the OF21 case due to the low oxygen concentration used in the latter combustion case. As a second step of this research program, a comprehensive CFD modelling study was undertaken by integrating the combustion of pulverized dry lignite in several combustion environments. Four different cases were investigated: an air-fired and three different oxy-fuel combustion environments (25 % vol. O2 concentration (OF25), 27 % vol. O2 concentration (OF27), and 29 % vol. O2 concentration (OF29) were considered. The available experimental results from a lab-scale 100 kW firing lignite unit (Chalmers’ furnace) were selected for the validation of these simulations. The findings showed reasonable agreement with the qualitative and quantitative measurements of temperature distribution profiles and species concentration profiles at the most intense combustion locations inside the furnace. Through the use of Computational Fluid Dynamics (CFD), it is concluded that the resident time, stoichiometry, and recycled flue gas rates are relevant parameters to optimize the design of oxy-fuel furnaces. Under the same oxy-fuel combustion conditions as above, the appropriate mathematical models with the related kinetics parameters were implemented. The purpose was to accurately calculate the temperature distributions, species concentrations (O2, CO2, CO, H2O, and H2), NOx emission concentrations, and the radiation heat transfer. In this step of the modelling investigation, the multi-step chemical reaction mechanisms were conducted on the gas-phase and solid-phase of the coal reaction in one-, two-, and three-step reaction schemes. The predicted results showed a reasonably good agreement against the measured data for all the combustion cases, but in the three-step scheme the results were highly improved, particularly in the flame envelope zone. In the final step of the investigation, a computational fluid dynamics (CFD) modelling study has been developed to investigate Victorian brown coal combustion in a 550 MW utility boiler under air-fired and three oxy-fuel-fired scenarios. User-defined functions (UDFs) were written and incorporated into the CFD code in order to calculate the following mathematical models: the PC devolatilization, char burnout, multi-step chemical reactions, mass and heat transfer, carbon in fly-ash, and NOx formation/destruction. A level of confidence of the CFD model was achieved by validating four different parameters of the conventional combustion case. The numerical results of OF29 combustion condition were considerably similar to the reference firing results in terms of gas temperature levels and radiative heat transfer relative to the OF25 and OF27 combustion cases. In addition, a significant increase in the CO2 concentrations and a noticeable decrease in the NOx formation were observed under all oxy-fuel combustion scenarios. This study of oxy-fuel combustion in a large-scale tangentially-fired boiler is important prior to its implementation in real-life.

This paper presents the first simulation model of a tilting pad journal bearing (TPJB) using three-dimensional (3D) computational fluid dynamics (CFD), including multiphase flow, thermal-fluid, transitional turbulence, and thermal deformation of the shaft and pads employing two-way fluid–structure interaction (FSI). Part I presents a modeling method for the static performance. The model includes flow between pads BP, which eliminates the use of an uncertain, mixing coefficient (MC) in Reynold's equation approaches. The CFD model is benchmarked with Reynold's model with a 3D thermal-film, when the CFD model boundary conditions are consistent with the Reynolds boundary conditions. The Reynolds model employs an oversimplified MC representation of the three-dimensional mixing effect of the BP flow and heat transfer, and it also employs simplifying assumptions for the flow and heat transfer within the thin film between the journal and bearing. This manufactured comparison shows good agreement between the CFD and Reynold's equation models. The CFD model is generalized by removing these fictitious boundary conditions on pad inlets and outlets and instead models the flow and temperature between pads. The results show that Reynold's model MC approach can lead to significant differences with the CFD model including detailed flow and thermal modeling between pads. Thus, the CFD approach provides increased reliability of predictions. The paper provides an instructive methodology including detailed steps for properly applying CFD to tilt pad bearing modeling. Parts I and II focus on predicting static and dynamic response characteristic responses, respectively.

Optimal shape design of multi-element trawl-doors using local surrogate models

The constant development of aero engines towards lighter but yet more compact designs, without decreasing their efficiency, has led to gradually increased demands of the lubrication systems, such as the bearing chambers of the aero engine. For this reason, it is of particular importance to increase our level of understanding of the flow field inside the bearing chambers in order to optimize its design and performance. The flow field in such cases is of a complicated nature since there is a strong interaction between air-flow and lubricant oil together with the geometrical configurations and the shaft rotational speed inside the bearing chamber. The behavior of this interaction must be investigated in order to understand the flow field development inside the aero engine bearing and, at a next step, optimize its performance in relation to the lubrication and heat transfer capabilities. Such an effort is presented in this work where an investigation of the air-flow field development inside the front bearing chamber of an aero engine is attempted. The front bearing chamber is divided in two separate smaller sections where the flow passes from the first section partially through the bearing and the holding structure, to the second one where the vent and the scavenge are placed. The investigation was performed with the combined use of experimental measurements and Computational Fluid Dynamics (CFD) modeling. The experimental measurements were carried out with the use of a Laser Doppler Anemometry (LDA) system in an experimental rig modeling the front bearing chamber of an aero engine for real operating conditions taking into account both air-flow and lubricant oil-flow and for a varying number of shaft rotating speeds. The CFD modeling was performed with the use of a commercial CFD package. The air-flow inside the bearing was modeled with the adoption of a porous medium assumption. The experimental measurements and the CFD computations presented similar flow patterns and satisfactory quantitative agreement. At the same time the effect of the important parameters such as the air and oil mass flow together with the shaft rotation speed and the effect of the chamber inside geometry were identified. These conclusions can be exploited in future attempts in combination with the developed CFD model, in order to optimize the efficiency of the lubricant and cooling system. The latter forms the main target of this work which is the development of a useful engineering tool capable of predicting the flow field inside the aero engine bearing so as to be used for optimization efforts.

Urban environmental problem became one of major issues during its urbanization processes. Environmental impacts are assessed during recent urban planning and development. Though the environmental impact assessment considers meteorological impact as a minor component, changes in wind environment during development can largely affect the distribution pattern of air temperature, humidity, and pollutants. Impact assessment of local wind is, therefore, a major element for impact assessment prior to any other meteorological impact assessment. Computational Fluid Dynamics (CFD) models are utilized in various fields such as in wind field assessment during a construction of a new building and in post analysis of a fire event over a mountain. CFD models require specially formatted input data and produce specific output files, which can be analyzed using special programs. CFD's huge requirement in computing power is another hurdle in practical use. In this study, a CFD model and related software processors were automated and integrated as a microscale wind environmental impact assessment system. A supercomputer system was used to reduce the running hours of the model. Input data processor ingests development plans in CAD or GIS formatted files and produces input data files for the CFD model. Output data processor produces various analytical graphs upon user requests. The system was used in assessing the impacts of a new building near an observatory on wind fields and showed the changes by the construction visually and quantitatively. The microscale wind assessment system will evolve, of course, incorporating new improvement of the models and processors. Nevertheless the framework suggested here can be utilized as a basic system for the assessment.

A computationally efficient technique for multiobjective design optimization of narrowband antennas is presented. In our approach, the corrected low-fidelity antenna model [obtained through coarse-discretization electromagnetic (EM) simulations] is enhanced using frequency scaling and response correction, sampled, and utilized to obtain a fast response surface approximation (RSA) antenna surrogate. The RSA model is constructed in the reduced design space. The initial set of Pareto-optimal designs is then utilized to confine the design space further by identifying designs that satisfy minimum requirements with respect to antenna reflection response. The updated RSA model set up in the confined space is subsequently optimized to yield the final Pareto set at the low computational cost.

This paper investigates the simulation, response surface modeling, and optimization of the metalorganic chemical vapor deposition (MOCVD) process for the deposition of gallium nitride (GaN). Trimethylgallium (TMGa) and ammonia (NH3) are the precursors carried by hydrogen into the rotating-disk reactor. The deposition rate of GaN film and its uniformity form the focus of this study. Computational fluid dynamics (CFD) model simulates the deposition of the GaN film. CFD model is employed to identify two design variables, inlet velocity and inlet precursor concentration ratio, which significantly affect the deposition rate and uniformity of GaN film. Compromise response surface method (CRSM) is used to generate response surfaces for average deposition rate and uniformity. These response surfaces are used to generate the Pareto front for the conflicting objectives of optimal rate of average deposition and uniformity. Pareto front captures the trade-off between deposition rate and uniformity of the GaN film. It is observed that for the whole range of design variables, there are numerous options to get stable uniformity levels than deposition rate. The optimal inlet velocity and precursor concentration for different objective functions considered tend to be near the upper bounds.

In the present study, a Computational Fluid Dynamics (CFD) model was set up to analyse the dynamic behaviour of LNG (Liquefied Natural Gas) tanks adopted in heavy trucks refuelling stations in normal operating conditions or during process upsets. The CFD model was based on the Volume-Of-Fluid (VOF) method to simulate the transient behaviour of a multi-phase flow in a 100 m3 vertical cylindrical tank. The model first considered a fully insulated vessel, and subsequently it was modified to take into account possible damages of the insulation layer, this being a critical upset scenario. The CFD model allowed predicting the dynamic pressure build up in the damaged tank, thus obtaining key indications for the evaluation of the vessel mechanical resistance and set up of safety measures. The model also provided the prediction of vapor temperature with respect to time and external heat source, that is a key issue to assess the effectiveness of the thermal insulation.

This critical review delves into the impact of Computational Fluid Dynamics (CFD) modeling techniques, specifically 2D, 2.5D, and 3D simulations, on the performance and vortex dynamics of Darrieus turbines. The central aim is to dissect the disparities apparent in numerical outcomes derived from these simulation methodologies when assessing the power coefficient (Cp) within a defined velocity ratio (λ) range. The examination delves into the prevalent turbulence models shaping Cp values, and offers insightful visual aids to expound upon their influence. Furthermore, the review underscores the predominant rationale behind the adoption of 2D CFD modeling, attributed to its computationally efficient nature vis-à-vis the more intricate 2.5D or 3D approaches, particularly when gauging the turbine’s performance within the designated λ realm. Moreover, the study meticulously curates a compendium of findings from an expansive collection of over 250 published articles. These findings encapsulate the evolution of pivotal parameters, including Cp, moment coefficient (Cm), lift coefficient (Cl), and drag coefficient (Cd), as well as the intricate portrayal of velocity contours, pressure distributions, vorticity patterns, turbulent kinetic energy dynamics, streamlines, and Q-criterion analyses of vorticity. An additional focal point of the review revolves around the discernment of executing 2D parametric investigations to optimize Darrieus turbine efficacy. This practice persists despite the emergence of turbulent flow structures induced by geometric modifications. Notably, the limitations inherent to the 2D methodology are vividly exemplified through compelling CFD contour representations interspersed throughout the review. Vitally, the review underscores that gauging the accuracy and validation of CFD models based solely on the comparison of Cp values against experimental data falls short. Instead, the validation of CFD models rests on time-averaged Cp values, thereby underscoring the need to account for the intricate vortex patterns in the turbine’s wake—a facet that diverges significantly between 2D and 3D simulations. In a bid to showcase the extant disparities in CFD modeling of Darrieus turbine behavior and facilitate the selection of the most judicious CFD modeling approach, the review diligently presents and appraises outcomes from diverse research endeavors published across esteemed scientific journals.