Journal articles by Andrei Horvat
The ITER Ion Cyclotron Range of Frequencies (ICRF) antenna’s capacity to couple power to plasma i... more The ITER Ion Cyclotron Range of Frequencies (ICRF) antenna’s capacity to couple power to plasma is determined by the plasma Scrape-Off Layer (SOL) profile, shaping of the front strap array, layout of the Port Plug (PP) and detailed design of its RF components. The first two factors are taken into account by the Torino Polytechnic Ion Cyclotron Antenna (TOPICA) calculated strap array Scattering/Impedance24-port (S24 × 24−/Z24 × 24−) matrices, while this paper deals with the optimisation of the PP layout and components. The RF modeling techniques are explained and used to maximize the coupled power under a set of constraints on RF quantities inside the PP. The total PP RF surface conductive and volumetric dielectric losses are calculated. The resulting S-parameters at the rear RF PP flanges are evaluated as input for the design of the pre-match, decoupling and matching network outside the PP. A discussion of the effect of errors on the PP excitation on the coupled power is also included.
The Duct Liner for the International Thermonuclear Experimental Reactors (ITER) Heating Neutral B... more The Duct Liner for the International Thermonuclear Experimental Reactors (ITER) Heating Neutral Beam (HNB) system is a key component in the beam transport system. Power loading on the top and bottom beam scraping panels of the Duct Liner occurs primarily due to direct interception of the HNB and it is highest at the extreme steering angles of the beam. Furthermore, power loading due to direct interception is dependent on the size and orientation of the scraper panels with respect to the neutral beam axis. This paper outlines the design features of the proposed Duct Liner and describes the analysis performed to optimize the compatibility of the top and bottom scraper panels (also known as Duct Liner Modules) with a normal beam operation scenario. Thermo-mechanical analyses have been performed to validate the design of Duct Liner Modules incorporating deep-drilled cooling technology with a peak power density of 1.2 MW/m2 and incident power of 0.27 MW, and also to verify its conformity with ITER structural design criteria. Furthermore, numerical simulations of the transient draining procedure were performed by using a one-dimensional thermo-hydraulic code to demonstrate complete emptying of the proposed parallel-layout cooling circuit without any reliance on conventional gravity draining.
In enclosure fires, density-driven vent flow through an opening to the fire compartment is direct... more In enclosure fires, density-driven vent flow through an opening to the fire compartment is directly dependent on the state of the fire and the evacuation of smoke and hot gases. If a fire is strongly under-ventilated, there may be heavy production of flammable gases. If a sudden opening occurs, e.g. a window breaks or a fireman opens a door to the fire compartment, fresh air enters the compartment and mixes with hot gases, thus creating a flammable mixture that might ignite and create a backdraft. In this article, we consider the critical flow approach to solve the classical hydraulic equations of density-driven flows in order to determine the gravity controlled inflow in a shipping container full of hot unburnt gases. One-third of the container’s height is covered by the horizontal opening. For the initial condition, i.e., just before opening the hatch,
zero velocity is prescribed everywhere. When the hatch is opened, the incoming air flows down to the container floor and the hot gas flows out. The interface in between them (the neutral plane) can move up like a free surface in internal flows, making it possible to use the techniques of open channel hydraulics devised by Pedersen [1].
The phenomenon of backdraft is closely linked to the formation of a flammable region due to the m... more The phenomenon of backdraft is closely linked to the formation of a flammable region due to the mixing process between the unburnt gases accumulated in the compartment and the fresh air entering the compartment through a recently created opening. The flow of incoming fresh air is called the gravity current. Gravity current prior to backdraft has already been studied, Fleischmann (1993, Backdraft phenomena, NIST-GCR-94-646. University of California, Berkeley) and Fleischmann
(1999, Numerical and experimental gravity currents related to backdrafts, Fire Safety Journal); Weng et al. (2002, Exp Fluids 33:398–404), but all simulations and experiments found in the current literature are systematically based on a perfectly regular volume, usually parallelipedic in shape, without any piece of furniture or equipment in the compartment. Yet, various obstacles are normally found in real compartments and
the question is whether they affect the gravity current velocity and the level of mixing between fresh and vitiated gases. In the work reported here, gravity current prior to backdraft in compartment with obstacles is investigated by means of three-dimensional CFD numerical simulations. These simulations use as a reference case the backdraft experiment test carried out by Gojkovic (2000, Initial Backdraft. Department of Fire Safety Engineering, Lunds Tekniska Hoegskola Universitet, Report 3121). The Froude number, the transit time and the ignition time are obtained from the computations and compared to the tests in order to validate the model.
A computational model of flashover is presented that closely follows the experimental setup at CN... more A computational model of flashover is presented that closely follows the experimental setup at CNRS-ENSMA-Poitiers. A propane burner with thermal power of 55 kW is used as a primary source of fire and square beech wood samples (30 mm × 30 mm × 5 mm) as fire spread targets. The computational model describes the wood pyrolysis with a progress variable. Using the conservation of heat fluxes at the solid-gas interface, the thermal diffusion in the wood samples is coupled with the convective and the radiative heat transfer in the ambient gas phase. The incoming heat flux at the upper surface of the wood samples reaches values between 20 and 30 kW/m2. With the ignition and subsequent combustion of the pyrolysis volatiles, the heat flux increases by approx. 12 kW/m2. The results show that the ignition of the wood samples is triggered at an approx. surface temperature of 650 K. Due to large local variations in incident heat flux, significant differences in the ignition times of the wood samples are observed. The comparison of the calculated and the experimentally measured temperature shows a good agreement for the first wood sample and the model predicts the ignition time very well. But for the second and the third wood samples the model overpredicts the temperature, which leads to a premature ignition of these wood samples.
The analysis focuses on the importance of accurate description of walls' thermal behaviour with p... more The analysis focuses on the importance of accurate description of walls' thermal behaviour with particularly relevance to fire safety cases. To accurately predict thermal loads on the walls in a complex fire situation, time dependent phenomena of thermal convection, diffusion, radiation and heat transfer across the wall have to be taken into account. The heat transfer
across the walls is often neglected as it requires fine grid resolution to approximate large temperature gradients. The analysis shows that this can lead to substantial errors and proposes a transient semi-analytical approximation of thermal wall behaviour. Unsteady heat diffusion equation is solved by semi-analytical approximation of a temperature profile in a wall. The temperature profile is approximated in each time step and for each wall grid node separately using the temperature at the wall internal side and Biot number at the external side for
boundary conditions. Wall heat transfer coefficient is calculated as a temperature profile derivative and then used in the coupled numerical model of the fire. The semi-analytical approach does not rely on a discretisation procedure and the accuracy of its solutions is independent of grid spacing. Therefore, such semi-analytical technique is computationally
less expensive and especially attractive for long transient calculations.
Two-fluid model of the WAHA Code for simulation of water hammer transients
A new thermal hydraulic computer code, WAHA, was developed within the WAHALoads project of the Eu... more A new thermal hydraulic computer code, WAHA, was developed within the WAHALoads project of the European Union Fifth Framework Program. The code's aim is simulation of water hammer transients in piping systems and is based on a one-dimensional, two-fluid, six-equation model of the two-phase flow. The WAHA code can describe two-phase flows in long piping systems (1D geometry) with variable cross section. The code contains correlations for heat, mass, and momentum transfer between the phases and for wall friction in dispersed and horizontally stratified flow regimes. The WAHA code physical model takes into account the elasticity of the pipe through Korteweg's equation; it takes into account water properties and the unsteady wall friction, and contains a set of subroutines for calculation of forces on the piping system. Special models in the WAHA code are implemented for abrupt area changes and branches, for constant pressure (tank), for constant velocity (pump), and for valve closure boundary conditions. The physical model and the main correlations and closure laws, which are applied in the present work, are described and verified with several experiments. The code was successfully applied for simulations of two-phase critical flow, several column-separation types of water hammer, and for condensation-induced water hammer experiments.
This paper describes the numerical scheme used in the WAHA code that was developed within the WAH... more This paper describes the numerical scheme used in the WAHA code that was developed within the WAHALoads project for simulations of fast transients in 1D piping systems. Two-fluid model equations described in a companion paper entitled "Two-Fluid Model of the WAHA Code for Simulations of Water Hammer Transients," are solved with an operator splitting procedure: the non-conservative characteristic upwind scheme is used to solve the hyperbolic part of the equations with the non-relaxation source terms, while the relaxation source terms are treated in the second step of the operator splitting procedure. Water properties are calculated with a newly developed set of subroutines that use pretabulated water properties. Special models that were developed for treatment of the abrupt area changes and branches in the piping systems are described. Various test cases, which were used to test the accuracy of the basic numerical scheme and the accompanying numerical models, are described and discussed together with the typical results of simulations.
The article describes full-scale backdraft experiments in a shipping container using methane as a... more The article describes full-scale backdraft experiments in a shipping container using methane as a fuel. Numerical modelling has followed the experimental setup. The numerical
simulations show the initial gravity current, the ignition, the spreading of flame in the enclosure, the external fireball, and the subsequent decay. The Detached Eddy Simulation
(DES) approach has been used to model turbulence. In order to describe the combustion process of the mixture from the local ignition to progressive deflagration, three separate
combustion models have been implemented for laminar, low- and high-intensity turbulence flow regimes. The calculated ignition time is slightly shorter than the average ignition time observed in the experiments. The fire front progresses through the combustible mixture, generating a cloud of hot gases that are accelerated from the container into the external environment. The velocity increases up to 20 m/s. When the fire front reaches the door, combustion continues outside the enclosure as the fuel has been pushed through the door. The comparison between the calculated time history of relative pressure and the pressure sensor record shows that the numerical simulations slightly overpredict the flame front speed, with a stronger pressure pulse and higher temperatures than the observations.
This paper reports preliminary computational fluid dynamics (CFD) simulations of backdraft observ... more This paper reports preliminary computational fluid dynamics (CFD) simulations of backdraft observed in an experimental rig at Lund University. The analysis was performed with the CFX software using the Detached Eddy Simulation (DES) turbulence model, a hybrid of Large Eddy Simulation (LES) and RANS, in combination with the EDM combustion model. The DES model uses a RANS formulation in wall proximity to avoid computationally expensive grid resolution that is necessary for realistic LES predictions in wall layers.
The preliminary results are qualitatively promising. The simulations began at the instant at which the door opens. A stream of fresh and cold air enters the enclosure as a gravity current. In the rig, ignition was triggered by flammable conditions existing at a wire, which was constantly heated. In the CFD model the ignition time is computed automatically when flammability conditions are reached inside the enclosure, at the wire, as part of the analysis. Subsequently, the fire front is formed. The deflagration expels fuel-rich mixture into
environment, and the combustion continues outside the enclosure as a typical ‘secondary’ event. Considering that backdraft is a very complex phenomenon, the outcome is considered by the authors to be encouraging.
Transient numerical simulations of fluid and heat flow were performed for eight heat exchanger se... more Transient numerical simulations of fluid and heat flow were performed for eight heat exchanger segments with cylindrical and wing-shaped tubes in staggered arrangement. Their hydraulic diameters were from 0.5824 to 3.899 cm for the cylindrical tubes, and from 0.5413 to 3.594 cm for the wing-shaped tubes. Based on the recorded time distributions of velocity and temperature, time average Reynolds number Re, drag coefficient Cd, and Stanton number St were calculated. In general, the drag coefficient and the Stanton number are smaller for the wing-shaped tubes than for the cylindrical tubes. However, with an increasing hydraulic diameter, these differences between both forms of tubes diminish. The time average values were further used to construct the drag coefficient and the Stanton number as polynomial functions Cd(
dh,Re) and St(dh,Re).
A solid-gas-phase model for thick wood gasification/combustion is extensively studied, after a re... more A solid-gas-phase model for thick wood gasification/combustion is extensively studied, after a re-examination of the kinetic constants for the char gasification reactions. The solid-phase model, which includes the description of all the relevant heat and mass transfer phenomena and chemical reactions, is coupled with a CFD code for the gas-phase processes. Both the gasification and combustion of single wood logs are simulated (log radius in the range of 0.06-0.1 m, initial moisture content, on a dry basis, 1-81%, inlet gas temperature 1253-1613K, inlet gas velocity 0.5-1.0 m/s, and various compositions of the gaseous mixture). For comparison purposes, a solid-phase model, with global heat and mass transfer coefficients and a constant-property gas phase, is also considered. Although both models predict the mass loss dynamics to be qualitatively similar, the solid-phase model overestimates the total heat flux and underestimates the char combustion rate. Extensive experimental validation of both models is carried out in terms of conversion time and average mass-loss rates. Acceptable
agreement is obtained for the comprehensive model, whereas in the other case, the conversion times are generally
underestimated and the average mass loss rates are overestimated. However, improvements in the predictive
capabilities of the solid-phase model could be achieved through the introduction of corrective factors for the external heat and mass transfer coefficients.
Transient numerical simulations of fluid and heat flow are performed for a number of heat exchang... more Transient numerical simulations of fluid and heat flow are performed for a number of heat exchanger segments with cylindrical and ellipsoidal form of tubes in a staggered arrangement. Based on the recorded time distributions of velocity and temperature, time-average values of Reynolds number, drag coefficient, and Stanton number are calculated. The drag coefficient and the Stanton number are smaller for the ellipsoidal tubes than for the cylindrical tubes. With an increasing hydraulic diameter, the difference between the two forms of tubes diminishes. To validate the selected numerical approach, the calculated time-average values are compared with experimental data. The time-average values are further used to construct the drag coefficient and the Stanton number as polynomial functions of Reynolds number and hydraulic diameter. The polynomial functions obtained are to be used as input correlations for a heat exchanger integral model.
Detailed transient numerical simulations of fluid and heat flow were performed for a number of he... more Detailed transient numerical simulations of fluid and heat flow were performed for a number of heat exchanger segments with cylindrical, ellipsoidal and wing-shaped tubes in a staggered arrangement. The purpose of the analysis was to get an insight of local heat transfer and fluid flow conditions in a heat exchanger and to establish widely applicable drag coefficient and Stanton number correlations for the heat exchanger integral model, based on average flow variables. The simulation results revealed much more complex flow behavior than reported in current literature. For each of the almost 100 analyzed cases, the time distributions of the Reynolds number, the drag coefficient and the Stanton number were recorded, and their average values calculated. Based on these average values, the drag coefficient and the Stanton number correlations were constructed as polynomial functions of the Reynolds number and the hydraulic diameter. The comparison of the collected results also allows more general conclusions on efficiency and stability of the heat transfer process in tube bundles.
A conjugate heat transfer model of fluid flow across a solid heat conducting structure has been b... more A conjugate heat transfer model of fluid flow across a solid heat conducting structure has been built. Two examples are presented: a.) air-stream cooling of the solid structure and b.) flow across rods with volumetric heat generation. To construct
the model, a Volume Average Technique (VAT) has been applied to the momentum and the energy transport equations for a fluid and a solid phase to develop a specific form of porous media flow equations. The model equations have been solved with the semi-analytical Galerkin method. The detailed velocity and temperature fields in the fluid flow and the solid structure have been obtained. Using the solution fields, the whole-section drag coefficient Cd and the whole-section Nusselt number Nu have been also calculated. To validate the developed solution procedure, the results have been compared to the results of the finite volume method and to the experimental data. The comparison demonstrates an excellent agreement.
An alternate approach based on hierarchic modeling is proposed to simulate fluid and heat flow in... more An alternate approach based on hierarchic modeling is proposed to simulate fluid and heat flow in heat exchangers. On the first level, the direct simulations have been performed for the geometry that is similar to a segment of the examined
heat sink. Based on the obtained results, the Reynolds number dependencies of the scaling factors f and St Pr^0.66 have been established. On the second level, the integral model of the whole heat sink has been built using the volume averaging technique (VAT). The averaging of the transport equations leads to a closure problem. The direct model Reynolds number dependencies f and St Pr^0.66 have been used to calculate the local values of the drag coefficient and the heat transfer coefficient, which are needed in the integral model. The example calculations have been performed for 14 different pressure drops across the aluminum heat sink. The whole-section drag coefficient and Nusselt number have been calculated and compared with the experimental data. A good agreement between the modeling results and the experiment data has been reached with same discrepancies in the transitional regime. The constructed computational algorithm offers possibilities for geometry improvements and optimization, to achieve higher thermal effectiveness.
A mathematical model of fluid flow across a rod bundle with volumetric heat generation has been b... more A mathematical model of fluid flow across a rod bundle with volumetric heat generation has been built. The rods are heated with volumetric internal heat generation. To construct the model, a volume average technique (VAT) has been applied to momentum and energy transport equations for a fluid and a solid phase to develop a specific form of porous media flow equations. The model equations have been solved with a semi-analytical Galerkin method. The detailed velocity and temperature fields in the fluid flow and the solid structure have been obtained. Using the solution fields, a whole-section drag coefficient Cd and a whole-section Nusselt number Nu have also been calculated. To validate the developed solution procedure, the results have been compared to the results of a finite volume method. The comparison shows an excellent agreement. The present results demonstrate that the selected Galerkin approach is capable of performing calculations of heat transfer in a cross-flow where thermal conductivity and internal heat generation in a solid structure has to be taken into account. Although the Galerkin method has limited applicability in complex geometries, its highly accurate solutions are an important benchmark on which other numerical results can be tested.
Direct numerical simulation of a passive scalar in fully developed turbulent channel flow is used... more Direct numerical simulation of a passive scalar in fully developed turbulent channel flow is used to show that Nusselt number is not only a function of Reynolds and Prandtl number, but also depends on properties of a heating wall. Variable thickness of the heating wall and variable heater properties, combined in a fluid–solid thermal activity ratio can change the Nusselt number of the turbulent channel flow for up to 1% at the same Reynolds and Prandtl number and at the same wall heat flux.
A fast-running computational algorithm based on the volume averaging technique (VAT) is developed... more A fast-running computational algorithm based on the volume averaging technique (VAT) is developed and solutions are obtained using the Galerkin method (GM). The goal is to extend
applicability of the GM to the area of heat exchangers in order to provide a reliable benchmark for numerical calculations of conjugate heat transfer problems. Using the VAT, the computational algorithm is fast-running, but still able to present a detailed picture of temperature fields in air flow as well as in the solid structure of the heat sink. The calculated whole-section drag coefficient and Nusselt number were compared with finite-volume method (FVM) results and with experimental data to verify the computational model. The comparison shows good agreement. The present results demonstrate that the selected Galerkin approach is capable to perform heat exchanger calculations where the thermal conductivity of the solid structure has to be taken into account.
International Journal of Heat and Mass Transfer, 2003
A fast running computational algorithm based on the volume averaging technique (VAT) is developed... more A fast running computational algorithm based on the volume averaging technique (VAT) is developed to simulate conjugate heat transfer process in an electronic device heat sink. The goal is to improve computational capability in the area of heat exchangers and to help eliminate some of empiricism that leads to overly constrained designs with resulting economic penalties.
Uploads
Journal articles by Andrei Horvat
zero velocity is prescribed everywhere. When the hatch is opened, the incoming air flows down to the container floor and the hot gas flows out. The interface in between them (the neutral plane) can move up like a free surface in internal flows, making it possible to use the techniques of open channel hydraulics devised by Pedersen [1].
(1999, Numerical and experimental gravity currents related to backdrafts, Fire Safety Journal); Weng et al. (2002, Exp Fluids 33:398–404), but all simulations and experiments found in the current literature are systematically based on a perfectly regular volume, usually parallelipedic in shape, without any piece of furniture or equipment in the compartment. Yet, various obstacles are normally found in real compartments and
the question is whether they affect the gravity current velocity and the level of mixing between fresh and vitiated gases. In the work reported here, gravity current prior to backdraft in compartment with obstacles is investigated by means of three-dimensional CFD numerical simulations. These simulations use as a reference case the backdraft experiment test carried out by Gojkovic (2000, Initial Backdraft. Department of Fire Safety Engineering, Lunds Tekniska Hoegskola Universitet, Report 3121). The Froude number, the transit time and the ignition time are obtained from the computations and compared to the tests in order to validate the model.
across the walls is often neglected as it requires fine grid resolution to approximate large temperature gradients. The analysis shows that this can lead to substantial errors and proposes a transient semi-analytical approximation of thermal wall behaviour. Unsteady heat diffusion equation is solved by semi-analytical approximation of a temperature profile in a wall. The temperature profile is approximated in each time step and for each wall grid node separately using the temperature at the wall internal side and Biot number at the external side for
boundary conditions. Wall heat transfer coefficient is calculated as a temperature profile derivative and then used in the coupled numerical model of the fire. The semi-analytical approach does not rely on a discretisation procedure and the accuracy of its solutions is independent of grid spacing. Therefore, such semi-analytical technique is computationally
less expensive and especially attractive for long transient calculations.
simulations show the initial gravity current, the ignition, the spreading of flame in the enclosure, the external fireball, and the subsequent decay. The Detached Eddy Simulation
(DES) approach has been used to model turbulence. In order to describe the combustion process of the mixture from the local ignition to progressive deflagration, three separate
combustion models have been implemented for laminar, low- and high-intensity turbulence flow regimes. The calculated ignition time is slightly shorter than the average ignition time observed in the experiments. The fire front progresses through the combustible mixture, generating a cloud of hot gases that are accelerated from the container into the external environment. The velocity increases up to 20 m/s. When the fire front reaches the door, combustion continues outside the enclosure as the fuel has been pushed through the door. The comparison between the calculated time history of relative pressure and the pressure sensor record shows that the numerical simulations slightly overpredict the flame front speed, with a stronger pressure pulse and higher temperatures than the observations.
The preliminary results are qualitatively promising. The simulations began at the instant at which the door opens. A stream of fresh and cold air enters the enclosure as a gravity current. In the rig, ignition was triggered by flammable conditions existing at a wire, which was constantly heated. In the CFD model the ignition time is computed automatically when flammability conditions are reached inside the enclosure, at the wire, as part of the analysis. Subsequently, the fire front is formed. The deflagration expels fuel-rich mixture into
environment, and the combustion continues outside the enclosure as a typical ‘secondary’ event. Considering that backdraft is a very complex phenomenon, the outcome is considered by the authors to be encouraging.
dh,Re) and St(dh,Re).
agreement is obtained for the comprehensive model, whereas in the other case, the conversion times are generally
underestimated and the average mass loss rates are overestimated. However, improvements in the predictive
capabilities of the solid-phase model could be achieved through the introduction of corrective factors for the external heat and mass transfer coefficients.
the model, a Volume Average Technique (VAT) has been applied to the momentum and the energy transport equations for a fluid and a solid phase to develop a specific form of porous media flow equations. The model equations have been solved with the semi-analytical Galerkin method. The detailed velocity and temperature fields in the fluid flow and the solid structure have been obtained. Using the solution fields, the whole-section drag coefficient Cd and the whole-section Nusselt number Nu have been also calculated. To validate the developed solution procedure, the results have been compared to the results of the finite volume method and to the experimental data. The comparison demonstrates an excellent agreement.
heat sink. Based on the obtained results, the Reynolds number dependencies of the scaling factors f and St Pr^0.66 have been established. On the second level, the integral model of the whole heat sink has been built using the volume averaging technique (VAT). The averaging of the transport equations leads to a closure problem. The direct model Reynolds number dependencies f and St Pr^0.66 have been used to calculate the local values of the drag coefficient and the heat transfer coefficient, which are needed in the integral model. The example calculations have been performed for 14 different pressure drops across the aluminum heat sink. The whole-section drag coefficient and Nusselt number have been calculated and compared with the experimental data. A good agreement between the modeling results and the experiment data has been reached with same discrepancies in the transitional regime. The constructed computational algorithm offers possibilities for geometry improvements and optimization, to achieve higher thermal effectiveness.
applicability of the GM to the area of heat exchangers in order to provide a reliable benchmark for numerical calculations of conjugate heat transfer problems. Using the VAT, the computational algorithm is fast-running, but still able to present a detailed picture of temperature fields in air flow as well as in the solid structure of the heat sink. The calculated whole-section drag coefficient and Nusselt number were compared with finite-volume method (FVM) results and with experimental data to verify the computational model. The comparison shows good agreement. The present results demonstrate that the selected Galerkin approach is capable to perform heat exchanger calculations where the thermal conductivity of the solid structure has to be taken into account.