The application of multiphysical modelling is steadily increasing in the last decade, which also ... more The application of multiphysical modelling is steadily increasing in the last decade, which also leads to a corresponding increase of the complexity and of the diversity of software packages used. To deal with this complexity, users of supercomputing clusters are often challenged to couple two or more software systems of different software vendors together. However, the combined use of complex software systems usually raises additional limitations, thus reducing considerably the efficiency of the parallel simulations. In the present work, an example of such complex software utilization has been shown and the particular limitations are identified. The most severe limitation for the current supercomputing simulations has been the relatively high RAM requirement per computing core. At this stage of the numerical investigation, in order to overcome the limitations, the software packages have been ported to a different, more suitable hardware architecture with increased RAM per node. This way, the efficient use of the parallel computational resources has been guaranteed which was confirmed by means of strong scaling tests.
High Performance Computing in Science and Engineering ´16, 2016
The application of multiphysical modelling is steadily increasing in the last decade, which also ... more The application of multiphysical modelling is steadily increasing in the last decade, which also leads to a corresponding increase of the complexity and of the diversity of software packages used. To deal with this complexity, users of supercomputing clusters are often challenged to couple two or more software systems of different software vendors together. However, the combined use of complex software systems usually raises additional limitations, thus reducing considerably the efficiency of the parallel simulations. In the present work, an example of such complex software utilization has been shown and the particular limitations are identified. The most severe limitation for the current supercomputing simulations has been the relatively high RAM requirement per computing core. At this stage of the numerical investigation, in order to overcome the limitations, the software packages have been ported to a different, more suitable hardware architecture with increased RAM per node. This way, the efficient use of the parallel computational resources has been guaranteed which was confirmed by means of strong scaling tests.
High Performance Computing in Science and Engineering ‘14, 2014
The paper presents results from three-dimensional Direct Numerical Simulations of turbulent lean ... more The paper presents results from three-dimensional Direct Numerical Simulations of turbulent lean premixed hydrogen, propane and methane flames. The three fuels studied have Lewis numbers which range from small values (hydrogen) through near-unity values (methane) up to values larger than unity (propane). The computations make use of reduced chemical mechanisms with species number ranging from 9 to 28. It has been found that for the present time-explicit numerical method, the CPU-time scales non-linearly – with the power of 1.6 – with the number of chemical species.A new method to produce turbulent vortexes directly in the computational domain is presented. The vortexes are produced through localized forcing in physical space. This allows the full control of the length-scale of the vortexes as well as of their spatial and time distribution. The parallelization of the method requires no programming efforts, i.e. the parallelization follows automatically. The set of vortexes created for this particular investigation is identical for all flames studied: two methane, one hydrogen and one propane flames. Despite the identical turbulent vortex set, the flames of the different fuels react quite differently to it, exhibiting qualitatively different combustion regimes as well as different growth of the flame front area and of the consumption flame speed with time.The computations are carried out on the CRAY XE6 HERMIT supercomputer at the High Performance Computing Center Stuttgart (HLRS). Numerical issues and performance results for the three chemical mechanisms, corresponding to the three fuels used, are presented and discussed.
High Performance Computing in Science and Engineering ‘13, 2013
The paper presents results from the Direct Numerical Simulation of lean premixed hydrogen and met... more The paper presents results from the Direct Numerical Simulation of lean premixed hydrogen and methane flames. A new turbulence-generating technique, based on step-wise body forcing in physical space is introduced. It is stable, easy to implement and straight forward to use in the case of parallelization strategy based on domain decomposition.
The flow field of a turbulent jet emerging from a straight round pipe into a laminar crossflow is... more The flow field of a turbulent jet emerging from a straight round pipe into a laminar crossflow is investigated by means of large eddy simulations. The concentration of a passive scalar, introduced with the jet, is calculated in order to quantify the mixing of the jet and the crossflow. In the jet, swirl is introduced by means of body forces and a range of jet swirl numbers from S=0 up to S=0.6 is studied. The impact of the jet swirl on the flow field, on the coherent structures, and on the mixing efficiency is investigated and quantified by means of various analyses. It is found that for all swirl numbers larger than zero a clear asymmetry appears in all quantities studied. Additional to the two hanging vortices at both sides of the jet a third vortex is introduced by the swirling pipe flow which interacts with the former. This feature is described in detail as it is not mentioned in the literature. For the strongest swirl investigated a recirculation zone near the jet exit is obser...
ABSTRACT Investigation of the mixing process is one of the main issues in chemical engineering an... more ABSTRACT Investigation of the mixing process is one of the main issues in chemical engineering and combustion and the configuration of a jet into a cross-flow (JCF) is often employed for this purpose. Experimental data are gained for the symmetry plane in a JCF-arrangement of an air flow using a combination of particle image velocimetry (PIV) with laser-induced fluorescence (LIF). The experimental data with thoroughly measured boundary conditions are complemented with direct numerical simulations, which are based on idealized boundary conditions. Two similar cases are studied with a fixed jet-to-cross-flow velocity ratio of 3.5 and variable cross-flow Reynolds numbers equal to 4,120 and 8,240; in both cases the jet issues from the pipe at laminar conditions. This leads to a laminar-to-turbulent transition, which depends on the Reynolds number and occurs quicker for the case with higher Reynolds number in both experiments and simulations as well. It was found that the Reynolds number only slightly affects the jet trajectory, which in the case with the higher Reynolds number is slightly deeper. It is attributed to the changed boundary layer shape of the cross-flow. Leeward streamlines bend toward the jet and are responsible for the strong entrainment of cross-flow fluid into the jet. Velocity components are compared for the two Reynolds numbers at the leeward side at positions where strongest entrainment is present and a pressure minimum near the jet trajectory is found. The numerical simulations showed that entrainment is higher for the case with the higher Reynolds number. The latter is attributed to the earlier transition in this case. Fluid entrainment of the jet in cross-flow is more than twice stronger than for a similar flow of a jet issuing into a co-flowing stream. This comparison is made along the trajectory of the two jets at a distance of 5.5 jet diameters downstream and is based on the results from the direct numerical simulations and recently published experiments of a straight jet into a co-flow. Mixing is further studied by means of second-order statistics of the passive scalar variance and the Reynolds fluxes. Windward and leeward sides of the jet exhibit different signs for the time-averaged streamwise Reynolds flux 〈v x ′c′〉. The large coherent structures which contribute to this effect are investigated by means of timely correlated instantaneous PIV-LIF camera snapshots and their contribution to the average statistics of 〈v x ′c′〉 are discussed. The discussion on mixing capabilities of the jet in cross-flow is supported by simulation results showing the instantaneous three-dimensional coherent structures defined in terms of the pressure fluctuations.
The application of multiphysical modelling is steadily increasing in the last decade, which also ... more The application of multiphysical modelling is steadily increasing in the last decade, which also leads to a corresponding increase of the complexity and of the diversity of software packages used. To deal with this complexity, users of supercomputing clusters are often challenged to couple two or more software systems of different software vendors together. However, the combined use of complex software systems usually raises additional limitations, thus reducing considerably the efficiency of the parallel simulations. In the present work, an example of such complex software utilization has been shown and the particular limitations are identified. The most severe limitation for the current supercomputing simulations has been the relatively high RAM requirement per computing core. At this stage of the numerical investigation, in order to overcome the limitations, the software packages have been ported to a different, more suitable hardware architecture with increased RAM per node. This way, the efficient use of the parallel computational resources has been guaranteed which was confirmed by means of strong scaling tests.
High Performance Computing in Science and Engineering ´16, 2016
The application of multiphysical modelling is steadily increasing in the last decade, which also ... more The application of multiphysical modelling is steadily increasing in the last decade, which also leads to a corresponding increase of the complexity and of the diversity of software packages used. To deal with this complexity, users of supercomputing clusters are often challenged to couple two or more software systems of different software vendors together. However, the combined use of complex software systems usually raises additional limitations, thus reducing considerably the efficiency of the parallel simulations. In the present work, an example of such complex software utilization has been shown and the particular limitations are identified. The most severe limitation for the current supercomputing simulations has been the relatively high RAM requirement per computing core. At this stage of the numerical investigation, in order to overcome the limitations, the software packages have been ported to a different, more suitable hardware architecture with increased RAM per node. This way, the efficient use of the parallel computational resources has been guaranteed which was confirmed by means of strong scaling tests.
High Performance Computing in Science and Engineering ‘14, 2014
The paper presents results from three-dimensional Direct Numerical Simulations of turbulent lean ... more The paper presents results from three-dimensional Direct Numerical Simulations of turbulent lean premixed hydrogen, propane and methane flames. The three fuels studied have Lewis numbers which range from small values (hydrogen) through near-unity values (methane) up to values larger than unity (propane). The computations make use of reduced chemical mechanisms with species number ranging from 9 to 28. It has been found that for the present time-explicit numerical method, the CPU-time scales non-linearly – with the power of 1.6 – with the number of chemical species.A new method to produce turbulent vortexes directly in the computational domain is presented. The vortexes are produced through localized forcing in physical space. This allows the full control of the length-scale of the vortexes as well as of their spatial and time distribution. The parallelization of the method requires no programming efforts, i.e. the parallelization follows automatically. The set of vortexes created for this particular investigation is identical for all flames studied: two methane, one hydrogen and one propane flames. Despite the identical turbulent vortex set, the flames of the different fuels react quite differently to it, exhibiting qualitatively different combustion regimes as well as different growth of the flame front area and of the consumption flame speed with time.The computations are carried out on the CRAY XE6 HERMIT supercomputer at the High Performance Computing Center Stuttgart (HLRS). Numerical issues and performance results for the three chemical mechanisms, corresponding to the three fuels used, are presented and discussed.
High Performance Computing in Science and Engineering ‘13, 2013
The paper presents results from the Direct Numerical Simulation of lean premixed hydrogen and met... more The paper presents results from the Direct Numerical Simulation of lean premixed hydrogen and methane flames. A new turbulence-generating technique, based on step-wise body forcing in physical space is introduced. It is stable, easy to implement and straight forward to use in the case of parallelization strategy based on domain decomposition.
The flow field of a turbulent jet emerging from a straight round pipe into a laminar crossflow is... more The flow field of a turbulent jet emerging from a straight round pipe into a laminar crossflow is investigated by means of large eddy simulations. The concentration of a passive scalar, introduced with the jet, is calculated in order to quantify the mixing of the jet and the crossflow. In the jet, swirl is introduced by means of body forces and a range of jet swirl numbers from S=0 up to S=0.6 is studied. The impact of the jet swirl on the flow field, on the coherent structures, and on the mixing efficiency is investigated and quantified by means of various analyses. It is found that for all swirl numbers larger than zero a clear asymmetry appears in all quantities studied. Additional to the two hanging vortices at both sides of the jet a third vortex is introduced by the swirling pipe flow which interacts with the former. This feature is described in detail as it is not mentioned in the literature. For the strongest swirl investigated a recirculation zone near the jet exit is obser...
ABSTRACT Investigation of the mixing process is one of the main issues in chemical engineering an... more ABSTRACT Investigation of the mixing process is one of the main issues in chemical engineering and combustion and the configuration of a jet into a cross-flow (JCF) is often employed for this purpose. Experimental data are gained for the symmetry plane in a JCF-arrangement of an air flow using a combination of particle image velocimetry (PIV) with laser-induced fluorescence (LIF). The experimental data with thoroughly measured boundary conditions are complemented with direct numerical simulations, which are based on idealized boundary conditions. Two similar cases are studied with a fixed jet-to-cross-flow velocity ratio of 3.5 and variable cross-flow Reynolds numbers equal to 4,120 and 8,240; in both cases the jet issues from the pipe at laminar conditions. This leads to a laminar-to-turbulent transition, which depends on the Reynolds number and occurs quicker for the case with higher Reynolds number in both experiments and simulations as well. It was found that the Reynolds number only slightly affects the jet trajectory, which in the case with the higher Reynolds number is slightly deeper. It is attributed to the changed boundary layer shape of the cross-flow. Leeward streamlines bend toward the jet and are responsible for the strong entrainment of cross-flow fluid into the jet. Velocity components are compared for the two Reynolds numbers at the leeward side at positions where strongest entrainment is present and a pressure minimum near the jet trajectory is found. The numerical simulations showed that entrainment is higher for the case with the higher Reynolds number. The latter is attributed to the earlier transition in this case. Fluid entrainment of the jet in cross-flow is more than twice stronger than for a similar flow of a jet issuing into a co-flowing stream. This comparison is made along the trajectory of the two jets at a distance of 5.5 jet diameters downstream and is based on the results from the direct numerical simulations and recently published experiments of a straight jet into a co-flow. Mixing is further studied by means of second-order statistics of the passive scalar variance and the Reynolds fluxes. Windward and leeward sides of the jet exhibit different signs for the time-averaged streamwise Reynolds flux 〈v x ′c′〉. The large coherent structures which contribute to this effect are investigated by means of timely correlated instantaneous PIV-LIF camera snapshots and their contribution to the average statistics of 〈v x ′c′〉 are discussed. The discussion on mixing capabilities of the jet in cross-flow is supported by simulation results showing the instantaneous three-dimensional coherent structures defined in terms of the pressure fluctuations.
Uploads
Papers by Jordan Denev