We investigated the ability of four popular Machine Learning methods i.e.,Deep Neural Networks(DNNs),Random Forest-based regressors(RFRs),Extreme Gradient Boosting-based regressors(XGBs),and stacked ensembles of DNNs,...We investigated the ability of four popular Machine Learning methods i.e.,Deep Neural Networks(DNNs),Random Forest-based regressors(RFRs),Extreme Gradient Boosting-based regressors(XGBs),and stacked ensembles of DNNs,to model the radiative heat transfer based on view factors in bi-and polydisperse particle beds including walls.Before training and analyzing the predictive capability of each method,an adjustment of markers used in monodisperse systems,as well as an evaluation of new markers was performed.On the basis of our dataset that considers a wide range of particle radii ratios,system sizes,particle volume fractions,as well as different particle-species volume fractions,we found that(i)the addition of particle size information allows the transition from monodisperse to bi-and polydisperse beds,and(ii)the addition of particle volume fraction information as the fourth marker leads to very accurate predictions.In terms of the overall performance,DNNs and RFRs should be preferred compared to the other two options.For particle-particle view factors,DNN and RFR are on par,while for particle-wall the RFR is superior.We demonstrate that DNNs and RFRs can be built to meet or even exceed the prediction quality standards achieved in a monodisperse system.展开更多
This work focuses on implementing a particle-based method able to characterize viscoelastic materials whose rheological properties,such as storage modulus G′and loss modulus G″,are known.It is based on the bonded pa...This work focuses on implementing a particle-based method able to characterize viscoelastic materials whose rheological properties,such as storage modulus G′and loss modulus G″,are known.It is based on the bonded particle model,with the elastic constitutive relation here substituted with a viscoelastic one to capture time-scale effects.The Burgers model,vastly used in literature to model viscoelastic systems,is discretized and implemented.The test case used for calibration comprises of a cubic lattice,sheared with a periodic motion,to mimic the effect of a shear rheometer.After appropriate filtering of the stress response,the rheological properties are obtained,highlighting the effect of the lattice geometry,as well as the particle size,on the accuracy of the model.Moreover,the Burgers parameters are calibrated by analytically fitting the experimental dataset,showing the limitation of the Burgers model.The micro-contact parameters are obtained from the macro parameters through appropriate scaling.After completing a frequency sweep,the simulated G′and G″show a relatively large error,around 25%for G’for example.For this reason,a more robust model,namely the generalized Maxwell model,has been implemented.The calibration procedure is performed in the same fashion as for the Burgers model.Moreover,the tangential micro-contact parameters are scaled w.r.t.the normal ones.This scaling parameter,calledα,is calibrated by minimizing the root mean square error between simulation and experimental data,giving errors below 10%in both G′and G″for a large dataset.Additionally,a full ring plate-plate rheometer setup is simulated,and the simulation is compared with the given experimental dataset,again finding a good agreement.展开更多
The P1 approximation is a computationally efficient model for thermal radiation.Here,we present a P1 formulation in the context of the combined computational fluid dynamics and discrete element method(CFD-DEM),includi...The P1 approximation is a computationally efficient model for thermal radiation.Here,we present a P1 formulation in the context of the combined computational fluid dynamics and discrete element method(CFD-DEM),including closures for dependent scattering and coarse-graining.Using available analytical and semi-analytical solutions,we find agreement for steady-state and transient quantities in sizedisperse systems.Heat flux is identified as the most sensitive quantity to predict,displaying unphysical spatial oscillations.These oscillations are due to a temperature slip at the locations of abrupt change in solid fraction.We propose two techniques that mitigate this effect:smoothing of the radiative properties,and pseudo-scattering.Furthermore,using up to a million times enlarged particles,we demonstrate practically limitless compatibility with coarse-graining.Finally,we compare predictions made with our code to experimental data for a pebble bed under vacuum conditions,and in presence of nitrogen.We find that a carefully calibrated simulation can replicate trends observed in experiments,with relative temperature error of less than 10%.展开更多
Accurately predicting heat flux in coarse-grained CFD-DEM simulations is a significant challenge.Specifically,the rates of fluid-particle heat exchange,the effective thermal conductivity of a bed of particles,as well ...Accurately predicting heat flux in coarse-grained CFD-DEM simulations is a significant challenge.Specifically,the rates of fluid-particle heat exchange,the effective thermal conductivity of a bed of particles,as well as radiative heat transfer rates are difficult to predict.By using a novel algorithm,we significantly improve the accuracy and stability of such simulations by using a heat exchange limiter.This limiter enables realistic predictions even at time steps that are three orders of magnitude larger than a typical fluid heat relaxation time.Additionally,view-factor-based corrections for radiative heat exchange computations are developed.These corrections ensure an effective thermal bed conductivity with less than 3%error for a coarse-graining ratio of 10.The applicability of the P1 radiation model in coarse-grained settings is also examined,leading to recommendations for the CFD grid resolution to ensure accurate predictions.Our methods significantly enhance stability,accuracy,and computational efficiency,making coarse-grained CFD-DEM simulations more viable for industrial applications.These advancements enable more reliable modeling of high-temperature processes,accelerate optimization studies,and enable virtual equipment design of such processes.展开更多
文摘We investigated the ability of four popular Machine Learning methods i.e.,Deep Neural Networks(DNNs),Random Forest-based regressors(RFRs),Extreme Gradient Boosting-based regressors(XGBs),and stacked ensembles of DNNs,to model the radiative heat transfer based on view factors in bi-and polydisperse particle beds including walls.Before training and analyzing the predictive capability of each method,an adjustment of markers used in monodisperse systems,as well as an evaluation of new markers was performed.On the basis of our dataset that considers a wide range of particle radii ratios,system sizes,particle volume fractions,as well as different particle-species volume fractions,we found that(i)the addition of particle size information allows the transition from monodisperse to bi-and polydisperse beds,and(ii)the addition of particle volume fraction information as the fourth marker leads to very accurate predictions.In terms of the overall performance,DNNs and RFRs should be preferred compared to the other two options.For particle-particle view factors,DNN and RFR are on par,while for particle-wall the RFR is superior.We demonstrate that DNNs and RFRs can be built to meet or even exceed the prediction quality standards achieved in a monodisperse system.
基金funded by the EU Horizon 2020 MSCA ITN program CALIPER with grant number 812638.
文摘This work focuses on implementing a particle-based method able to characterize viscoelastic materials whose rheological properties,such as storage modulus G′and loss modulus G″,are known.It is based on the bonded particle model,with the elastic constitutive relation here substituted with a viscoelastic one to capture time-scale effects.The Burgers model,vastly used in literature to model viscoelastic systems,is discretized and implemented.The test case used for calibration comprises of a cubic lattice,sheared with a periodic motion,to mimic the effect of a shear rheometer.After appropriate filtering of the stress response,the rheological properties are obtained,highlighting the effect of the lattice geometry,as well as the particle size,on the accuracy of the model.Moreover,the Burgers parameters are calibrated by analytically fitting the experimental dataset,showing the limitation of the Burgers model.The micro-contact parameters are obtained from the macro parameters through appropriate scaling.After completing a frequency sweep,the simulated G′and G″show a relatively large error,around 25%for G’for example.For this reason,a more robust model,namely the generalized Maxwell model,has been implemented.The calibration procedure is performed in the same fashion as for the Burgers model.Moreover,the tangential micro-contact parameters are scaled w.r.t.the normal ones.This scaling parameter,calledα,is calibrated by minimizing the root mean square error between simulation and experimental data,giving errors below 10%in both G′and G″for a large dataset.Additionally,a full ring plate-plate rheometer setup is simulated,and the simulation is compared with the given experimental dataset,again finding a good agreement.
基金funded through Marie SKEODOWSKA-CURIE Innovative Training Network MATHEGRAM,the People Programme(Marie SKLODOWSKA-CURIE Actions)of the European Union's Horizon 2020 Programme H2020 under REA grant agreement No.813202.
文摘The P1 approximation is a computationally efficient model for thermal radiation.Here,we present a P1 formulation in the context of the combined computational fluid dynamics and discrete element method(CFD-DEM),including closures for dependent scattering and coarse-graining.Using available analytical and semi-analytical solutions,we find agreement for steady-state and transient quantities in sizedisperse systems.Heat flux is identified as the most sensitive quantity to predict,displaying unphysical spatial oscillations.These oscillations are due to a temperature slip at the locations of abrupt change in solid fraction.We propose two techniques that mitigate this effect:smoothing of the radiative properties,and pseudo-scattering.Furthermore,using up to a million times enlarged particles,we demonstrate practically limitless compatibility with coarse-graining.Finally,we compare predictions made with our code to experimental data for a pebble bed under vacuum conditions,and in presence of nitrogen.We find that a carefully calibrated simulation can replicate trends observed in experiments,with relative temperature error of less than 10%.
基金Rouven Weiler and Dominik Weis for their valuable insights,discussions,and feedback,which contributed to this work.We also gratefully acknowledge the financial support provided by BASF SE.
文摘Accurately predicting heat flux in coarse-grained CFD-DEM simulations is a significant challenge.Specifically,the rates of fluid-particle heat exchange,the effective thermal conductivity of a bed of particles,as well as radiative heat transfer rates are difficult to predict.By using a novel algorithm,we significantly improve the accuracy and stability of such simulations by using a heat exchange limiter.This limiter enables realistic predictions even at time steps that are three orders of magnitude larger than a typical fluid heat relaxation time.Additionally,view-factor-based corrections for radiative heat exchange computations are developed.These corrections ensure an effective thermal bed conductivity with less than 3%error for a coarse-graining ratio of 10.The applicability of the P1 radiation model in coarse-grained settings is also examined,leading to recommendations for the CFD grid resolution to ensure accurate predictions.Our methods significantly enhance stability,accuracy,and computational efficiency,making coarse-grained CFD-DEM simulations more viable for industrial applications.These advancements enable more reliable modeling of high-temperature processes,accelerate optimization studies,and enable virtual equipment design of such processes.
