Intensification of catalytic reactors: A synergic effort of Multiscale Modeling, Machine Learning and Additive Manufacturing

Bracconi, Mauro

doi:10.1016/j.cep.2022.109148

Cited by 19 publications

(14 citation statements)

References 197 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Theories about possible mechanisms have been proposed, with some authors attempting to fit ZnS:Cu-PDMS behavior into the conventional ML model (Chandra et al 2015a, Krishnan et al 2017 while others have proposed an entirely unique mechanism arising from the triboelectric effect (Sohn et al 2016). The concept of triboelectricity-induced luminescence seems promising as it explains lack of ML response from ZnS:Cu when embedded in a rigid matrix, contrary to the behavior of SrAl 2 O 4 :Eu or ZnS:Mn, and has been accepted by other researchers (Sohn et al 2016, Zhang et al 2017, 2022, Cheng et al 2019b, Wang et al 2020, Zhou et al 2023.…”

Section: Introductionmentioning

confidence: 99%

Machine-learned constitutive relations for mechanoluminescent ZnS:Cu–PDMS composites

Hoover,

Huang,

Ryu

2023

Smart Mater. Struct.

View full text Add to dashboard Cite

Materials with novel properties, such as emerging smart materials, offer a design challenge to researchers who want to make use of their unique behaviors. The complex nature of these material responses can be difficult to model from a physics-based understanding as a full description of the multi-physics, multi-scale, and non-linear phenomena requires expertise from various scientific disciplines. Some new smart materials, such as the mechanoluminescent (ML) copper-doped zinc sulfide (ZnS:Cu)-embedded in polydimethylsiloxane (PDMS) (ZnS:Cu-PDMS), lack a constitutive model or an agreement on the mechanisms of action behind the unique material properties. As constitutive equations are essential to engineer devices, with existing knowledge gap in underlying physics of smart materials, a viable approach is to use empirical data for deriving constitutive equations. However, it is challenging to derive constitutive equations on non-linear, multi-variate, and multi-physics relationship using conventional data processing approaches due to the size and complexity of the empirical data. In this work, a machine learning framework is proposed for ones to de- rive constitutive equations using empirical data for novel materials. The framework is validated by creating constitutive models for ZnS:Cu-PDMS elastomeric composites undergoing a variety of tensile load patterns. To avoid confinement of the models to the programming environment, in which they are developed, numerical fits of the machine-learned models are created as constitutive equations for the non-linear, multi-variate, and multi-physics ML properties. These models can be used when designing ML ZnS:Cu-PDMS to develop devices to harness the unique ML properties. 

show abstract

Section: Introductionmentioning

confidence: 99%

Machine-learned constitutive relations for mechanoluminescent ZnS:Cu–PDMS composites

Hoover,

Huang,

Ryu

2023

Smart Mater. Struct.

View full text Add to dashboard Cite

show abstract

“…Conventional transformation minimizes the error of transformed values MATE instead. 3 Therefore, its predictions are two orders of magnitude less accurate, as measured by the relative error. Because MATE is not a relevant measure for the application in reactor simulations, the slightly better MATE score of the conventional approach poses no considerable advantage over the latent approach.…”

Section: Modeling Co Source Termsmentioning

confidence: 99%

“…3 , O 2 , H 2 O, NO, N 2 O and N 2 , steady state surface coverages are calculated. This is done by integrating equation 11 in time until dθl /dt = 0.…”

mentioning

confidence: 99%

Efficient Neural Network Models of Chemical Kinetics Using a Latent asinh Rate Transformation

Döppel

Votsmeier

2023

Preprint

View full text Add to dashboard Cite

We propose a new modeling strategy to build efficient neural network representations of chemical kinetics. Instead of fitting the logarithm of rates, we embed the hyperbolic sine function into neural networks and fit the actual rates. We demonstrate this approach on two detailed surface mechanisms: the preferential oxidation of CO in the presence of H2 and the ammonia oxidation under industrially relevant conditions of the Ostwald process. Implementing the surrogate models into reactor simulations shows accurate results with a speed-up of 100 000. Overall, the approach proposed in this work will significantly facilitate the application of detailed mechanistic knowledge to the simulation-based design of realistic catalytic systems. We foresee that combining this approach with neural ordinary differential equations will allow building machine learning representations of chemical kinetics directly from experimental data.

show abstract

“…Computational fluid dynamics (CFD), able to properly account for the transport phenomena at the reactor and catalyst scales, is coupled with the accurate description of the chemical kinetics granted by microkinetic models or kinetic Monte Carlo (kMC) simulations . However, the CFD of reactive flows is challenging.…”

Section: Introductionmentioning

confidence: 99%

Increasing Computational Efficiency of CFD Simulations of Reactive Flows at Catalyst Surfaces through Dynamic Load Balancing

Micale,

Bracconi,

Maestri

2024

ACS Eng. Au

Self Cite

View full text Add to dashboard Cite

We propose a numerical strategy based on dynamic load balancing (DLB) aimed at enhancing the computational efficiency of multiscale CFD simulation of reactive flows at catalyst surfaces. Our approach employs DLB combined with a hybrid parallelization technique, integrating both MPI and OpenMP protocols. This results in an optimized distribution of the computational load associated with the chemistry solution across processors, thereby minimizing computational overheads. Through assessments conducted on fixed and fluidized bed reactor simulations, we demonstrated a remarkable improvement of the parallel efficiency from 19 to 87% and from 19 to 91% for the fixed and fluidized bed, respectively. Owing to this improved parallel efficiency, we observe a significant computational speed-up of 1.9 and 2.1 in the fixed and fluidized bed reactor simulations, respectively, compared to simulations without DLB. All in all, the proposed approach is able to improve the computational efficiency of multiscale CFD simulations paving the way for a more efficient exploitation of high-performance computing resources and expanding the current boundaries of feasible simulations.

show abstract

Intensification of catalytic reactors: A synergic effort of Multiscale Modeling, Machine Learning and Additive Manufacturing

Cited by 19 publications

References 197 publications

Machine-learned constitutive relations for mechanoluminescent ZnS:Cu–PDMS composites

Machine-learned constitutive relations for mechanoluminescent ZnS:Cu–PDMS composites

Efficient Neural Network Models of Chemical Kinetics Using a Latent asinh Rate Transformation

Increasing Computational Efficiency of CFD Simulations of Reactive Flows at Catalyst Surfaces through Dynamic Load Balancing

Contact Info

Product

Resources

About