Knowledge extraction in catalysis utilizing design of experiments and machine learning

MacQueen, Blake; Jayarathna, Rasika; Lauterbach, Jochen A.

doi:10.1016/j.coche.2021.100781

Cited by 15 publications

(11 citation statements)

References 52 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The design of the experiment (DOE) reduces the material and time costs of experiments. Taguchi, response surface, and factorial design are common methods for the DOE . The training of ML methods requires experimental data, which in some cases are inadequate and require traditional or high throughput experimentation ( i.e ., HTCT).…”

Section: Development Pathway Toward the Machine-learning-aided Design...mentioning

confidence: 99%

“…The training of ML methods requires experimental data, which in some cases are inadequate and require traditional or high throughput experimentation ( i.e ., HTCT). In the alternative, autonomous experimentation (AE) generates “smart data” by using automated characterization or reaction data to iterate the next set of experiments (active learning) . A venue for the screening of catalysts consists of applying ML to evaluate the importance of descriptors that influence the catalytic activity, followed by the design of the experiment and the optimization of the catalyst (Figure ).…”

Section: Development Pathway Toward the Machine-learning-aided Design...mentioning

confidence: 99%

“…Taguchi, response surface, and factorial design are common methods for the DOE. 171 The training of ML methods requires experimental data, which in some cases are inadequate and require traditional or high throughput experimentation (i.e., HTCT 172 ). In the alternative, autonomous experimentation (AE) generates "smart data" by using automated characterization or reaction data to iterate the next set of experiments (active learning).…”

Section: Development Pathway Toward Thementioning

confidence: 99%

See 2 more Smart Citations

Experimental and Computational Synergistic Design of Cu and Fe Catalysts for the Reverse Water–Gas Shift: A Review

et al. 2022

View full text Add to dashboard Cite

Strategies to capture and sequester ever-increasing anthropogenic CO2 emissions include adsorbing CO2 onto inorganic substrates and then storing it in reservoirs, changing land use to promote forestry, and converting CO2 to chemicals and fuels. The reverse water–gas shift (RWGS) reaction is a conversion strategy for producing CO from CO2 that provides the highest technology readiness level. Cu and alkali metals promote CO2 adsorption, Fe improves the thermal stability, and reducible supports like CeO2 accelerate the reaction rate. Density functional theory (DFT) is a practical modeling tool for evaluating the catalytic properties of materials at the atomic scale. The active phases of the Cu- and Fe-based catalysts, the effect of bimetallic compositions, the presence of promotors, and the influence of the support material are evaluated using observations from DFT simulations and experimental data. An optimal RWGS catalyst favors (1) CO2 adsorption, (2) the dissociation of CO2 or intermediate carbonate species to CO, and (3) CO desorption. Typically, a single-component catalytic plane is unfavorable for all these criteria, thus necessitating the design of an optimal multicomponent RWGS catalyst. Future DFT research is directed toward multifacet catalytic systems to understand the structural configuration of a highly active RWGS system. Experimental and characterization results complement DFT studies in the design of the optimal RWGS catalyst. Machine learning trained by literature data provides an automated approach for the inverse design of high-performance, stable, and economic catalysts for the RWGS reaction. This review encompasses experimental and computational approaches to understand the activity of RWGS catalysts.

show abstract

Section: Development Pathway Toward the Machine-learning-aided Design...mentioning

confidence: 99%

Section: Development Pathway Toward the Machine-learning-aided Design...mentioning

confidence: 99%

Section: Development Pathway Toward Thementioning

confidence: 99%

See 1 more Smart Citation

Experimental and Computational Synergistic Design of Cu and Fe Catalysts for the Reverse Water–Gas Shift: A Review

et al. 2022

View full text Add to dashboard Cite

show abstract

“…Data-driven catalyst design and discovery of hidden trends using machine learning (ML) engineering and/or data management are hoped to guide direct access to the goal of achieving desired performance more effectively than using conventional approaches. [1][2][3][4][5][6] In fact, the combined use of informatics techniques can suggest "unreported" areas in some cases, and can suggest unexpected areas in catalyst research. The exploration of these areas can engender new motivations for revealing important hidden characteristics in components known to have catalyst performance.…”

Section: Introductionmentioning

confidence: 99%

Leveraging machine learning engineering to uncover insights into heterogeneous catalyst design for oxidative coupling of methane

Nishimura

Ohyama

et al. 2023

Catal. Sci. Technol.

View full text Add to dashboard Cite

show abstract

“…Machine learning has had a large impact in catalyst design, where catalysts are designed from the trends and patterns found in catalyst data. − The idea of supervised machine learning is to essentially solve y = f ( x ), where y and x are described as objective and descriptor variables, respectively. − However, the challenge in machine learning lies with descriptor variables, where catalyst descriptors for predicting catalytic activities still remain a mystery. − If catalyst descriptors can be determined, one can consider that further accurate machine learning models can be designed in principle. Here, catalyst descriptors are explored with the use of physical quantities from the periodic table.…”

Section: Introductionmentioning

confidence: 99%

Designing Catalyst Descriptors for Machine Learning in Oxidative Coupling of Methane

et al. 2022

View full text Add to dashboard Cite

Catalysts descriptors for representing catalytic activities have been challenging in regard to machine learning. Machine learning and catalyst big data generated from high-throughput experiments are combined to explore the catalyst descriptors. Catalyst descriptors are designed using the physical quantities from the periodic table in the oxidative coupling of methane (OCM) reaction. Machine learning unveils the five key physical quantities representing ethylene/ethane selectivity (C2s) in the OCM reaction, where machine learning predicted three catalysts to have high C2s values. Experiments confirm that the proposed three catalysts have high C2s values in the OCM reaction. Hence, the physical quantities can be used as alternative descriptors for designing heterogeneous catalysts.

show abstract

Knowledge extraction in catalysis utilizing design of experiments and machine learning

Cited by 15 publications

References 52 publications

Experimental and Computational Synergistic Design of Cu and Fe Catalysts for the Reverse Water–Gas Shift: A Review

Experimental and Computational Synergistic Design of Cu and Fe Catalysts for the Reverse Water–Gas Shift: A Review

Leveraging machine learning engineering to uncover insights into heterogeneous catalyst design for oxidative coupling of methane

Designing Catalyst Descriptors for Machine Learning in Oxidative Coupling of Methane

Contact Info

Product

Resources

About