Multi-output machine learning models for kinetic data evaluation : A Fischer–Tropsch synthesis case study

Chakkingal, Anoop; Janssens, Pieter; Poissonnier, Jeroen; Virginie, Mirella; Khodakov, Andreï Y.; Thybaut, Joris W.

doi:10.1016/j.cej.2022.137186

Cited by 23 publications

(14 citation statements)

References 59 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this work, three ML models (Figure S9): back-propagation neural network (BPNN), 58 cascade forward neural network (CFNN), 59 and support vector regression (SVR) 47 are trained in Matlab 2022 and compared to select the most suitable model for establishing the relationship between t m and influencing parameters. The input parameters are N, Q, μ, R, and N S , and the output parameter is t m .…”

Section: Machine Learning (Ml) Models and Model Interpretabilitymentioning

confidence: 99%

“…Fortunately, model-agnostic methods have been proposed to interpret the prediction results of ML models. Originated from the cooperative game theory, the SHAP (Shapley additive explanations) method is an effective model-agnostic method and has been widely adopted in ML model interpretability in the study of Fischer–Tropsch synthesis, separation, chemical absorption, porous carbon materials, etc. Model interpretability helps researchers screen out important factors in the research process and build trust in the prediction results of ML models.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Micromixing Performance and Residence Time Distribution in a Miniaturized Magnetic Reactor: Experimental Investigation and Machine Learning Modeling

Chen

Deng

Luo

2023

Ind. Eng. Chem. Res.

View full text Add to dashboard Cite

Miniaturization of mixers is a hot topic in process intensification, but efficient mixing at operational conditions with relatively large viscosity μ and large flow rate ratio R between two mixed solutions is still a great challenge for micromixers. Herein, a miniaturized magnetic reactor (MMR) is developed, and its micro- and macromixing performances are characterized by the Villermaux/Dushman reaction and residence time distribution experiments, respectively. The results show that the miniaturization of the premixing unit and the stirred unit is vital. By arranging the stirred units in series, the MMR is effective in reducing axial dispersion, intensifying mixing at operational conditions with large μ and R and achieving efficient energy utilization without introducing extra energy input. The micromixing time t m in the optimized MMR ranges from 0.24 to 1.42 ms when 4.72 mPa·s < μ < 73.30 mPa·s and 8 < R < 20, indicating superior micromixing performance. Moreover, machine learning models are established and compared for predicting t m with excellent accuracy and guiding mixer design, and the SHAP method is adopted for model interpretation.

show abstract

Section: Machine Learning (Ml) Models and Model Interpretabilitymentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Micromixing Performance and Residence Time Distribution in a Miniaturized Magnetic Reactor: Experimental Investigation and Machine Learning Modeling

Chen

Deng

Luo

2023

Ind. Eng. Chem. Res.

View full text Add to dashboard Cite

show abstract

“…Compared with the random forest model and support vector machine model, the prediction accuracy of the DNN model is the highest, and the average R 2 of the optimized DNN model is higher than 0.90. Thybaut et al systematically compared the prediction results of MLmethods in Fischer–Tropsch reaction models 45 . The results show that the deep belief network (DBN) model is superior to other verifying indicators 45,46 .…”

Section: Introductionmentioning

confidence: 99%

“…Thybaut et al systematically compared the prediction results of MLmethods in Fischer–Tropsch reaction models 45 . The results show that the deep belief network (DBN) model is superior to other verifying indicators 45,46 . Furthermore, multiple linear regression, support vector machines, extreme gradient boosting (XGboost), and RF models have been employed to guide the optimization of product distribution and improve the yield of target products in the catalytic conversion process 47,48 .…”

Section: Introductionmentioning

confidence: 99%

“…45 The results show that the deep belief network (DBN) model is superior to other verifying indicators. 45,46 Furthermore, multiple linear regression, support vector machines, extreme gradient boosting (XGboost), and RF models have been employed to guide the optimization of product distribution and improve the yield of target products in the catalytic conversion process. 47,48 Based on the above research background, selecting appropriate artificial intelligence modeling methods and developing ML prediction models with good accuracy will play an essential role in improving the yield of target products in the chemical process.…”

mentioning

confidence: 99%

See 1 more Smart Citation

A hybrid deep learning framework driven by data and reaction mechanism for predicting sustainable glycolic acid production performance

Zhou

Feng

et al. 2023

AIChE Journal

View full text Add to dashboard Cite

Selective oxidation at low temperatures without alkali of biomass is a promising and sustainable avenue to manufacture glycolic acid (GA), a biodegradable functional material to protect the environment. However, producing glycolic acid with high selectivity and yield using the traditional research and development approach is timeconsuming and labor-intensive. To this context, a hybrid deep learning framework driven by data and reaction mechanisms for predicting GA production was proposed, considering the lack of related reaction mechanisms in the machine learning algorithms. The proposed hybrid deep learning framework involves the kinetic reaction mechanism, catalyst properties, and reaction conditions. Results showed that the fully connected residual network exhibited superior performance (average R 2 = 0.98) for the prediction of conversion rate and product yields. Then, the multi-objective optimization and experimental verification guided the research are carried out. The experimental verification is comparable to the modeling results, with errors of less than 4% for conversion rate and GA yields. The life cycle assessment further identifies that using the optimized operating parameters, the fossil energy demand and greenhouse emissions have decreased by 2.96% and 3.00%, respectively. This work provides new insight and strategy to accelerate the engineered selective oxidation for desired GA production.

show abstract

Predicting FTS products through artificial neural network modelling

Moretta,

Grainca,

Manenti

et al. 2024

Computer Aided Chemical Engineering

View full text Add to dashboard Cite

Multi-output machine learning models for kinetic data evaluation : A Fischer–Tropsch synthesis case study

Cited by 23 publications

References 59 publications

Micromixing Performance and Residence Time Distribution in a Miniaturized Magnetic Reactor: Experimental Investigation and Machine Learning Modeling

Micromixing Performance and Residence Time Distribution in a Miniaturized Magnetic Reactor: Experimental Investigation and Machine Learning Modeling

A hybrid deep learning framework driven by data and reaction mechanism for predicting sustainable glycolic acid production performance

Predicting FTS products through artificial neural network modelling

Contact Info

Product

Resources

About