A machine learning-based methodology for multi-parametric solution of chemical processes operation optimization under uncertainty

Shokry, Ahmed; Medina-González, Sergio; Baraldi, Piero; Zio, Enrico; Moulines, Éric; Espuña, Antonio

doi:10.1016/j.cej.2021.131632

Cited by 17 publications

(7 citation statements)

References 50 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Due to the uncertainty associated with the formal statistical analysis, the study adopted the contemporary and advanced machine learning technique to understand the regressors’ interactions and data prediction. , Of the popular algorithms, the DNN showed effectiveness in solving numerous regression and classification problems by capturing the patterns from the data and performing nonlinear transformations. , The model functions in a style that applies the nonlinearity at each hidden layer to obtain the abstract representation . The study employed the feedforward neural network model to predict the fructose yield, fructose selectivity, and glucose conversion responses generated by the impact of the varied MgO concentrations on CaO at different time and temperature intervals.…”

Section: Resultsmentioning

confidence: 99%

Section: ■ Experimental Sectionmentioning

confidence: 99%

See 1 more Smart Citation

MgO/CaO Nanocomposite Facilitates Economical Production of d-Fructose and d-Allulose Using Glucose and Its Response Prediction Using a DNN Model

Arumugam¹,

Singh

Mahala

et al. 2022

Ind. Eng. Chem. Res.

View full text Add to dashboard Cite

This study presents a method for the economical production of fructose and allulose (a valuable byproduct) directly from glucose over a MgO/CaO nanocomposite under an aqueous condition. The catalyst containing MgO and CaO at equal proportions helped manipulate the inherent characteristics of CaO, particularly strong basicity and surface properties. The analytical characterizations revealed that the structural assembly is such that MgO settles at the surface to initiate the isomerization reaction by providing a higher number of weak/medium base sites. The CaO present beneath undertakes the sequential conversion of the enolintermediate to ultimate fructose and byproducts (mannose and allulose). Thus, the catalyst accelerated the glucose interconversion to obtain a fructose yield as high as 33 wt % with 80% selectivity within 15 min. At the same time, it also initiated the C-3 fructose epimerization to yield allulose (a low-calorie sugar molecule). Moreover, the adopted deep neural network modeling well predicted the catalytic response with the MAE <5%. The technoeconomic analysis estimated the minimum selling price of different products to be US $ ∼4/kg (fructose), $ ∼4/kg (mannose), and $ ∼10/kg (allulose).

show abstract

Section: Resultsmentioning

confidence: 99%

Section: ■ Experimental Sectionmentioning

confidence: 99%

MgO/CaO Nanocomposite Facilitates Economical Production of d-Fructose and d-Allulose Using Glucose and Its Response Prediction Using a DNN Model

Arumugam¹,

Singh

Mahala

et al. 2022

Ind. Eng. Chem. Res.

View full text Add to dashboard Cite

show abstract

“…However, machine learning (ML) has the potential to model these complex systems due to its powerful predictive performance. To date, ML algorithms such as artificial neural network (ANN) and support vector machine (SVM) have been introduced to the property modeling study of many complex systems. − In addition, some ML models have been successfully used in different applications including solvent design, , catalyst design, , material design, , process design, , control, , and optimization . Most recently, two ML models that combine an ANN algorithm and GC method were developed to predict the phase behavior of polymer-electrolyte ATPS involving biomolecules.…”

Section: Computer-aided Atps Designmentioning

confidence: 99%

Optimal Design of Aqueous Two-Phase Systems for Biomolecule Partitioning

Chen

Kontogeorgis

2023

Ind. Eng. Chem. Res.

View full text Add to dashboard Cite

Aqueous two-phase systems (ATPS) have exhibited superior performance in many biotechnological applications. To promote the implementation of these powerful platforms by industry in the downstream processing, an optimal design method is developed to tailor high-performance ATPS for partitioning biomolecules in this work. In this design method, two machine learning (ML) models that combine the artificial neural network (ANN) algorithm and group contribution (GC) method are respectively employed to predict the phase equilibrium composition of polymer-electrolyte ATPS and the partition of biomolecules in these aqueous systems. By integrating these two ANN-GC models into the computer-aided design technique, the optimal ATPS is identified by solving an optimization-based mixed-integer non-linear programming (MINLP) problem. As a proof of concept, results of partitioning cefazolin and β-amylase are presented. In the case of cefazolin, the partitioning performance of our tailored ATPS (PPG600 + KNaSO4 + H2O) is nearly 7 times greater than that of the reported ATPS (PEG6000 + Na3C6H5O7 + H2O). Meanwhile, the ATPS of PPG600 + KNaSO4 + H2O gives a cefazolin recovery of 95.0 wt % and an agent input of 0.154 kg/kg aqueous solution, and for the ATPS of PEG6000 + Na3C6H5O7 + H2O, these values are 90.6 and 0.233, respectively. For the second case, the partition coefficient of β-amylase in our proposed ATPS (PPG400 + KNaHPO4 + H2O) is about 13.5 times higher than that of the reported ATPS (PEG10000 + KH2PO4 + H2O). In addition, the ATPS of PPG600 + KNaSO4 + H2O gives an β-amylase recovery of 97.3 wt % at a cost of 0.387 kg agent input/kg aqueous solution, and for the ATPS of PEG6000 + Na3C6H5O7 + H2O, they are 66.3 and 0.252, respectively.

show abstract

“…Deterministic approaches (e.g., artificial neural networks) have emerged in the realm of optimization because of their ability to maintain high accuracy even in the presence of complex basis functions. Nevertheless, various uncertainties in the data still exist in practical situations and are ignored by deterministic modeling approaches . The uncertainties include inherent properties (e.g., kinetic rates, heat transfer constants, blending effects, and intermolecular forces), − process operation fluctuations (e.g., recipe variations, temperatures, and equipment efficiencies), , and stream properties (e.g., density, sulfur content, and composition), as well as external uncertainties, such as resources, prices, and demand. , …”

Section: Introductionmentioning

confidence: 99%

“…Nevertheless, various uncertainties in the data still exist in practical situations and are ignored by deterministic modeling approaches. 30 The uncertainties include inherent properties (e.g., kinetic rates, heat transfer constants, blending effects, and intermolecular forces), 31−34 process operation fluctuations (e.g., recipe variations, temperatures, and equipment efficiencies), 35,36 and stream properties (e.g., density, sulfur content, and composition), 37 as well as external uncertainties, such as resources, prices, and demand. 38,39 Robust optimization, which does not require the accurate probability distribution of uncertain parameters beforehand, has been used in energy transactions, 40 unit commitment and dispatch of multi-energy systems and microgrids, 41 and optimal load dispatch of the community energy hub.…”

mentioning

confidence: 99%

Modified Hybrid Strategy Integrating Online Adjustable Oil Property Characterization and Data-Driven Robust Optimization under Uncertainty: Application in Gasoline Blending

et al. 2022

View full text Add to dashboard Cite

With the increasing challenges in environmental protection, product quality, and market profit, the development of a feasible optimization approach coupling accurate oil property online prediction has become crucial to the intelligent production of refineries. However, traditional property modeling with near-infrared (NIR) spectroscopy cannot dynamically extract the synergistic effect between wavelengths and cause inaccurate property prediction. The success rate of the data-driven robust optimization (DDRO) of the refining process is influenced by process uncertainty. Thus, we proposed a modified hybrid strategy on online NIR prediction modeling with adjustable characteristic wavelength selection and DDRO under uncertainty for the refining process. An adjustable feature space variable extraction (AFSVE) method was first proposed to dynamically select the characteristic wavelengths for constructing accurate property prediction models. A data-driven robust optimization of gasoline blending model was rendered through dual transformation using the uncertainty set of oil property derived from principal component analysis combined with robust kernel density estimation. An industrial application showed the best prediction accuracy of property models with the AFSVE method, and a high blending success rate was obtained by the proposed modified hybrid strategy, which hedges against uncertainties, including measuring error and blending effect, and boosts environmentally friendly and intelligent process operation.

show abstract

A machine learning-based methodology for multi-parametric solution of chemical processes operation optimization under uncertainty

Cited by 17 publications

References 50 publications

MgO/CaO Nanocomposite Facilitates Economical Production of d-Fructose and d-Allulose Using Glucose and Its Response Prediction Using a DNN Model

MgO/CaO Nanocomposite Facilitates Economical Production of d-Fructose and d-Allulose Using Glucose and Its Response Prediction Using a DNN Model

Optimal Design of Aqueous Two-Phase Systems for Biomolecule Partitioning

Modified Hybrid Strategy Integrating Online Adjustable Oil Property Characterization and Data-Driven Robust Optimization under Uncertainty: Application in Gasoline Blending

Contact Info

Product

Resources

About