Modeling Of Chemical Processes Research Articles

Modelling chemical processes in explicit solvents with machine learning potentials

Solvent effects influence all stages of the chemical processes, modulating the stability of intermediates and transition states, as well as altering reaction rates and product ratios. However, accurately modelling these effects remains challenging. Here, we present a general strategy for generating reactive machine learning potentials to model chemical processes in solution. Our approach combines active learning with descriptor-based selectors and automation, enabling the construction of data-efficient training sets that span the relevant chemical and conformational space. We apply this strategy to investigate a Diels-Alder reaction in water and methanol. The generated machine learning potentials enable us to obtain reaction rates that are in agreement with experimental data and analyse the influence of these solvents on the reaction mechanism. Our strategy offers an efficient approach to the routine modelling of chemical reactions in solution, opening up avenues for studying complex chemical processes in an efficient manner.

An Interpretable Light Attention–Convolution–Gate Recurrent Unit Architecture for the Highly Accurate Modeling of Actual Chemical Dynamic Processes

To equip data-driven dynamic chemical process models with strong interpretability, we develop a light attention–convolution–gate recurrent unit (LACG) architecture with three sub-modules—a basic module, a brand-new light attention module, and a residue module—that are specially designed to learn the general dynamic behavior, transient disturbances, and other input factors of chemical processes, respectively. Combined with a hyperparameter optimization framework, Optuna, the effectiveness of the proposed LACG is tested by distributed control system data-driven modeling experiments on the discharge flowrate of an actual deethanization process. The LACG model provides significant advantages in prediction accuracy and model generalization compared with other models, including the feedforward neural network, convolution neural network, long short-term memory (LSTM), and attention-LSTM. Moreover, compared with the simulation results of a deethanization model built using Aspen Plus Dynamics V12.1, the LACG parameters are demonstrated to be interpretable, and more details on the variable interactions can be observed from the model parameters in comparison with the traditional interpretable model attention-LSTM. This contribution enriches interpretable machine learning knowledge and provides a reliable method with high accuracy for actual chemical process modeling, paving a route to intelligent manufacturing.

System simulation and parametric analysis of low-concentration coal mine gas fed SOFC-CHP with deoxygenation and enrichment pretreatment

Solid oxide fuel cell (SOFC) combined heat and power (CHP) can efficiently convert low-concentration coal mine gas (LC-CMG) into clean energy, avoiding the greenhouse effect of LC-CMG. Deoxygenation and methane enrichment pretreatment of LC-CMG can prevent explosion and anode oxidation risk. Herein, a chemical process model of LC-CMG fed SOFC-CHP system with the pretreatment was established to predict the performances under different operating conditions. Performance evaluation and exergy analysis were performed considering the use of oxygen-rich gas from the pretreatment at SOFC cathode. The effects of key operation parameters on system performances were investigated in terms of residual oxygen concentration, inlet methane concentration, fuel flow and fuel utilization. The results indicated that net electrical efficiency and heat supply efficiency can respectively reach 49.13 % and 23.90 % using LC-CMG of 15 % CH4 with enhancing effect of oxygen-rich gas at cathode. The decrease of methane concentration and the increase of residual oxygen concentration and flow rate caused the decreased electrical efficiency but the increased heat supply efficiency due to the enlargement of concentration polarization and heat supply. The optimal operation parameters were recognized by comprehensively considering diverse performance indicators. This work can guide system integration and operation optimization of LC-CMG fed SOFC-CHP.

All‐nonlinear static‐dynamic neural networks versus Bayesian machine learning for data‐driven modelling of chemical processes

AbstractIn recent decades, the utilization of machine learning (ML) and artificial intelligence (AI) approaches have been explored for process modelling applications. However, different types of ML models may have contrasting advantages and disadvantages, which become critical during the optimal selection of a specific data‐driven model for a particular application as well as estimation of parameters during model training. This paper compares and contrasts two different types of data‐driven modelling approaches, namely the series/parallel all‐nonlinear static‐dynamic neural network models and models from a Bayesian ML approach. Both types of AI modelling approaches considered in this work have shown to significantly outperform several state‐of‐the‐art steady‐state and dynamic data‐driven modelling techniques for various performance measures, specifically, model sparsity, predictive capabilities, and computational expense. The performances of the proposed model structures and algorithms have been evaluated for two nonlinear dynamic chemical engineering systems—a plug‐flow reactor for vapour phase cracking of acetone for production of acetic anhydride and a pilot‐plant for post‐combustion CO2 capture using monoethanolamine as the solvent. For the validation data from the CO2 capture pilot plant, root mean squared error (RMSE) for flue gas outlet temperature, flowrate and CO2 concentration is 0.05%, 1.07%, and 5.0%, respectively, for the all‐nonlinear static‐dynamic neural networks and 0.1%, 1.75%, and 14.14%, respectively, for the Bayesian ML models. For the plug flow reactor data, the Bayesian ML models yield superior RMSE compared to the all‐nonlinear static‐dynamic neural networks when the measurement data are corrupted with Gaussian, auto‐correlated, or cross‐correlated noise.

Reliable identification based intelligent PID tuning for long-period process control under different working conditions

Background: Chemical process models change frequently with operating conditions. However, the industry mostly adopts an off-line calibration and online fixed value commissioning scheme, which can be difficult to maintain control quality. Therefore, a long-cycle online intelligent PID controller calibration scheme is necessary.Methods: Firstly, a high-order linear dynamic model is used as the identification model, and online recursive identification technology is utilized. Secondly, a comprehensive performance index is selected to ensure both stability and rapidity of dynamic transitions. Finally, the slow rate updated PID parameters are obtained during the process control, and the Levy Memory Particle Swarm Optimization (LMPSO) search is applied to harvest the optimal solution.Significant findings: This paper has presented an intelligent PID tuning method based on reliable identification technology. The main contributions include: (i) By using Lyapunov theory, the boundedness of identification error using the proposed algorithm is guaranteed; (ii) The LMPSO algorithm is proposed to enhance global search capability and search efficiency; (iii) A novel optimization scheme is proposed for a full-process online closed-loop intelligent PID controller. The scheme aims to improve the industrial controller's performance without altering its structure.

Cryogenic Process Optimization for Natural Gas Purification: Predictive Modeling of Methane–CO2 Solid–Vapor Phase Equilibrium Using Response Surface Methodology

Cryogenic Process Optimization for Natural Gas Purification: Predictive Modeling of Methane–CO₂ Solid–Vapor Phase Equilibrium Using Response Surface Methodology

Termolecular Eley–Rideal pathway for catalytic oxidation of nitric oxide on [Pt2]0,± dimers using O2

AbstractA comprehensive density functional theory (DFT) investigation of the catalytic oxidation of NO to NO2 on neutral and charged [Pt2]0,± dimers has been considered employing M06L/def2TZVP level of theory. Single as well as co‐adsorption energies of NO and O2 molecules suggest that the traditional Langmuir Hinshelwood (LH) mechanism and less explored termolecular Eley–Rideal (TER) and termolecular Langmuir Hinshelwood (TLH) mechanisms, are suitable for a full catalytic reaction pathway in which two NO molecules are converted to two NO2 molecules, initiated by an activated O2 molecule. Activation barrier reveals that the TER mechanism is found to be more reliable in converting two NO molecules into two NO2 molecules on [Pt2]¯ dimer. In addition to shedding light on the intrinsic characteristics of Pt2 dimers, the study will serve as a benchmark for investigating the oxidation process of NO to NO2 utilizing models of termolecular chemical processes.

UNIFAC PREDICTION OF VAPOR LIQUID EQUILIBRIA INVOLVING GAMMA-VALEROLACTONE DERIVATIVE SYSTEMS

Researchers have studied the hydrogenation of Gamma-Valerolactone (GVL) over Ru/C to produce 2-methyltetrahydrofuran (2-MTHF), a biomass-based platform chemical with potential as a biofuel and green solvent. Other byproducts of this reaction include 2-butanol (2-BuOH), 2-pentanol (2-PeOH), and 1,4-pentanediol (1,4-PDO). In this study, UNIFAC activity coefficient models were used to predict the vapor-liquid equilibrium of several systems included in the process to explain the phase behavior which is important in purification process. The accuracy of the UNIFAC model is tested by comparing the experimental boiling points for the binary system of 2-propanol + 1-butanol and the vapor-liquid equi-librium of the GVL + 2-MTHF system. From this comparison, the Root Mean Square Deviation (RMSD) values for the boiling point measurements are obtained to be 3.153%, and for the liquid and vapor phases in the vapor-liquid equilibrium measurements, the RMSD values are 1.934% and 0.298% respectively. These RMSD values indicate the level of accuracy of the UNIFAC model in representing the experimental data for both boiling point measurements and vapor-liquid equilibrium phase behavior. The prediction results of vapor-liquid equilibrium data for GVL derivative systems showed that the 2-BuOH + 2-MTHF system does form an azeotrope when the mol fraction of 2-BuOH is 0.0031 at 79.780C. The calculation was performed using ChemCAD commercial software for chemical process modeling and simulation

Experimentally validated predictive PI-PD control strategy for delay-dominant chemical processes

Controlling delay dominant chemical processes is a challenging control problem. Hence, a new method for designing a Smith predictor (SP)-based proportional-integral proportional-derivative (PI-PD) control strategy is proposed in this study. The inner-loop PD parameters are determined using the phase margin (PM) and maximum sensitivity (Ms) specifications. To determine the outer-loop PI parameters, the moment matching technique is augmented with maximum sensitivity considerations. The performance and robustness of this design is compared with that of its contemporaries using four benchmark chemical process models including that of a stirred tank reactor and a heat exchanger. The resilience of this control strategy amid uncertainties in process parameters is studied and performance improvement achieved over contemporary works is quantified. Finally, the proposed design is also experimentally validated on a two-tank level control loop.

MALTA: A Zonally Averaged Global Atmospheric Transport Model for Long‐Lived Trace Gases

AbstractWe present a two‐dimensional, zonally averaged global model of atmospheric transport named MALTA: Model of Averaged in Longitude Transport in the Atmosphere. It aims to be accessible to a broad community of users, with the primary function of quantifying emissions of greenhouse gases and ozone depleting substances. The model transport is derived from meteorological reanalysis data and flux‐gradient experiments using a three‐dimensional transport model. Atmospheric sinks are prescribed loss frequency fields. The zonally averaged model simulates important large‐scale transport features such as the influence on trace gas concentrations of the quasi‐biennial oscillation and variations in inter‐hemispheric transport rates. Stratosphere‐troposphere exchange is comparable to a three‐dimensional model and inter‐hemispheric transport is faster by up to 0.3 years than typical transport times of three‐dimensional models, depending on the metric used. Validation of the model shows that it can estimate emissions of CFC‐11 from an incorrect a priori emissions field well using three‐dimensional (3D) mole fraction fields generated using a different 3D model than which the flux gradient relationships were derived. The model is open source and is expected to be applicable to a wide range of studies requiring a fast, simple model of atmospheric transport and chemical processes for estimating associated emissions or mole fractions.

Adaptive soft sensor modeling of chemical processes based on an improved just‐in‐time learning and random mapping partial least squares

AbstractThe just‐in‐time learning‐based partial least squares (JIT‐PLS) has been extensively applied to adaptive soft sensor modeling of complex nonlinear processes. However, it still has the problems of unreasonable relevant samples selection and unsatisfactory local modeling. Aiming at these problems, this paper proposes an improved just‐in‐time learning‐based random mapping partial least squares (IJIT‐RMPLS), including an improved relevant samples selection strategy and a random mapping PLS (RMPLS) model. On the one hand, considering the different correlation degrees between input variables and output variable, this method applies mutual information to evaluate the importance of each input variable and designs a variable‐weighted Euclidean distance to select relevant samples for local modeling. On the other hand, in order to prompt the prediction precision of local soft sensor models, this method combines the idea of nonlinear random mapping in extreme learning machines with PLS and builds a RMPLS with multiple activation functions. Applications on a numerical example and a real chemical process show that the proposed IJIT‐RMPLS has smaller prediction error compared with traditional JIT‐PLS.

Language models can identify enzymatic binding sites in protein sequences

Recent advances in language modeling have had a tremendous impact on how we handle sequential data in science. Language architectures have emerged as a hotbed of innovation and creativity in natural language processing over the last decade, and have since gained prominence in modeling proteins and chemical processes, elucidating structural relationships from textual/sequential data. Surprisingly, some of these relationships refer to three-dimensional structural features, raising important questions on the dimensionality of the information encoded within sequential data. Here, we demonstrate that the unsupervised use of a language model architecture to a language representation of bio-catalyzed chemical reactions can capture the signal at the base of the substrate-binding site atomic interactions. This allows us to identify the three-dimensional binding site position in unknown protein sequences. The language representation comprises a reaction-simplified molecular-input line-entry system (SMILES) for substrate and products, and amino acid sequence information for the enzyme. This approach can recover, with no supervision, 52.13% of the binding site when considering co-crystallized substrate-enzyme structures as ground truth, vastly outperforming other attention-based models.

Моделирование установки замедленного коксования в программе MatLab

In order to prevent the occurrence of an emergency situation in the oil and gas industry, it is necessary to use control, automation and alarm systems. In this regard, artificial intelligence technologies have recently become in-creasingly popular. Neural networks are of particular interest. To implement the task of regulating, automating and predicting technological processes in the oil industry, it is possible to use neural network modeling of chemical and technological processes. Examples of using neural network modeling in practice are given. The results of neural network modeling of a delayed coking plant of one of the operating enterprises are presented. A delayed coking installation was modeled in the UniSim Design software package, which allowed us to obtain an initial data array for a neural network. The neural network was built in the MatLab program, and the program code was created. Graphs of error and regression are presented. The analysis of the results presented on the regression and error graphs is given. As a result of testing the model, minimal discrepancies between experimental and predicted data were obtained, which indicates the adequacy of the neural network model. An additional test of the program was also performed. The results of training and testing of the model are presented. The results obtained can later be used to create programs at different levels of control, since the model allows you to estimate the amount of losses during the operation of the installation at a certain flow rate of feed water supplied to the installation as a turbocharger.

Optimization-based multi-source transfer learning for modeling of nonlinear processes

This work develops an optimization-based multi-source transfer learning scheme for modeling of nonlinear chemical processes. Specifically, a transfer learning neural network model for a target process with limited data is developed using the pre-trained model obtained with multiple source processes. Since the performance of transfer learning models depends on the quality of the pre-trained models, we propose a novel Bayesian optimization problem to optimize the selection of multi-source data for the pre-trained models by first deriving a generalization error bound for multi-source domain adaptation using Y-discrepancy distance. Subsequently, the optimization problem is formulated using the theoretical error bound to select the optimal set of multiple sources, which can provide a good initial guess of the weight parameters for transfer learning model. Finally, a simulation study of a chemical reactor process in Aspen Plus Dynamics is conducted to illustrate the effectiveness of the optimization-based multi-source transfer learning scheme.

Mathematical modeling of the creation process and aging of polymer composite materials

Based on the analysis of the literature on the possibility of using neural networks to create new materials with high functional properties, the article considers a solution to the problem of determining the operational stability of polymeric composite materials by creating physical and chemically sound mathematical prediction models. Epoxy resins of the UP-637 and EA brands with an isophorone diamine hardener were chosen as the matrix of the model composite material, and oligobutadiene rubber of the SKN-10 KTR brand was chosen as the modifier. It justifies directions of work necessary for development of new materials creation methodology with optimal characteristics, building a model for changing the properties of materials at variation of composition and implementation of full-scale mathematical modeling of physical and chemical processes of polymer composite materials aging at changing level and time of climatic factors influence. Verification of the obtained dependence of service characteristics on the composition of the material and the level of influencing climatic factors was carried out on the basis of data from full-scale tests in a temperate climate. The proposed methodology for modelling the properties of polymer composite materials will reduce the development time of new materials and allow creation of polymer composites based on epoxy resin containing fillers of various natures (carbon, mineral and polymer) with high performance parameters.

NEURAL NETWORK MODELING OF PLASMA-CHEMICAL PROCESSES OF OBTAINING NANOSYSTEMS

Plasma-chemical technology is a new area of industrial chemical technology. Chemical processes in low-temperature plasma, the regularities of reactions, and the basis of plasma-chemical technology will require computer simulation. Building adequate simulation models of plasma-chemical processes for developing nanosystems and computer simulations with them allows the development of applied research studies in this subject area. Computer intelligence technologies provide an opportunity to use non-classical approaches to building mathematical models of chemical processes. Neural network technologies make it possible to create mathematical simulation models of various processes.

Process structure-based fully connected neural network for the modelling of chemical processes: A comparison between global and modular configurations

BackgroundChemical process modeling is essential for the optimization, precise control and prediction of processes performance. As the modeling based on the first principles is time-consuming and labor-intensive, utilizing neural networks for the process modeling was proposed. However, the poor interpretability of the neural network models is may a problem in the modelling of chemical processes where the accuracy and reliability of the model are highly required. MethodInspired by the functional partitioning of neural networks in human brain, in this paper, global fully connected neural network (GFNN) and modular fully connected neural network (MFNN) models are constructed for the production process of ethylbenzene and that of isopropanol with cycle process, respectively. For the chemical processes with complex cycle processes, the quality of the product can be easily affected by small feed changes due to the highly nonlinear model which is called snowball effect. To avoid snowball effect, the cycle-avoidance MFNN is constructed, and it shows better modelling performance than GFNN. Significant findingsMFNN model has more advantages than GFNN model in the modelling of large-scale chemical processes without cycle processes. In addition, it is clear that MFNN model has a partial interpretability which means the local physical meaning for different parts of the whole process.

Robust reduced-order machine learning modeling of high-dimensional nonlinear processes using noisy data

Autoencoder-based reduced-order machine learning models have been developed for modeling and predictive control of nonlinear chemical processes with high dimensionality such as discretization of reaction–diffusion processes. However, in the presence of data noise, autoencoders may over-fit the training data and subsequently learn an inaccurate low-dimensional representation of the process variables. This leads to an inaccurate prediction model when the models are integrated with model predictive control (MPC). To address this issue, this work develops a novel machine-learning-based reduced-order modeling method by integrating SpectralDense layers into autoencoders and incorporating them with recurrent neural networks. We demonstrate that the new architecture of autoencoders using SpectralDense layers is more robust against over-fitting than conventional autoencoders in the presence of data noise, which improves the prediction accuracy in MPC. A diffusion–reaction process simulation example is used to demonstrate that the robust autoencoders outperform those using conventional layers for reduced-order modeling in predictive control.

A new Correlation-Similarity Conjoint Algorithm for developing Encoder-Decoder based deep learning multi-step prediction model of chemical process

As an important procedure for the multi-step modeling of chemical process, the feature selection approach for unequal-length process variable series is frequently studied based on statistical analysis rather than chemical mechanism. As such, in this contribution, a convolution-based Correlation-Similarity Conjoint Algorithm (CSCA) considering chemical process mechanism including underlying variable interactions is developed to perform correlation analyses between unequal-length variable series, so that the most correlated features with minimum redundancy can be detected for the multi-step prediction modeling of variable series. Aided by CSCA, the chemical encoder-decoder based deep learning models are built for multi-step predicting flowrates, pressures and temperatures in the actual industrial process. The proposed CSCA proves to be effective in series feature selection, promoting the developments of the multi-step prediction encoder-decoder based deep learning model of different process variables, which are of great value for prospective model predictive control and real-time optimization in chemical engineering.

Family of Skeletal Reaction Mechanisms for Methane–Oxygen Combustion in Rocket Propulsion

Analyzing methane–oxygen rocket propellant combinations requires suitable modeling of the major chemical reaction processes. Although several detailed kinetic mechanisms for methane oxidation in air exist, most do not reproduce the reaction pathways of high-pressure methane–oxygen combustion, typical of liquid rocket engines. Moreover, when large-scale computational fluid dynamics simulations are pursued, detailed reaction schemes are not computationally viable. In the present study, we identify a reliable detailed kinetic scheme for liquid rocket applications, and then we perform a wide reduction campaign leveraging computational singular perturbation theory. Enforcing various reduction targets, we obtain a family of seven skeletal schemes, including 11–39 species. Each mechanism targets different combustion modes, namely, homogeneous ignition, complex flows and flame extinction, premixed burning, reaction processes under intense turbulent mixing, and largely off-stoichiometric mixtures, typical of rocket engine preburners. We test the skeletal mechanisms against meaningful validation targets, attaining appreciable predictive accuracy compared with the detailed parent scheme. We expect the proposed family of skeletal schemes to offer a wide and flexible range of solutions—in terms of size, accuracy, and dominant combustion mode—for performing large-scale yet cost-affordable computational fluid dynamics of methane–oxygen flames under rocket-engine-relevant conditions.