Machine learning model and optimization of a PSA unit for methane-nitrogen separation

How Can (or Why Should) Process Engineering Aid the Screening and Discovery of Solid Sorbents for CO2 Capture?

ConspectusAdsorption using solid sorbents is emerging as a serious contender to amine-based liquid absorption for postcombustion CO2 capture. In the last 20+ years, significant efforts have been invested in developing adsorption processes for CO2 capture. In particular, significant efforts have been invested in developing new adsorbents for this application. These efforts have led to the generation of hundreds of thousands of (hypothetical and real) adsorbents, e.g., zeolites and metal-organic frameworks (MOFs). Identifying the right adsorbent for CO2 capture remains a challenging task. Most studies are focused on identifying adsorbents based on certain adsorption metrics. Recent studies have demonstrated that the performance of an adsorbent is intimately linked to the process in which it is deployed. Any meaningful screening should thus consider the complexity of the process. However, simulation and optimization of adsorption processes are computationally intensive, as they constitute the simultaneous propagation of heat and mass transfer fronts; the process is cyclic, and there are no straightforward design tools, thereby making large-scale process-informed screening of sorbents prohibitive.This Account discusses four papers that develop computational methods to incorporate process-based evaluation for both bottom-up (chemistry to engineering) screening problems and top-down (engineering to chemistry) inverse problems. We discuss the development of the machine-assisted adsorption process learning and emulation (MAPLE) framework, a surrogate model based on deep artificial neural networks (ANNs) that can predict process-level performance by considering both process and material inputs. The framework, which has been experimentally validated, allows for reliable, process-informed screening of large adsorbent databases. We then discuss how process engineering tools can be used beyond adsorbent screening, i.e., to estimate the practically achievable performance and cost limits of pressure vacuum swing adsorption (PVSA) processes should the ideal bespoke adsorbent be made. These studies show what conditions stand-alone PVSA processes are attractive and when they should not be considered. Finally, recent developments in physics-informed neural networks (PINNS) enable the rapid solution of complex partial differential equations, providing tools to potentially identify optimal cycle configurations. Ultimately, we provide areas where further developments are required and emphasize the need for strong collaborations between chemists and chemical engineers to move rapidly from discovery to field trials, as we do not have much time to fulfill commitments to net-zero targets.

Molecular and Artificial Neural Networks Modeling of Sorption and Diffusion of Small Alkanes, Alkenes and Their Ternary Mixtures in ZIF-8 at Different Temperatures

Numerical computations comprising of Grand Canonical-Monte Carlo (GCMC) and Canonical Statistical Ensemble (NVT) Molecular Dynamics (MD) simulations were used to study the diffusion and sorption characteristics primarily for methane, ethane, and ethene molecules and for their ternary mixtures at different temperatures in ZIF-8 porous material. Methane as pure component or in mixture proved to be the sorbed hydrocarbon with the higher molecular mobility at the temperature range of 273-373 K among alkanes and alkenes. In addtion, alkenes were the hydrocarbons with the higher self-diffusion coefficients compared to the respective alkanes. In the ternary mixtures ethane was preferentially sorbed in ZIF-8 at all temperatures studied. Direct comparisons of the self-diffusivity data obtained from the NVT-MD simulations with recently reported Magic Angle Spinning Pulsed Field Gradient Nuclear Magnetic Resonance (MAS PFG NMR) measurements showed reasonable agreement. Furthermore, the NVT-MD self-diffusivity coefficients in conjunction with the aforementioned MAS PFG NMR experimental measurements, and sorption thermodynamic data obtained from the present GCMC simulations were utilized for the development of individual Artificial Neural Networks (ANNs) predictive modeling procedures in order to provide additional quantitative and qualitative information regarding the diffusion and sorption of small alkanes, alkenes in ZIF-8. The ANNs predictions were in good agreement with the experimental measurements and with the molecular simulation data. The modeling and analysis capabilities of ANNs along with their fast computations using moderate computer resources can significantly assist the irreplaceable molecular simulation and experimental approaches to cope with complicated problems at the molecular level.

Optimization and Recovery of a Pressure Swing Adsorption Process for the Purification and Production of Bioethanol

Today, there are new technologies to produce bioethanol: one of them is the Pressure Swing Adsorption (PSA) process. This process has displaced other separation technologies due to the use of natural adsorbents and its methodology to obtain high purities with a lower energy cost. The aim of this work focuses on the optimization of the PSA process (experimental case) to obtain a higher recovery and production of bioethanol using lower energy consumption. The results are favorable since the energy cost is reduced to a range of 60% and 62%, obtaining purities above 99% wt of ethanol and recovery between 75% and 77.41%. The bioethanol produced and purified in the different scenarios meets international standards to be used as a fuel or oxygenating additive.

A Review of Numerical Research on the Pressure Swing Adsorption Process

The pressure swing adsorption (PSA) process has been considered a promising method for gas separation and purification. However, experimental methods are time-consuming, and it is difficult to obtain the detailed changes in variables in the PSA process. This review focuses on the numerical research developed to realize the modelling, optimization and control of the cyclic PSA process. A complete one-dimensional mathematical model, including adsorption bed, auxiliary devices, boundary conditions and performance indicators, is summarized as a general modelling approach. Key simplified assumptions and special treatments for energy balance are discussed for model reliability. Numerical optimization models and control strategies are reviewed for the PSA process as well. Relevant attention is given to the combination of deep-learning technology with artificial-intelligence-based optimization algorithms and advanced control strategies. Challenges to further improvements in the adsorbent database establishment, multiscale computational mass transfer model, large-scale PSA facility design, numerical computations and algorithm robustness are identified.

Machine Learning-Based Dynamic Modeling for Process Engineering Applications: A Guideline for Simulation and Prediction from Perceptron to Deep Learning

A misusage of machine learning (ML) strategies is usually observed in the process systems engineering literature. This issue is even more evident when dynamic identification is performed. The root of this problem is the gradient explode and vanishing issue related to the recurrent neural networks training. However, after the advent of deep learning, these issues were mitigated. Furthermore, the problem of data structuration is often overlooked during the machine learning model identification in this field. In this scenario, this work proposes a guideline for identifying ML models for the different applications in process systems engineering, which are usually for simulation or prediction purposes. While using the proposed guideline, the work also identifies a virtual analyzer for a pressure swing adsorption unit. In these types of adsorption separations, it is usual that the measurement of the main properties is not done online. Therefore, the virtual analyzer is another contribution of this manuscript. The overall results demonstrate that even though the test provides good performance during the ML model identification, its quality might degenerate over the application domain if the model application is overlooked.

Artificial Intelligence and Cyber-Physical Systems: A Review and Perspectives for the Future in the Chemical Industry

Modern society is living in an age of paradigm changes. In part, these changes have been driven by new technologies, which provide high performance computing capabilities that enable the creation of complex Artificial Intelligence systems. Those developments are allowing the emergence of new Cyber Systems where the continuously generated data is utilized to build Artificial Intelligence models used to perform specialized tasks within the system. While, on one hand, the isolated application of the cyber systems is becoming widespread, on the other hand, their synchronical integration with other cyber systems to build a concise and cognitive structure that can interact deeply and autonomously with a physical system is still a completely open question, only addressed in some works from a philosophical point of view. From this standpoint, the AI can play an enabling role to allow the existence of these cognitive CPSs. This review provides a look at some of the aspects that will be crucial in the development of cyber-physical systems, focusing on the application of artificial intelligence to confer cognition to the system. Topics such as control and optimization architectures and digital twins are presented as components of the CPS. It also provides a conceptual overview of the impacts that the application of these technologies might have in the chemical industry, more specifically in the purification of methane.

Performance-Based Screening of Porous Materials for Carbon Capture

Computational screening methods have changed the way new materials and processes are discovered and designed. For adsorption-based gas separations and carbon capture, recent efforts have been directed toward the development of multiscale and performance-based screening workflows where we can go from the atomistic structure of an adsorbent to its equilibrium and transport properties at different scales, and eventually to its separation performance at the process level. The objective of this work is to review the current status of this new approach, discuss its potential and impact on the field of materials screening, and highlight the challenges that limit its application. We compile and introduce all the elements required for the development, implementation, and operation of multiscale workflows, hence providing a useful practical guide and a comprehensive source of reference to the scientific communities who work in this area. Our review includes information about available materials databases, state-of-the-art molecular simulation and process modeling tools, and a complete catalogue of data and parameters that are required at each stage of the multiscale screening. We thoroughly discuss the challenges associated with data availability, consistency of the models, and reproducibility of the data and, finally, propose new directions for the future of the field.

Methodology to Solve the Multi-Objective Optimization of Acrylic Acid Production Using Neural Networks as Meta-Models

It is paramount to optimize the performance of a chemical process in order to maximize its yield and productivity and to minimize the production cost and the environmental impact. The various objectives in optimization are often in conflict, and one must determine the best compromise solution usually using a representative model of the process. However, solving first-principle models can be a computationally intensive problem, thus making model-based multi-objective optimization (MOO) a time-consuming task. In this work, a methodology to perform the multi-objective optimization for a two-reactor system for the production of acrylic acid, using artificial neural networks (ANNs) as meta-models, is proposed in an effort to reduce the computational time required to circumscribe the Pareto domain. The performance of the meta-model confirmed good agreement between the experimental data and the model-predicted values of the existent relationships between the eight decision variables and the nine performance criteria of the process. Once the meta-model was built, the Pareto domain was circumscribed based on a genetic algorithm (GA) and ranked with the net flow method (NFM). Using the ANN surrogate model, the optimization time decreased by a factor of 15.5.

Extracting Valuable Information from Big Data for Machine Learning Control: An Application for a Gas Lift Process

The present work investigated the use of an echo state network for a gas lift oil well. The main contribution is the evaluation of the network performance under conditions normally faced in a real production system: noisy measurements, unmeasurable disturbances, sluggish behavior and model mismatch. The main pursued objective was to verify if this tool is suitable to compose a predictive control scheme for the analyzed operation. A simpler model was used to train the neural network and a more accurate process model was used to generate time series for validation. The system performance was investigated with distinct sample sizes for training, test and validation procedures and prediction horizons. The performance of the designed ESN was characterized in terms of slugging, setpoint changes and unmeasurable disturbances. It was observed that the size and the dynamic content of the training set tightly affected the network performance. However, for data sets with reasonable information contents, the obtained ESN performance could be regarded as very good, even when longer prediction horizons were proposed.