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.

Structure-Preserving Joint Non-negative Tensor Factorization to Identify Reaction Pathways Using Bayesian Networks.

Extracting meaningful information from spectroscopic data is key to species identification as a first step to monitoring chemical reactions in unknown complex mixtures. Spectroscopic data obtained over multiple process modes (temperature, residence time) from different sensors [Fourier transform infrared (FTIR), proton nuclear magnetic resonance (1H NMR)] comprise hidden complementary information of the underlying chemical system. This work proposes an approach to jointly capture these hidden patterns in a structure-preserving and interpretable manner using coupled non-negative tensor factorization to achieve uniqueness in decomposition. Projections onto the modes of spectral channels, specific to each sensor, are interpreted as pseudo-component spectra, while projections onto the shared process modes are interpreted as the corresponding pseudo-component concentrations across temperature and residence times. Causal structure inference among these pseudo-component spectra (using Bayesian networks) is then used to identify plausible reaction pathways among the identified species representing each pseudo-component. Tensor decomposition of the FTIR data enables the development of reaction sequences based on the identified functional groups, while that of 1H NMR by itself is lacking in mechanism development as it solely reveals the proton environments in a pseudo-component. However, jointly parsing spectra from both the sensors is seen to capture complementary information, wherein insights into the proton environment from 1H NMR disambiguate pseudo-components that have similar FTIR peaks. A scalable method of parallelizing tensor decomposition to handle high-dimensional modes in process data by using grid tensor factorization, while being robust to process data artifacts like outliers, noise, and missing data, has been developed.

Discovery of biomarker combinations that predict periodontal health or disease with high accuracy from GCF samples based on high-throughput proteomic analysis and mixed-integer linear optimization.

AIM: To identify optimal combination(s) of proteomic based biomarkers in gingival crevicular fluid (GCF) samples from chronic periodontitis (CP) and periodontally healthy individuals and validate the predictions through known and blind test sets. MATERIALS AND METHODS: GCF samples were collected from 96 CP and periodontally healthy subjects and analysed using high-performance liquid chromatography, tandem mass spectrometry and the PILOT_PROTEIN algorithm. A mixed-integer linear optimization (MILP) model was then developed to identify the optimal combination of biomarkers which could clearly distinguish a blind subject sample as healthy or diseased. RESULTS: A thorough cross-validation of the MILP model capability was performed on a training set of 55 samples and greater than 99% accuracy was consistently achieved when annotating the testing set samples as healthy or diseased. The model was then trained on all 55 samples and tested on two different blind test sets, and using an optimal combination of 7 human proteins and 3 bacterial proteins, the model was able to correctly predict 40 out of 41 healthy and diseased samples. CONCLUSIONS: The proposed large-scale proteomic analysis and MILP model led to the identification of novel combinations of biomarkers for consistent diagnosis of periodontal status with greater than 95% predictive accuracy.

Novel protein identification methods for biomarker discovery via a proteomic analysis of periodontally healthy and diseased gingival crevicular fluid samples.

AIM: To identify possible novel biomarkers in gingival crevicular fluid (GCF) samples from chronic periodontitis (CP) and periodontally healthy individuals using high-throughput proteomic analysis. MATERIALS AND METHODS: Gingival crevicular fluid samples were collected from 12 CP and 12 periodontally healthy subjects. Samples were trypically digested with trypsin, eluted using high-performance liquid chromatography, and fragmented using tandem mass spectrometry (MS/MS). MS/MS spectra were analysed using PILOT_PROTEIN to identify all unmodified proteins within the samples. RESULTS: Using the database derived from Homo sapiens taxonomy and all bacterial taxonomies, 432 human (120 new) and 30 bacterial proteins were identified. The human proteins, angiotensinogen, clusterin and thymidine phosphorylase were identified as biomarker candidates based on their high-scoring only in samples from periodontal health. Similarly, neutrophil defensin-1, carbonic anhydrase-1 and elongation factor-1 gamma were associated with CP. Candidate bacterial biomarkers include 33 kDa chaperonin, iron uptake protein A2 and phosphoenolpyruvate carboxylase (health-associated) and ribulose biphosphate carboxylase, a probable succinyl-CoA:3-ketoacid-coenzyme A transferase, or DNA-directed RNA polymerase subunit beta (CP-associated). Most of these human and bacterial proteins have not been previously evaluated as biomarkers of periodontal conditions and require further investigation. CONCLUSIONS: The proposed methods for large-scale comprehensive proteomic analysis may lead to the identification of novel biomarkers of periodontal health or disease.

Can a computer "learn" nonlinear chromatography?: Physics-based deep neural networks for simulation and optimization of chromatographic processes.

The design and optimization of chromatographic processes is essential for enabling efficient separations. To this end, hyperbolic partial differential equations (PDEs) along with nonlinear adsorption isotherms must be solved using computationally expensive numerical solvers to understand, simulate, and design the complex behavior of solute movement in chromatographic columns. In this study, physics-based artificial neural network framework for adsorption and chromatography emulation (PANACHE) is used to simulate and optimize chromatographic processes in a computationally faster and reliable manner. The proposed approach relies on learning the underlying PDEs in the form of a physics-constrained loss function to improve the accuracy of process simulations. The effectiveness of this approach is demonstrated by considering the complex dynamics of binary solute mixtures for generic pulse injections subjected to different isotherm systems, namely, the four cases of the generalized Langmuir isotherms. Unique neural network models were developed for each isotherm and the models accurately predicted the spatiotemporal concentrations of solute mixture in chromatographic columns for an arbitrary feed concentrations and injection volumes by facilitating up to 250 times computational speed-ups. Moreover, the neural network models were incorporated with process optimization routines to precisely determine the optimal injection volumes to enable baseline separation of solute components of the feed mixture.

A Comparative Theoretical and Computational Study on Robust Counterpart Optimization: III. Improving the Quality of Robust Solutions.

In this paper, we study the solution quality of robust optimization problems when they are used to approximate probabilistic constraints and propose a novel method to improve the quality. Two solution frameworks are first compared: (1) the traditional robust optimization framework which only uses the a priori probability bounds and (3) the approximation framework which uses the a posteriori probability bound. We illustrate that the traditional robust optimization method is computationally efficient but its solution is in general conservative. On the other hand, the a posteriori probability bound based method provides less conservative solution but it is computationally more difficult because a nonconvex optimization problem is solved. Based on the comparative study of the two methods, we propose a novel iterative solution framework which combines the advantage of the a priori bound and the a posteriori probability bound. The proposed method can improve the solution quality of traditional robust optimization framework without significantly increasing the computational effort. The effectiveness of the proposed method is illustrated through numerical examples and applications in planning and scheduling problems.

A Comparative Theoretical and Computational Study on Robust Counterpart Optimization: II. Probabilistic Guarantees on Constraint Satisfaction.

Probabilistic guarantees on constraint satisfaction for robust counterpart optimization are studied in this paper. The robust counterpart optimization formulations studied are derived from box, ellipsoidal, polyhedral, "interval+ellipsoidal" and "interval+polyhedral" uncertainty sets (Li, Z., Ding, R., and Floudas, C.A., A Comparative Theoretical and Computational Study on Robust Counterpart Optimization: I. Robust Linear and Robust Mixed Integer Linear Optimization, Ind. Eng. Chem. Res, 2011, 50, 10567). For those robust counterpart optimization formulations, their corresponding probability bounds on constraint satisfaction are derived for different types of uncertainty characteristic (i.e., bounded or unbounded uncertainty, with or without detailed probability distribution information). The findings of this work extend the results in the literature and provide greater flexibility for robust optimization practitioners in choosing tighter probability bounds so as to find less conservative robust solutions. Extensive numerical studies are performed to compare the tightness of the different probability bounds and the conservatism of different robust counterpart optimization formulations. Guiding rules for the selection of robust counterpart optimization models and for the determination of the size of the uncertainty set are discussed. Applications in production planning and process scheduling problems are presented.

A Comparative Theoretical and Computational Study on Robust Counterpart Optimization: I. Robust Linear Optimization and Robust Mixed Integer Linear Optimization.

Robust counterpart optimization techniques for linear optimization and mixed integer linear optimization problems are studied in this paper. Different uncertainty sets, including those studied in literature (i.e., interval set; combined interval and ellipsoidal set; combined interval and polyhedral set) and new ones (i.e., adjustable box; pure ellipsoidal; pure polyhedral; combined interval, ellipsoidal, and polyhedral set) are studied in this work and their geometric relationship is discussed. For uncertainty in the left hand side, right hand side, and objective function of the optimization problems, robust counterpart optimization formulations induced by those different uncertainty sets are derived. Numerical studies are performed to compare the solutions of the robust counterpart optimization models and applications in refinery production planning and batch process scheduling problem are presented.