Mohammad Fakhroleslam

Benchmark tables are used to justify descriptor choice, model superiority, and candidate prioritization in MOF machine learning, yet the reproducibility of those conclusions is rarely examined as training data are added.

Amine absorption remains the default for post-combustion CO2 capture, but its regeneration energy penalty motivates alternatives, and integrated adsorption-cryogenic systems are potential alternatives, since they avoid solvent degradation and deliver CO2 already in a transport-ready dense-phase. To our knowledge, no prior study has jointly optimized an industrial-scale, hydrodynamically-constrained vacuum pressure swing adsorption (VPSA) unit with a cryogenic purification stage within a rigorous plant-wide optimization framework benchmarked directly against amine absorption; this study addresses this gap. We develop a surrogate-assisted multi-objective optimization framework coupling a dynamic VPSA unit, scaled to industrial fluidization and geometric limits, with a downstream flash-based cryogenic carbon purification unit (CCPU) for CO2 capture from steam methane reforming (SMR) flue gas, achieving the required purity without a dedicated refrigeration cycle or complex distillation columns. Gaussian process regression models trained on Latin hypercube-sampled data from each unit are structurally linked and embedded in a non-dominated sorting genetic algorithm (NSGA-II) to search a ten-dimensional design space. The resulting Pareto front traces the recovery-cost trade-off, with specific energy consumption increasing rapidly as recovery exceeds 90%. Benchmarked against a literature-based monoethanolamine benchmark ($127.81/tonneCO2), a maximum-recovery design (93.06% recovery, 450.3 kWh/tonneCO2) carries a 0.1% cost premium ($127.88/tonneCO2), while a minimum-cost design (90.02% recovery, 307.4 kWh/tonneCO2) undercuts it by 17.1% at $105.90/tonneCO2. These results demonstrate that a properly optimized, industrial-scale integrated VPSA-cryogenic architecture can be economically competitive with amine absorption under the stated assumptions and system boundary, while also producing dense-phase CO2 suitable for downstream transport, a capability solvent-based capture cannot offer directly.

Post-combustion CO2 capture from steam-methane-reforming (SMR) flue gas remains challenging because dilute, low-pressure streams impose a trade-off between separation performance, energy consumption, and product CO2 quality. Integrated adsorption–cryogenic configurations provide a potential solution by combining low-regeneration-energy adsorption with downstream purification to produce dense-phase CO2. However, existing studies have generally evaluated such systems using predefined adsorption hardware and process configurations, limiting the representation of operating-condition-dependent equipment feasibility. This study develops a hydrodynamically constrained optimization and techno-economic framework for an industrial-scale integrated vacuum pressure swing adsorption (VPSA)–cryogenic carbon purification process. A process-driven sizing procedure first determines feasible VPSA train configurations, bed dimensions, and hydrodynamic limits for each candidate operating condition. A multi-objective particle swarm optimization with crowding distance (MOPSO-CD) is then coupled with dynamic VPSA and steady-state cryogenic models to optimize system recovery and specific energy consumption. The resulting VPSA product is sequentially processed in a flash-based cryogenic carbon purification unit without inter-unit recycle, external refrigeration, or downstream purification columns. Two representative designs were selected from the >90% recovery Pareto-optimal region. The maximum-recovery configuration achieved 91.74% overall CO2 recovery with a specific energy consumption of 373.8 kWh/tonne CO2 and an avoided cost of $113.64/tonne CO2, while the minimum-energy configuration reduced energy consumption to 311.6 kWh/tonne CO2 and achieved an avoided cost of $107.25/tonne CO2 at 90.03% recovery. Compared with standalone cryogenic capture, the integrated architecture reduced specific energy consumption by up to 54.9%. The proposed framework demonstrates the importance of incorporating adsorption hardware feasibility directly into process optimization for realistic large-scale CO2 capture design.

Albumin-based nanoparticles are promising drug-delivery carriers owing to their biocompatibility, biodegradability, and drug-delivery capacity. However, their design and optimization remain challenging because of complex parameter interactions and a largely empirical development landscape. The application of machine learning to albumin nanoparticle formulation is constrained by the lack of curated, and openly accessible datasets. To address this gap, we present a curated dataset of albumin-based nanoparticle formulations for drug delivery. The dataset comprises 373 unique formulations encompassing 82 mostly small-molecule drugs, extracted from 159 peer-reviewed articles. For each formulation, the dataset includes drug properties, formulation parameters, physicochemical characterization (e.g., size, polydispersity index, zeta potential, encapsulation efficiency, and drug loading), and, where available, biological and drug-release profiles. By integrating heterogeneous experimental data into a standardized, machine-readable format, this dataset provides a robust, reusable repository for feature engineering and predictive modeling. This framework aims to drive comparative nanomedicine research and accelerate machine learning.

Large scale deployment of spiral wound forward osmosis (FO) modules is hindered by intertwined hydraulic and mass transfer limitations and by the lack of computationally efficient, yet predictive, leaf-scale models for engineering applications. A multiscale quasi-3D framework is developed that resolves spacer filled channel hydraulics at moderate computational cost. Fluid flow in the feed and draw channels is described by 3D Darcy-Forchheimer porous media momentum equations, from which depth-averaged velocities are used to drive 2D convection-diffusion transport on bulk planes. Through-thickness transport across boundary films, active and support layers are captured by 1D-film and solution-diffusion ODEs, accounting for external and internal concentration polarization within a unified framework. The model is validated against experimental data from a spiral wound FO module and benchmarked against a representative surface discretized literature model reproduced in this study. The effects of leaf-scale geometry (length-width ratio, glue line length and position, channel heights) and flow configuration under normal and elevated inlet concentrations are quantified via 32 parametric studies and in-depth analysis of transport phenomena. The results show that near-square shape leaves, glue line lengths below ~0.8 m, and draw-to-feed channel height ratios around 2 minimize total pressure drop with negligible flux penalty, while cross-counter-current configuration offers the best compromise between leaf-averaged flux and hydraulic losses. In contrast, fully counter-current configurations provide only marginal additional flux at substantially higher pressure drops. The proposed framework thus enables rational screening of non-standard leaf geometries and flow arrangements and is readily extensible to RO, PRO, and other membrane-based processes.

Globally, hydrogen production relies predominantly on steam methane reforming (SMR) as its main pathway, yet it remains a substantial contributor to stationary carbon dioxide (CO2) emissions. This study evaluates an integrated adsorption-cryogenic process for large-scale CO2 capture from SMR flue gas. To address the thermodynamic inefficiencies of standalone units, a vacuum-pressure swing adsorption (VPSA) unit is evaluated through dynamic process modeling and rigorous industrial scale-up, and a cryogenic carbon purification unit (CCPU) is simulated using a steady-state framework.,Operating as a bulk pre-concentrator, a four-train parallel VPSA configuration utilizing commercial 13X zeolite is designed within hydrodynamic limits to upgrade the dilute, dry flue gas feed from 18.22 mol% to 47.09 mol% of CO2. This pre-concentration minimizes the volumetric inert gas load, substantially lowering the compression energy required in the downstream cryogenic section. Receiving this intermediate stream, the CCPU reaches final product specifications through a tightly coupled mechanical design, employing multi-stage compression, strict CO2 partial pressure regulation, and expansion turbine energy recovery.,Following sensitivity analysis, the integrated VPSA-CCPU plant achieved an overall CO2 capture rate of 90.05% and a purity of 98.25 mol%, delivered as a dense phase at 110 bar. The total specific energy consumption was 339.38 kWh/tonneCO2 (approximately 3.30 GJth/tonneCO2), which is highly competitive with traditional amine-based benchmarks (3.5-4.0 GJth/tonneCO2) encompassing the aggregate energy demand for capture, transport, and compression to 110 bar.