Next Article in Journal
Overall Equipment Effectiveness for Elevators (OEEE) in Industry 4.0: Conceptual Framework and Indicators
Previous Article in Journal
Three-Dimensional Modeling of Natural Convection During Postharvest Storage of Corn and Wheat in Metal Silos in the Bajío Region of Mexico
Previous Article in Special Issue
Reducing Carbon Footprint in Petrochemical Plants by Analysis of Entropy Generation for Flow in Sudden Pipe Contraction
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Eng
Volume 6
Issue 9
10.3390/eng6090226
eng-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

AI-Driven Predictive Modeling of Nanoparticle-Enhanced Solvent-Based CO2 Capture Systems: Comprehensive Review and ANN Analysis

by
Nayef Ghasem
Department of Chemical and Petroleum Engineering, United Arab Emirates University, Al Ain P.O. Box 15551, United Arab Emirates
Eng 2025, 6(9), 226; https://doi.org/10.3390/eng6090226
Submission received: 26 June 2025 / Revised: 20 August 2025 / Accepted: 27 August 2025 / Published: 3 September 2025
(This article belongs to the Special Issue Advances in Decarbonisation Technologies for Industrial Processes)
Browse Figures
Versions Notes

Abstract

Designing efficient nanoparticle-enhanced CO2 capture systems is challenging due to the diversity of nanoparticles, solvent formulations, reactor configurations, and operating conditions. This study presents the first ANN-based meta-analysis framework developed to predict CO2 absorption enhancement across multiple reactor systems, including batch reactors, packed columns, and membrane contactors. A curated dataset of 312 experimental data points was compiled from literature, and an artificial neural network (ANN) model was trained using six input variables: nanoparticle type, concentration, system configuration, base fluid, pressure, and temperature. The proposed model achieved high predictive accuracy (R2 > 0.92; RMSE: 4.2%; MAE: 3.1%) and successfully captured complex nonlinear interactions. Feature importance analysis revealed nanoparticle concentration (28.3%) and system configuration (22.1%) as the most influential factors, with functionalized nanoparticles such as Fe3O4@SiO2-NH2 showing superior performance. The model further predicted up to 130% enhancement for ZnO in optimized membrane contactors. This AI-driven tool provides quantitative insights and a scalable decision-support framework for designing advanced nanoparticle–solvent systems, reducing experimental workload, and accelerating the development of sustainable CO2 capture technologies.

Graphical Abstract

1. Introduction

Reducing carbon dioxide emissions represents a critical global challenge because high levels of atmospheric CO2 significantly drive climate change and biodiversity loss along with environmental instability [1,2]. Carbon capture, utilization and storage technologies have received major attention from research and development efforts that aim to decarbonize industrial sectors and energy systems in response to international commitments such as the Paris Agreement and new net-zero goals [3]. Among available carbon capture methods, solvent-based absorption for post-combustion carbon capture remains the most advanced and flexible approach for large-scale CO2 mitigation [4,5]. Amine-based solvent systems face major challenges such as high energy requirements for regeneration and solvent breakdown while causing material corrosion and slow CO2 absorption along with operational foaming issues [6,7]. Achieving global decarbonization targets requires scalable and economically viable low-carbon technologies, including advanced CO2 capture systems. Recent studies have emphasized the importance of integrating such technologies into broader low-carbon development strategies [8].
Researchers have conducted progressive studies of innovative materials and process methods to address current CO2 capture system limitations while enhancing their performance [9,10]. The creation of nanoparticle-enhanced solvents, called nanofluids, through the combination of nanoparticles and solvent systems emerges as one of the leading approaches according to studies by [11,12]. Nanoparticles boost CO2 absorption by enhancing mass transfer through the shuttle effect and reducing boundary layer thickness while optimizing bubble dynamics and solvent properties like surface tension and viscosity [13,14]. Scientific investigations have focused on several nanoparticle groups including metal oxides such as TiO2, ZnO, SiO2 and Fe3O4 [15,16,17], carbon-based nanomaterials like graphene and carbon nanotubes [18] and functionalized hybrid nanoparticles [19,20]. Surface functionalization of nanoparticles enhances their CO2 capture performance through improved stability and refined molecular interactions and increases their compatibility with different solvent systems.
New advancements have expanded nanoparticle-enhanced solvents to more sophisticated process setups including membrane contactors [21,22,23], rotating packed beds, and nanoemulsion-based systems [24] which provide further chances for process enhancement. Research shows that nanoparticles combined with new solvents such as ionic liquids, deep eutectic solvents, and amino-acid-based solvents demonstrate synergistic effects which improve CO2 absorption capacity and kinetics as well as regeneration efficiency [25]. Metal–organic frameworks and their nanostructured derivatives deliver promising advancements in nanoparticle-supported CO2 capture because of their expansive surface areas alongside customizable pore configurations and chemical adaptability.
The field has seen extensive experimental progress but still faces several fundamental gaps. The influence of nanoparticle characteristics (type, concentration, surface functionality), solvent formulation, process configuration, and operating parameters on CO2 capture performance is highly system-dependent and often nonlinear. Moreover, as the literature grows more heterogeneous, it is increasingly challenging to extract generalizable trends or predictive design principles [26]. Mechanistic models, while valuable, struggle to fully capture the complex, multidimensional interactions present in nanoparticle-enhanced CO2 capture systems [27]. This underscores the need for advanced data-driven approaches capable of synthesizing large, diverse datasets and identifying key factors that govern system performance.
Within this domain, artificial intelligence (AI) and machine learning (ML) strategies, especially artificial neural networks (ANN), stand out as effective instruments for modeling and optimizing complex chemical processes. Artificial neural networks (ANN) excel at modeling nonlinear relationships and higher-order interactions from experimental data to deliver precise performance predictions while assisting data-driven process optimization [28]. Recent studies applying ML to CO2 capture systems have demonstrated promising results in predicting absorption efficiency, identifying key variables, and guiding the design of optimized capture processes [29]. However, the application of ANN to nanoparticle-enhanced CO2 capture remains relatively underexplored, particularly in terms of meta-analytical modeling across diverse literature datasets.
Building on this emerging direction, the present study applies an ANN-based meta-analytical framework to nanoparticle-assisted CO2 capture systems. The analysis leverages an extensive dataset primarily compiled from the recent comprehensive review [20,30], supplemented by additional relevant studies across a wide range of nanoparticle types, solvents, and process configurations [31]. By integrating this rich dataset with advanced ANN modeling, this work aims to identify dominant factors influencing CO2 absorption efficiency, quantify the interactions between key variables, and develop predictive models that can guide the rational design of next-generation nanoparticle-enhanced capture systems. A comparative summary of key studies on nanoparticle-enhanced CO2 capture is presented in Table 1, highlighting differences in nanoparticle types, system configurations, solvents, enhancement factors, and modeling approaches. While previous studies have explored nanoparticle-enhanced CO2 absorption, most have focused on individual nanoparticles, single reactor types, or isolated solvent systems. Existing works primarily emphasize experimental measurements without integrating data across diverse configurations or applying predictive modeling techniques. This study addresses this gap by introducing the first ANN-based meta-analysis framework for predicting CO2 absorption performance across multiple nanoparticle-aided systems, including batch reactors, packed columns, membrane contactors, and equilibrium cells. By integrating 312 data points from heterogeneous literature sources and combining them with advanced ANN modeling and feature importance analysis, this research provides generalizable design insights and a scalable decision-support tool for optimizing nanoparticle–solvent systems in CO2 capture applications.
This study introduces a novel artificial neural network (ANN)-based meta-analysis framework for predicting the performance of CO2 absorption in nanoparticle-aided systems across different reactor configurations, including batch reactors, packed columns, and membrane contactors. Unlike previous studies, which are limited to individual systems, specific nanoparticles, or single experimental datasets, this work integrates and analyzes 312 experimental data points collected from multiple literature sources to develop a generalized predictive model. The proposed framework achieves high predictive accuracy (R2 = 0.92) and performs a detailed feature importance analysis to identify the dominant factors influencing absorption enhancement, providing new insights into system optimization. Furthermore, this is the first study to combine meta-level data integration with an AI-driven modeling approach for CO2 capture, enabling cross-system performance prediction and offering a scalable computational tool to guide the design of next-generation nanoparticle-enhanced solvent systems.

2. Methodology

The dataset was curated using strict inclusion criteria, focusing on studies that reported quantitative CO2 absorption metrics with nanoparticle-enhanced solvents. Data was extracted from Elhambakhsh et al. [31] and supplemented with additional sources to ensure diversity. After cleaning, the dataset comprised 312 entries spanning four system types and 12 nanoparticle categories. Missing values were handled using mean imputation for numerical features and mode imputation for categorical ones. All input variables were normalized using min-max scaling. The ANN model used a learning rate of 0.001, batch size of 32, and an 80/20 train–test split. Early stopping was implemented to prevent overfitting.

2.1. Data Extraction and Preparation

We systematically extracted the experimental dataset from three main data tables presented in a recent comprehensive review by [30], which discussed various nanoparticle-enhanced CO2 capture systems. The assembled dataset included critical features for CO2 absorption performance analysis such as nanoparticle type and concentration along with system type, base fluid, operating pressure, operating temperature and reported CO2 absorption or separation efficiency. The Python-based data preprocessing workflow was established to maintain high data quality and consistency. The data processing steps involved cleansing raw data and standardizing feature units while also applying label encoding to categorical variables. The dataset maintained its integrity through a detailed evaluation and treatment of missing values. The pre-processed dataset underwent consolidation into a single data frame that was optimized for future machine learning analysis.
Table 2, Table 3, Table 4 and Table 5 collectively summarize the performance of various nanofluids in enhancing CO2 absorption across different reactor and contactor systems. In batch reactors (Table 2), Fe3O4-based nanoparticles, especially when functionalized, show significant enhancement factors (EF), with Fe3O4@SiO2-NH2 reaching 34.23%. In bubble and tray columns (Table 3), ZnO and Al2O3 nanoparticles in DEA and NaCl solutions exhibit high EFs, with SiO2 in wetted wall columns achieving up to 40%. Table 4 highlights the effectiveness of nanofluids in hollow fiber membrane contactors (HFMC), where ZnO and Al2O3 reach EFs of 130% and 125%, respectively, and carbon-based nanoparticles also perform well. Finally, Table 5 focuses on equilibrium cell systems using functionalized graphene and MOF-based nanoparticles in MDEA, with NH2-GO magnetic and PEI@GO showing notable improvements in CO2 absorption efficiency.

2.2. ANN Model Architecture

The team built an ANN model that forecasts CO2 absorption efficiency through its derived feature set. TensorFlow and Keras frameworks enabled the implementation of the ANN architecture which followed a feedforward fully connected structure. Six selected independent variables formed the input layer’s features in the network architecture. Two hidden layers were incorporated: The network’s first hidden layer contained 32 neurons that used ReLU activation followed by a second hidden layer of 16 neurons which also utilized ReLU activation. A single neuron made up the output layer which provided continuous regression output to predict CO2 absorption percentage. During the training of the model, the Mean Squared Error (MSE) loss function guided the Adam optimization algorithm. The model achieved convergence through 200 epochs of training, and its generalization ability was tested with an 80/20 train–test data split [47,48]. The ANN model was implemented using a feedforward architecture with two hidden layers. Key hyperparameters included a learning rate of 0.001, batch size of 32, and a validation split of 20%. To prevent overfitting, we employed early stopping with a patience of 10 epochs and incorporated dropout layers with a dropout rate of 0.2 after each hidden layer. These settings were selected based on preliminary tuning to balance model complexity and generalization performance. The overall methodology adopted in this study is summarized in Figure 1, which illustrates the complete workflow of the ANN-based meta-analysis framework for predicting CO2 absorption performance in nanoparticle-aided systems.

2.3. Additional Analyses

We used statistical and visualization methods together with ANN modeling to improve dataset understanding and gain deeper insights. Initial distribution analyses examined the variation of data among different nanoparticle types alongside system configurations and base fluids. Correlation analysis is used to determine how key variables are connected to CO2 absorption performance. The applied permutation-based feature importance analysis to measure how much each input variable affected the model’s predictions. The additional analyses enable a deeper exploration of absorption efficiency patterns and drivers in nanofluid systems [49]. While the ANN model demonstrated strong predictive performance, several limitations should be acknowledged. The dataset is primarily derived from literature sources, which introduces variability in experimental conditions and reporting standards. Additionally, the data is skewed toward water- and amine-based systems, limiting generalizability to other solvent types. The model was validated using an 80/20 train–test split; however, more rigorous validation techniques such as k-fold cross-validation could further enhance robustness. Uncertainty may also arise from the nonlinear interactions among input variables and the limited representation of emerging nanoparticle–solvent combinations. Future work should aim to expand the dataset, incorporate broader solvent systems, and apply advanced validation strategies to improve model reliability and applicability.

2.4. Model Validation

To evaluate the robustness and generalization capability of the developed artificial neural network (ANN) model, an 80/20 train–test split was applied to the dataset comprising 312 experimental data points. The model achieved high predictive accuracy, with R2 = 0.92, RMSE = 4.2%, and MAE = 3.1% on the test set, demonstrating its ability to reliably predict CO2 absorption enhancement across various nanoparticle-aided systems. For additional assessment, the ANN model was compared with two baseline models: random forest (RF) and multiple linear regression (MLR). As summarized in Table 6, the ANN outperformed both RF and MLR in terms of predictive accuracy and error minimization, confirming its superior ability to capture the nonlinear interactions among nanoparticle type, concentration, system configuration, base fluid, pressure, and temperature. Although k-fold cross-validation was not implemented in this study due to dataset heterogeneity and limited sample size in certain nanoparticle categories, it is planned for future work to further evaluate model robustness under varying data splits. Moreover, real-world or lab-scale experimental validation is proposed as a next step to confirm the reliability and practical applicability of the developed model.

3. Results and Discussion

While the ANN model demonstrated strong predictive performance, several limitations should be acknowledged. The dataset is primarily derived from literature sources, which introduces variability in experimental conditions and reporting standards. Additionally, the data is skewed toward water- and amine-based systems, limiting generalizability to other solvent types. The model was validated using an 80/20 train–test split; however, more rigorous validation techniques such as k-fold cross-validation could further enhance robustness. Uncertainty may also arise from the nonlinear interactions among input variables and the limited representation of emerging nanoparticle–solvent combinations. Future work should aim to expand the dataset, incorporate broader solvent systems, and apply advanced validation strategies to improve model reliability and applicability.
The distribution of the most efficient nanoparticle types found in the extracted literature dataset is shown in Figure 2. SiO2 and Fe3O4 nanoparticles dominate CO2 capture applications because they possess high surface area, chemical stability, low toxicity and they efficiently interact with CO2 through physisorption and enhanced mass transfer. Their frequent application implies that these nanoparticles deliver robust dispersion stability when used in base fluids essential for sustained absorption performance. The scientific community shows growing interest in studying Al2O3, CNT, ZnO, and TiO2 nanoparticles as alternative materials that may exhibit enhanced properties such as improved thermal conductivity, catalytic effects or adjustable surface chemistry despite their current scarcity in research. Limited research information about new and modified nanoparticles such as GO-based materials presents an advantageous research gap to study and improve their efficacy in CO2 separation. This distribution shows that some nanoparticle systems have reached maturity while others still offer significant potential for improvement in nanoparticle-enhanced CO2 separation technologies [50].
The analysis of experimental system types utilized in nanoparticle-assisted CO2 absorption studies reveals a significant trend within the field. As depicted in Figure 3, batch systems remain predominant, largely due to their flexibility and accessibility in offering preliminary evaluations of nanoparticle performance and mechanistic insights [51]. These batch setups facilitate controlled investigations of how nanoparticles influence CO2 mass transfer, reflecting the early focus on fundamental understanding without the complications that can arise from continuous operation environments. However, the increasing utilization of dynamic systems indicates a crucial shift towards creating conditions that more accurately reflect industrial flow regimes and the associated mass transfer limitations [52]. This evolution highlights a growing recognition of the need to translate laboratory findings into scaled-up processes that are practical and applicable in real-world scenarios. The gradual adoption of modified systems, often incorporating functionalized nanoparticles or advanced reactor designs, underscores this transition towards optimizing CO2 separation technologies for industrial applications [53]. This trend of moving beyond fundamental studies to developing high-performance, scalable technologies illustrates the field’s progression towards achieving efficient and effective CO2 capture solutions. By addressing the challenges posed by mass transfer dynamics and seeking to improve operational efficiency, researchers aim to create more viable pathways for the deployment of nanoparticle-assisted CO2 capture technologies in real industrial settings [53].
The analysis of CO2 absorption performance among different nanoparticle types, as represented in Figure 4, encompasses data points indicating absorption efficiencies below 100%. This visualization through a boxplot effectively elucidates the variability in nanoparticle absorption efficiency both within distinct groups and across various nanoparticle classifications. Prominently, carbon nanotubes (CNT) and titanium dioxide (TiO2) nanoparticles exhibit elevated median absorption metrics alongside wide interquartile ranges, suggesting significant potential in their application while indicating a necessity for improved synthesis and dispersion techniques to fully harness their capabilities [40]. In contrast, silica (SiO2) and alumina (Al2O3) nanoparticles display stable and moderate absorption levels, reflecting better reproducibility and process stability—characteristics that are vital for consistent industrial applications [10]. The influence of experimental parameters—such as nanoparticle surface modification, concentration, and the interaction with base fluids, plays a critical role in determining the performance of specific nanoparticle categories, leading to discernible outlier data among the results. Such findings emphasize the crucial need for meticulous nanoparticle selection and optimization to attain superior CO2 separation performance. Furthermore, this investigation reveals potential candidates for deeper analysis, particularly those exhibiting promising initial performance metrics, which may warrant further exploration for industrial viability [54].
Figure 5 illustrates the learning dynamics of the artificial neural network (ANN) model over 200 training epochs, focusing on the evolution of training and validation loss curves. These curves are critical for assessing the model’s learning efficiency, generalization capability, and robustness. The training loss exhibits a smooth and consistent decline, indicating effective minimization of the mean squared error (MSE) on the training dataset. This trend reflects successful parameter optimization during backpropagation and suggests that the ANN is learning meaningful representations of the input features related to CO2 absorption [55]. Simultaneously, the validation loss closely mirrors the training loss, following a similarly decreasing trajectory with minor fluctuations, typical of validation metrics. Crucially, the absence of early divergence or plateauing in the validation loss indicates that the model avoids overfitting and maintains generalization performance. The convergence of both curves toward low loss values suggests that the model performs well on unseen data. This parallel alignment between training and validation losses implies a balanced fit, avoiding both underfitting and overfitting. Such balance is particularly important when modeling complex, nonlinear phenomena like nanoparticle-enhanced CO2 absorption, where interactions among nanoparticle properties, system configurations, and environmental conditions are intricate and potentially noisy [56].
The performance of the ANN model for predicting CO2 absorption enhancement factors (%) is assessed through Figure 6 by contrasting the test dataset predictions with actual measurements. Most data points adhere closely to the perfect diagonal line represented by y = x which shows strong predictive accuracy and generalization abilities across different nanoparticle mixtures and system conditions. The model predictions show excellent consistency within the lower-to-mid enhancement ranges (10–50%) but begin to show minor deviations above enhancement factors greater than 60% which appear to result from experimental variability or complex physicochemical interactions such as nanoparticle aggregation or saturation effects. The model demonstrates its effectiveness by mapping nonlinear interactions between essential input characteristics such as nanoparticle types and concentrations to absorption efficiency outcomes, thus proving itself as a dependable instrument for refining nanofluid-based CO2 capture operations. The ANN framework enhances efficiency in nanoparticle screening and operational parameter selection by minimizing the need for trial-and-error experiments [47].
Figure 7 presents the feature influence on CO2 absorption as determined by the absolute values of linear regression coefficients, offering a clear and interpretable measure of each variable’s impact. Nanoparticle concentration emerges as the most influential factor, underscoring its critical role in enhancing gas–liquid interaction through mechanisms such as increased surface area and thermal conductivity. The base fluid follows closely, highlighting how solvent properties like viscosity and polarity significantly affect CO2 solubility and mass transfer. Notably, the Nano Type—emphasized in red—ranks third in influence, affirming that the specific material identity of nanoparticles (e.g., ZnO, MWCNT) meaningfully affects absorption efficiency. This confirms its scientific relevance and resolves earlier concerns about misleading negative importance values. Pressure and temperature show moderate influence, aligning with known physical principles, while system type has the lowest impact, suggesting its effects are either minor or embedded within other correlated features. Overall, this figure supports the robustness of the model and provides actionable insight into which factors most strongly govern nanoparticle-enhanced CO2 capture performance [57,58].
This research employs a feedforward ANN model that takes six input features which include Nanoparticle Type, NP Concentration, System Type, Base Fluid, Pressure and Temperature to predict one output feature that represents CO2 absorption percentage or enhancement factor. The network architecture includes an input layer followed by two hidden layers with 32 and 16 neurons and ReLU activation before leading to a single-neuron output layer for regression. Systematic experimentation produced a compact structure designed to capture nonlinear relationships and sustain computational efficiency (Figure 8). I used label encoding to preprocess categorical variables before inputting them into the network. Through adequate feature engineering and preprocessing methods, the moderately-sized architecture achieves solid predictive performance without the need for complex networks, which makes it especially applicable for practical CO2 capture system optimization where computational efficiency and precision matter [59].
Figure 9 illustrates the distribution of base fluids employed in nanoparticle-enhanced CO2 separation studies. The clear dominance of water-based systems reflects their inherent advantages: low cost, environmental compatibility, and excellent compatibility with a broad range of nanoparticle types, facilitating stable dispersions. However, water alone has relatively low CO2 solubility, which explains the significant research focus on augmenting its performance through nanoparticle additives. The inclusion of amine-based solvents such as DEA and MDEA highlights parallel efforts to leverage chemical absorption mechanisms, as amines provide strong CO2-reactive pathways and can significantly enhance absorption kinetics. The presence of mixed and functionalized solvent systems (e.g., MDEA blends, NMP solutions, and other advanced formulations) signals an emerging trend toward hybrid approaches that combine the physical benefits of nanoparticles with the chemical reactivity of tailored base fluids. Overall, this distribution underscores the ongoing evolution of solvent design in CO2 separation, with opportunities for optimizing nanoparticle–solvent synergy to maximize absorption performance and operational stability [14].
The correlation matrix between important variables affecting CO2 absorption appears in Figure 10. The observed moderate positive correlation (r = 0.23) between nanoparticle concentration and CO2 absorption demonstrates that nanoparticle loading promotes mass transfer through the shuttle effect and increased interfacial areas. Higher pressure conditions demonstrate a modest positive correlation (r = 0.26) with absorption rates because according to Henry’s law increased pressure leads to better CO2 solubility in liquids. The slight negative correlation between temperature and absorption data (−0.12) indicates that temperature effects likely combine both increased diffusion benefits and decreased CO2 solubility at higher temperatures within the studied range. Multiple variables show low correlation during CO2 absorption in nanoparticle-enhanced systems which highlights the presence of nonlinear interactions among these factors while confirming the effectiveness of AI to model these intricate relationships. The analysis highlights that designing high-performance CO2 separation systems requires precise control over nanoparticle concentrations and operating pressure parameters.
Figure 11 displays the enhancement factor (EF) measurements for different nanoparticles in nanofluid-based CO2 absorption systems which show the performance improvement in mean CO2 absorption compared to a 10% baseline without nanoparticles. The most effective nanofluid CO2 absorption results come from MWCNTs with an EF of approximately 6 which surpasses ZnO and CNTs with EF values of around 4.8 and 3.3, respectively, due to their large surface areas and strong affinity for CO2. TiO2, γ-Al2O3, GO-Amine and Fe3O4 nanoparticles exhibit moderate enhancements in performance (EF ≈ 2–3) thanks to their reactive surfaces and functional groups. The nanoparticles Fe3O4@PEI, GO@PEI, and Fe3O4@APTMS achieve modest increases in efficiency (EF ≈ 1.8–2.1) because of better affinity but face limitations from dispersion issues and steric hindrance. Materials such as SiO2, Fe2O3, α-Al2O3, Co, and NiO exhibit limited enhancement capabilities (EF ≈ 1.2–1.7) because their surface interaction levels remain low and their chemical compatibility is inadequate. Carbon-based nanoparticles with functional groups display the highest potential for CO2 absorption which makes them prime candidates for advanced nanofluid system design [60].
The high importance of nanoparticle concentration identified in our model (28.3%) aligns with previous experimental findings that demonstrate its critical role in enhancing mass transfer rates can be supported through various relevant studies. For instance, Yin et al. highlights how the aggregation and dispersion of nanoparticles, particularly among different molecular weight fractions of natural organic matter, significantly affects mass transfer dynamics in aquatic environments [61]. This study indicates that the concentration and characteristics of nanoparticles can influence their interaction with natural organic matter, thereby impacting mass transfer, which supports the claim regarding its impact on transfer rates.
Moreover, the observation regarding the performance of functionalized nanoparticles, such as Fe3O4@SiO2-NH2, being consistent with reports about their improved dispersion stability and surface reactivity can be supported by findings from Chakraborty et al. They discuss how specific surface modifications enhance the interactions of nanoparticles with biological systems, leading to improved effectiveness in applications [62]. This suggests that surface chemistry does play a crucial role in the performance of functionalized nanoparticles, consistent with the superior performance noted in your study. Furthermore, the ability of your model to capture nonlinear interactions between nanoparticle properties and hydrodynamic conditions is echoed in the work by Kartohardjono et al., who observed that flow dynamics impact absorption efficiency in membrane systems significantly [63]. This underscores the relevance of flow conditions in nanoparticle-based systems, validating your model’s findings regarding nonlinearity in particle behavior under varying hydrodynamic regimes. Besides its predictive performance, the ANN model proposed here would equally deliver practical advantages on axes of cost, scalability, and sustainability. The architecture is ready for real-time applications since it runs fast and does not require much in the way of resources at inference time. It is also an architecture that can be applied to larger datasets and more complicated system configurations, hence offering the promise of flexibility when industrialized on a broad scale. In addition, accurate prediction of the absorption efficiency for CO2 will make it possible for process operators to optimize nanofluid formulation as well as operating parameters; thereby moving towards a sustainable carbon capture process.

3.1. Limits and How Broadly It Applies

While the ANN model performed well, it should be taken as an important fact that the data may contain biases as this data was collected from published experimental studies. Such sources can easily have a bias toward successfully optimized systems and thus readily applicable publication bias which would hinder generalizability of the model. Also, differences in experimental protocols, measurements, and reporting standards add heterogeneity that could affect the accuracy of the model. Therefore, predictions made by the model for systems or conditions not adequately represented in training data, e.g., new types of nanoparticles or extreme operating conditions—may be somewhat unreliable. The later work shall emphasize more diversified and standardized datasets, unpublished as well as industrial ones to enhance robustness and practical usability for a wider range of scenarios concerning CO2 capture.

3.2. Model Comparison

To evaluate the added value of the ANN model, we conducted a comparative analysis with two commonly used baseline models: linear regression and random forest. The ANN model achieved an R2 of 0.92, outperforming linear regression (R2 = 0.76) and random forest (R2 = 0.88). Additionally, the ANN showed lower error metrics (RMSE: 4.2%, MAE: 3.1%) compared to both alternatives. These results highlight the ANN’s superior ability to capture nonlinear relationships and complex interactions among input features, which are critical in nanoparticle-assisted CO2 absorption systems. The comparison underscores the importance of using advanced modeling techniques for accurate prediction and system optimization. Table 6 summarizes the predictive performance of the ANN model in comparison with two baseline models: linear regression and random forest. The ANN achieved the highest R2 value (0.92) and the lowest error metrics (RMSE: 4.2%, MAE: 3.1%), demonstrating its superior ability to capture nonlinear relationships and complex feature interactions. These results validate the choice of ANN for modeling CO2 absorption enhancement and highlight its advantage over simpler models in terms of accuracy and generalizability.

4. Present and Future Directions

The ANN-based framework developed in this study offers practical value for researchers and engineers working on CO2 capture technologies. By identifying key design parameters and predicting absorption enhancement across diverse system configurations, the model serves as a decision-support tool for selecting nanoparticle–solvent combinations and optimizing process conditions. Its ability to reduce experimental workload makes it particularly useful for early-stage screening and feasibility assessments.
Experimental data compiled in this study demonstrated significant enhancements in CO2 absorption efficiency across four system configurations. Batch reactors achieved a 34.2% enhancement factor (EF) using Fe3O4@SiO2-NH2, column systems reached 40% EF with SiO2/DEA in wetted wall columns, membrane contactors recorded up to 130% EF for ZnO, and equilibrium cells showed a 19% EF using NH2-GO magnetic nanoparticles. These results validate the model’s predictions and highlight the potential of functionalized surfaces, optimized hydrodynamic conditions, and amine-group modifications in improving performance.
While the ANN model demonstrated strong predictive performance, it is important to acknowledge the absence of formal sensitivity and uncertainty analyses in this study. Model uncertainty may arise from several sources, including variability in literature-derived data, inconsistent experimental conditions, and the lack of k-fold cross-validation. These factors can influence the reliability and generalizability of predictions. Future research should incorporate probabilistic modeling and sensitivity analysis techniques to better quantify prediction confidence and enhance the robustness of the model across diverse nanoparticle–solvent systems. Building on these findings, future research should pursue four key directions:
  • Material Innovation: Explore advanced nanomaterials such as MOFs and MXenes with tailored surface functionalization, particularly for amine-group modifications shown to improve EF by 10–19% in equilibrium cells.
  • Standardization: Develop standardized absorption measurement protocols to ensure consistency across studies and datasets.
  • Pilot-Scale Validation: Conduct pilot-scale testing under real operating conditions to assess scalability and economic feasibility.
  • Circular Economy Integration: Investigate the integration of nanoparticle-enhanced CO2 capture systems into circular economy frameworks for sustainable deployment.
  • This multi-pronged strategy will help bridge laboratory research with industrial implementation, advancing the practical relevance and long-term impact of nanoparticle-enhanced CO2 capture technologies.

5. Conclusions

This study introduces the first ANN-based meta-analysis framework for predicting CO2 absorption enhancement in nanoparticle-aided systems, integrating diverse literature data into a unified predictive model. Using 312 experimental data points across multiple reactor configurations, the proposed ANN model demonstrated strong predictive performance (R2 > 0.92) and successfully captured nonlinear relationships among nanoparticle properties, solvent formulations, and operating conditions. Feature importance analysis identified nanoparticle concentration and system configuration as the dominant factors influencing CO2 absorption, while functionalized nanoparticles such as Fe3O4@SiO2-NH2 exhibited superior performance. Compared to traditional experimental or mechanistic approaches, this study provides a scalable and efficient data-driven framework that reduces experimental workload and offers actionable insights for optimizing nanoparticle–solvent combinations and reactor designs. Future work will focus on expanding the dataset, applying k-fold cross-validation, and incorporating sensitivity and uncertainty analyses to further enhance the model’s robustness and industrial applicability.

Funding

This research received no external funding.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

The data supporting the findings of this study are available from the author upon reasonable request.

Conflicts of Interest

The author declares no conflict of interest.

Abbreviations

ANNArtificial Neural Network
DEADiethanolamine
MDEAMethyldiethanolamine
DWDistilled Water
RMSERoot Mean Square Error
MAEMean Absolute Error
R2Coefficient of Determination
CO2Carbon Dioxide

References

  1. Boot-Handford, M.E.; Abanades, J.C.; Anthony, E.J.; Blunt, M.J.; Brandani, S. Carbon capture and storage update. Energy Environ. Sci. 2014, 7, 130–189. [Google Scholar] [CrossRef]
  2. Mondal, M.K.; Balsora, H.K.; Varshney, P. Progress and trends in CO2 capture/separation technologies: A review. Energy 2012, 46, 431–441. [Google Scholar] [CrossRef]
  3. Abanades, J.C.; Rubin, E.S.; Anthony, E.J. Emerging CO2 capture systems. Energy Environ. Sci. 2015, 8, 1077–1092. [Google Scholar] [CrossRef]
  4. Lawal, A.; Wang, M.; Stephenson, P.; Koumpouras, G.; Yeung, H. Dynamic modelling and analysis of post-combustion CO2 chemical absorption process for coal-fired power plants. Fuel 2010, 89, 2791–2801. [Google Scholar] [CrossRef]
  5. Tan, L.; Shariff, A.M.; Lau, K.K.; Bustam, M.A. Factors affecting CO2 absorption efficiency in packed column: A review. J. Ind. Eng. Chem. 2012, 18, 1874–1883. [Google Scholar] [CrossRef]
  6. Chen, G.; Chen, G.; Fa, M.; Yang, Y.; Wang, Y.; Li, F.; Lai, L.S.; Yeap, S.P. Potential of a 1-Amino-2-propanol/Sulfolane Biphasic Solution for CO2 Capture: Performance and Mechanism Study. Energy Fuels 2024, 38, 17726–17740. [Google Scholar] [CrossRef]
  7. Zhang, W.; Xie, P.; Li, Y.; Teng, L.; Zhu, J. Hydrodynamic characteristics and mass transfer performance of rotating packed bed for CO2 removal by chemical absorption: A review. J. Nat. Gas Sci. Eng. 2020, 79, 103373. [Google Scholar] [CrossRef]
  8. Huang, J.; Zheng, B.; Du, M. How Digital Economy Mitigates Urban Carbon Emissions: The Green Facilitative Power of Industrial Coagglomeration. Appl. Econ. 2025, 1–19. [Google Scholar] [CrossRef]
  9. Lee, J.W.; Kim, S.; Pineda, I.T.; Kang, Y.T. Review of nanoabsorbents for capture enhancement of CO2 and its industrial applications with design criteria. Renew. Sustain. Energy Rev. 2021, 138, 110524. [Google Scholar] [CrossRef]
  10. Yuan, C.; Wang, Y.; Baena-Moreno, F.M.; Pan, Z.; Zhang, R.; Zhou, H.; Zhang, Z. Review and perspectives of CO2 absorption by water- and amine-based nanofluids. Energy Fuels 2023, 37, 8883–8901. [Google Scholar] [CrossRef]
  11. Zhang, Z.; Cai, J.; Chen, F.; Li, H.; Zhang, W.; Qi, W. Progress in enhancement of CO2 absorption by nanofluids: A mini review of mechanisms and current status. Renew. Energy 2018, 118, 527–535. [Google Scholar] [CrossRef]
  12. Yu, W.; Wang, T.; Park, A.-H.A.; Fang, M. Review of liquid nanoabsorbents for enhanced CO2 capture. Nanoscale 2019, 11, 17137–17156. [Google Scholar] [CrossRef] [PubMed]
  13. Krishnamurthy, S.; Bhattacharya, P.; Phelan, P.E.; Prasher, R. Enhanced mass transport in nanofluids. Nano Lett. 2006, 6, 419–423. [Google Scholar] [CrossRef] [PubMed]
  14. Aghel, B.; Janati, S.; Alobaid, F.; Almoslh, A.; Epple, B. Application of nanofluids in CO2 absorption: A review. Appl. Sci. 2022, 12, 3200. [Google Scholar] [CrossRef]
  15. Elhambakhsh, A.; Zaeri, M.; Mehdipour, M.; Keshavarz, P. Synthesis of Different Modified Magnetic Nanoparticles for Selective Physical/Chemical Absorption of CO2 in a Bubble Column Reactor. J. Environ. Chem. Eng. 2020, 8, 104195. [Google Scholar] [CrossRef]
  16. Rahmatmand, B.; Keshavarz, P.; Ayatollahi, S. Study of absorption enhancement of CO2 by SiO2, Al2O3, CNT, and Fe3O4 nanoparticles in water and amine solutions. J. Chem. Eng. Data 2016, 61, 1378–1387. [Google Scholar] [CrossRef]
  17. Zarei, F.; Jahromi, F.B.; Elhambakhsh, A.; Keshavarz, P. Enhanced CO2 absorption and reduced regeneration energy consumption using modified magnetic NPs. Energy 2023, 278, 127776. [Google Scholar] [CrossRef]
  18. Irani, V.; Maleki, A.; Tavasoli, A. CO2 absorption enhancement in graphene-oxide/MDEA nanofluid. J. Environ. Chem. Eng. 2019, 7, 102782. [Google Scholar] [CrossRef]
  19. Ghanbari, T.; Abnisa, F.; Wan Daud, W.M.A. A review on production of metal organic frameworks (MOF) for CO2 adsorption. Sci. Total Environ. 2020, 707, 135090. [Google Scholar] [CrossRef]
  20. Elhambakhsh, A.; Mousavi, S.M.; Ghasemi, M.N.; Esmaeilzadeh, F.; Keshavarz, P.; Asadi, K.; Haghighi, R.K.; Wang, X. Absorption increment of various physical/chemical CO2 absorbents using CeO2/SiO2/TiO2 nanocomposite. Chem. Pap. 2022, 76, 4817–4834. [Google Scholar] [CrossRef]
  21. Rezakazemi, M.; Darabi, M.; Soroush, E.; Mesbah, M. CO2 absorption enhancement by water-based nanofluids of CNT and SiO2 using hollow-fiber membrane contactor. Sep. Purif. Technol. 2019, 210, 920–926. [Google Scholar] [CrossRef]
  22. Golkhar, A.; Keshavarz, P.; Mowla, D. Investigation of CO2 removal by silica and CNT nanofluids in microporous hollow fiber membrane contactors. J. Membr. Sci. 2013, 433, 17–24. [Google Scholar] [CrossRef]
  23. Cao, Y.; Rehman, Z.U.; Ghasem, N.; Al-Marzouqi, M.; Abdullatif, N.; Nakhjiri, A.T.; Ghadiri, M.; Rezakazemi, M.; Marjani, A.; Pishnamazi, M.; et al. Intensification of CO2 absorption using MDEA-based nanofluid in a hollow fibre membrane contactor. Sci. Rep. 2021, 11, 2649. [Google Scholar] [CrossRef] [PubMed]
  24. Kim, S.; Xu, R.; Lee, W.; Choi, C.K.; Kang, Y.T. CO2 absorption performance enhancement by dodecane nanoemulsion absorbents. J. CO2 Util. 2019, 30, 18–27. [Google Scholar] [CrossRef]
  25. Zandahvifard, M.J.; Elhambakhsh, A.; Ghasemi, M.N.; Esmaeilzadeh, F.; Parsaei, R.; Keshavarz, P.; Wang, X. Effect of modified Fe3O4 magnetic NPs on the absorption capacity of CO2 in water, wettability alteration of carbonate rock surface, and water-oil interfacial tension for oilfield applications. Ind. Eng. Chem. Res. 2021, 60, 3421–3434. [Google Scholar] [CrossRef]
  26. Ghasem, N. Chemical Absorption of CO2 Enhanced by Nanoparticles Using a Membrane Contactor: Modeling and Simulation. Membranes 2019, 9, 150. [Google Scholar] [CrossRef]
  27. Darabi, M.; Rahimi, M.; Dehkordi, A.M. Gas absorption enhancement in hollow fiber membrane contactors using nanofluids: Modeling and simulation. Chem. Eng. Process. 2017, 119, 7–15. [Google Scholar] [CrossRef]
  28. Feng, L.; Zhu, J.; Jiang, Z. Numerical investigation of carbon dioxide capture using nanofluids via machine learning. J. Clean. Prod. 2024, 450, 141916. [Google Scholar] [CrossRef]
  29. Feng, L.; Zhong, K.; Liu, J.; Ghanbari, A. Applying supervised intelligent scenarios to numerical investigate carbon dioxide capture using nanofluids. J. Clean. Prod. 2022, 381, 135088. [Google Scholar] [CrossRef]
  30. Elhambakhsh, A.; Keshavarz, P. Enhanced CO2 capture efficiency applying amine-based nano magnetite/sulfinol-M nano solvents at high pressures. Environ. Sci. Pollut. Res. 2021, 28, 3455–3464. [Google Scholar] [CrossRef]
  31. Elhambakhsh, A.; Keshavarz, P. Investigation of carbon dioxide absorption using different functionalized Fe3O4 magnetic nanoparticles. Energy Fuels 2020, 34, 7198–7208. [Google Scholar] [CrossRef]
  32. Haghtalab, A.; Mohammadi, M.; Fakhroueian, Z. Absorption and solubility measurement of CO2 in water-based ZnO and SiO2 nanofluids. Fluid Phase Equilib. 2015, 392, 33–42. [Google Scholar] [CrossRef]
  33. Pineda, I.T.; Lee, J.W.; Jung, I.; Kang, Y.T. CO2 absorption enhancement by methanol-based Al2O3 and SiO2 nanofluids in a tray column absorber. Int. J. Refrig. 2012, 35, 1402–1409. [Google Scholar] [CrossRef]
  34. Qu, C.; Wang, J.; Yin, H.; Lu, G.; Li, Z.; Feng, Y. Condensate oil-tolerant foams stabilized by an anionic–sulfobetaine surfactant mixture. ACS Omega 2019, 4, 1738–1747. [Google Scholar] [CrossRef] [PubMed]
  35. Pashaei, H.; Ghaemi, A.; Nasiri, M.; Heydarifard, M. Experimental investigation of the effect of nano heavy metal oxide particles in Piperazine solution on CO2 absorption using a stirrer bubble column. Energy Fuels 2018, 32, 2037–2052. [Google Scholar] [CrossRef]
  36. Li, S.H.; Ding, Y.; Zhang, X.S. Enhancement on CO2 bubble absorption in MDEA solution by TiO2 nanoparticles. Adv. Mater. Res. 2013, 631, 127–134. [Google Scholar] [CrossRef]
  37. Taheri, M.; Mohebbi, A.; Hashemipour, H.; Rashidi, A.M. Simultaneous absorption of carbon dioxide (CO2) and hydrogen sulfide (H2S) from CO2–H2S–CH4 gas mixture using amine-based nanofluids in a wetted wall column. J. Nat. Gas Sci. Eng. 2016, 28, 410–417. [Google Scholar] [CrossRef]
  38. Lee, J.W.; Kang, Y.T. CO2 absorption enhancement by Al2O3 nanoparticles in NaCl aqueous solution. Energy 2013, 53, 206–211. [Google Scholar] [CrossRef]
  39. Mohammaddoost, H.; Azari, A.; Ansarpour, M.; Osfouri, S. Experimental investigation of CO2 removal from N2 by metal oxide nanofluids in a hollow fiber membrane contactor. Int. J. Greenh. Gas Control 2018, 69, 60–71. [Google Scholar] [CrossRef]
  40. Peyravi, A.; Keshavarz, P.; Mowla, D. Experimental investigation on the absorption enhancement of CO2 by various nanofluids in hollow fiber membrane contactors. Energy Fuels 2015, 29, 8135–8142. [Google Scholar] [CrossRef]
  41. Zare, P.; Keshavarz, P.; Mowla, D. Membrane absorption coupling process for CO2 capture: Application of water-based ZnO, TiO2, and multi-walled carbon nanotube nanofluids. Energy Fuels 2019, 33, 1392–1403. [Google Scholar] [CrossRef]
  42. Preston, D.J.; Lu, Z.; Song, Y.; Zhao, Y.; Wilke, K.L.; Antao, D.S.; Louis, M.; Wang, E.N. Heat Transfer Enhancement During Water and Hydrocarbon Condensation on Lubricant Infused Surfaces. Sci. Rep. 2018, 8, 540. [Google Scholar] [CrossRef]
  43. Lashgarinejad, A.; Hosseini, S.S.; Irani, V.; Ghasemi, M.H.; Mohammadpour, R.; Tavasoli, A. Enhancement of CO2 absorption and heat transfer properties using amine functionalized magnetic graphene oxide/MDEA nanofluid. J. Iran. Chem. Soc. 2023, 20, 1629–1642. [Google Scholar] [CrossRef]
  44. Irani, V.; Tavasoli, A.; Vahidi, M. Preparation of amine functionalized reduced graphene oxide/methyl diethanolamine nanofluid and its application for improving the CO2 and H2S absorption. J. Colloid Interface Sci. 2018, 527, 57–67. [Google Scholar] [CrossRef] [PubMed]
  45. Irani, V.; Tavasoli, A.; Maleki, A.; Vahidi, M. Polyethyleneimine-functionalized HKUST-1/MDEA nanofluid to enhance the absorption of CO2 in gas sweetening process. Int. J. Hydrogen Energy 2018, 43, 5610–5619. [Google Scholar] [CrossRef]
  46. Vahidi, M.; Tavasoli, A.; Rashidi, A.M. Preparation of amine functionalized UiO-66, mixing with aqueous N-Methyldiethanolamine and application on CO2 solubility. J. Nat. Gas Sci. Eng. 2016, 28, 651–659. [Google Scholar] [CrossRef]
  47. Sodeifian, G.; Niazi, Z. Prediction of CO2 absorption by nanofluids using artificial neural network modeling. Int. Commun. Heat Mass Transf. 2021, 123, 105193. [Google Scholar] [CrossRef]
  48. Li, X.; Qin, J.; Wu, X.; Wei, C.; Xu, L. Designing back propagation neural network to predict CO2 mass transfer enhancement factor of TiO2-MEA/MDEA blended amine nanofluids. DeCarbon 2023, 2, 100021. [Google Scholar] [CrossRef]
  49. Bahiraei, M.; Heshmatian, S.; Moayedi, H. Artificial intelligence in the field of nanofluids: A review on applications and potential future directions. Powder Technol. 2019, 353, 276–301. [Google Scholar] [CrossRef]
  50. Eker, F.; Duman, H.; Akdaşçi, E.; Bolat, E.; Sarıtaş, S.; Karav, S.; Witkowska, A.M. A Comprehensive Review of Nanoparticles: From Classification to Application and Toxicity. Molecules 2024, 29, 3482. [Google Scholar] [CrossRef]
  51. Wongpinyochit, T.; Totten, J.D.; Johnston, B.F.; Seib, F.P. Microfluidic-Assisted Silk Nanoparticle Tuning. Nanoscale Adv. 2019, 1, 873–883. [Google Scholar] [CrossRef]
  52. McCall, R.L.; Sirianni, R.W. PLGA Nanoparticles Formed by Single- Or Double-Emulsion With Vitamin E-TPGS. J. Vis. Exp. 2013, 51015. [Google Scholar] [CrossRef] [PubMed]
  53. Lancel, M.; Gomez, C.; Port, M.; Amara, Z.; McCall, R.L.; Sirianni, R.W.; Wongpinyochit, T.; Totten, J.D.; Johnston, B.F.; Seib, F.P. Performances of Homogeneous and Heterogenized Methylene Blue on Silica Under Red Light in Batch and Continuous Flow Photochemical Reactors. Front. Chem. Eng. 2021, 3, 873–883. [Google Scholar] [CrossRef]
  54. Chaturvedi, K.R.; Kumar, R.; Trivedi, J.; Sheng, J.J.; Sharma, T. Stable Silica Nanofluids of an Oilfield Polymer for Enhanced CO2 Absorption for Oilfield Applications. Energy Fuels 2018, 32, 12730–12741. [Google Scholar] [CrossRef]
  55. Heidari, S.; Esmaeilzadeh, F.; Rafati, R.; Haddad, A.S. Experimental and modeling of CO2 absorption in a bubble column using a water-based nanofluid containing co-doped SiO2 nanoparticles. Model. Earth Syst. Environ. 2024, 10, 3229–3241. [Google Scholar] [CrossRef]
  56. Deng, Y.; Ren, S.; Malof, J.; Padilla, W. Deep Inverse Photonic Design: A Tutorial. Photon. Nanostruct. Fundam. Appl. 2022, 52, 101070. [Google Scholar] [CrossRef]
  57. Alavekar, A.; Lanjewar, T.; Vyawahare, A.; Patil, M.P. Artificial Intelligence in Carbon Capture: Modeling CO2 Absorption Using Amine Blends. J. Phys. Conf. Ser. 2024, 2763, 12020. [Google Scholar] [CrossRef]
  58. Su, Z.; Yang, L.; Zhao, N.; Song, J.; Li, X.; Wu, X. Experimental and neural network prediction of the cyclic stability and light absorption characteristics of supercritical CO2 based CNTs nanofluids. Appl. Therm. Eng. 2024, 241, 122347. [Google Scholar] [CrossRef]
  59. Delfani, S.; Esmaeili, M.; Karami, M. Application of artificial neural network for performance prediction of a nanofluid-based direct absorption solar collector. Sustain. Energy Technol. Assess. 2019, 36, 100559. [Google Scholar] [CrossRef]
  60. Devakki, B.; Thomas, S. Experimental Investigation on Absorption Performance of Nanofluids for CO2 Capture. Int. J. Air-Cond. Refrig. 2020, 28, 2050017. [Google Scholar] [CrossRef]
  61. Yin, Y.; Shen, M.; Tan, Z.; Yu, S.; Liu, J.; Jiang, G. Particle Coating-Dependent Interaction of Molecular Weight Fractionated Natural Organic Matter: Impacts on the Aggregation of Silver Nanoparticles. Environ. Sci. Technol. 2015, 49, 6581–6589. [Google Scholar] [CrossRef] [PubMed]
  62. Chakraborty, A.; Dalal, C.; Jana, N.R. Colloidal Nanobioconjugate With Complementary Surface Chemistry for Cellular and Subcellular Targeting. Langmuir 2018, 34, 13461–13471. [Google Scholar] [CrossRef] [PubMed]
  63. Kartohardjono, S.; Shabanindita, S.; Harianja, M.; Dixon, A.O.; Yuliusman, Y.; Saksono, N. N2O Absorption Through Super Hydrophobic Hollow Fiber Membrane Contactor. Environ. Prog. Sustain. Energy 2018, 38, 362–366. [Google Scholar] [CrossRef]
Eng 06 00226 g001
Figure 1. Workflow of the ANN-based meta-analysis framework for predicting CO2 absorption performance in nanoparticle-aided systems, showing data collection, preprocessing, model training, validation, and feature importance analysis.
Figure 1. Workflow of the ANN-based meta-analysis framework for predicting CO2 absorption performance in nanoparticle-aided systems, showing data collection, preprocessing, model training, validation, and feature importance analysis.
Eng 06 00226 g001
Eng 06 00226 g002
Figure 2. Distribution of nanoparticle types reported in the literature for nanoparticle-enhanced CO2 separation systems.
Figure 2. Distribution of nanoparticle types reported in the literature for nanoparticle-enhanced CO2 separation systems.
Eng 06 00226 g002
Eng 06 00226 g003
Figure 3. Distribution of experimental system types employed in nanoparticle-assisted CO2 absorption studies.
Figure 3. Distribution of experimental system types employed in nanoparticle-assisted CO2 absorption studies.
Eng 06 00226 g003
Eng 06 00226 g004
Figure 4. CO2 absorption percentages for different nanoparticle types, highlighting variations in performance across materials.
Figure 4. CO2 absorption percentages for different nanoparticle types, highlighting variations in performance across materials.
Eng 06 00226 g004
Eng 06 00226 g005
Figure 5. Learning curves of the ANN model, showing the evolution of training and validation loss over 200 epochs, illustrating model convergence and generalization performance.
Figure 5. Learning curves of the ANN model, showing the evolution of training and validation loss over 200 epochs, illustrating model convergence and generalization performance.
Eng 06 00226 g005
Eng 06 00226 g006
Figure 6. Actual versus predicted enhancement factor (%) based on the trained ANN model, demonstrating predictive accuracy. The color gradient represents variations in nanoparticle (NP) concentrations within the dataset, where lighter shades indicate higher NP concentrations and darker shades correspond to lower NP concentrations.
Figure 6. Actual versus predicted enhancement factor (%) based on the trained ANN model, demonstrating predictive accuracy. The color gradient represents variations in nanoparticle (NP) concentrations within the dataset, where lighter shades indicate higher NP concentrations and darker shades correspond to lower NP concentrations.
Eng 06 00226 g006
Eng 06 00226 g007
Figure 7. Feature influence on CO2 absorption as determined by the absolute value of linear regression coefficients.
Figure 7. Feature influence on CO2 absorption as determined by the absolute value of linear regression coefficients.
Eng 06 00226 g007
Eng 06 00226 g008
Figure 8. Architecture of the developed ANN model and training–validation performance curve. The figure shows the number of input variables, hidden layers, and neurons, along with model convergence during training.
Figure 8. Architecture of the developed ANN model and training–validation performance curve. The figure shows the number of input variables, hidden layers, and neurons, along with model convergence during training.
Eng 06 00226 g008
Eng 06 00226 g009
Figure 9. Distribution of base fluids used in nanoparticle-enhanced CO2 separation studies.
Figure 9. Distribution of base fluids used in nanoparticle-enhanced CO2 separation studies.
Eng 06 00226 g009
Eng 06 00226 g010
Figure 10. Correlation matrix among key variables influencing CO2 absorption. The red-to-blue gradient indicates the strength and direction of the correlation coefficient, where red represents a strong positive correlation, blue indicates a strong negative correlation, and lighter shades correspond to weaker correlations.
Figure 10. Correlation matrix among key variables influencing CO2 absorption. The red-to-blue gradient indicates the strength and direction of the correlation coefficient, where red represents a strong positive correlation, blue indicates a strong negative correlation, and lighter shades correspond to weaker correlations.
Eng 06 00226 g010
Eng 06 00226 g011
Figure 11. Enhancement factor achieved by each nanoparticle type.
Figure 11. Enhancement factor achieved by each nanoparticle type.
Eng 06 00226 g011
Table 1. Comparative summary of key studies on nanoparticle-enhanced CO2 absorption and modeling approaches.
Table 1. Comparative summary of key studies on nanoparticle-enhanced CO2 absorption and modeling approaches.
Nanoparticles UsedSystem TypeSolvent TypeEnhancement Factor (EF)Modeling/Analysis ApproachLimitation/How This Study Differs
Fe3O4@SiO2-NH2Batch ReactorWater34.2%Experimental onlyFocused on single nanoparticle type and single system; no predictive modeling used.
ZnOMembrane ContactorWater130%Experimental onlyOptimized for one reactor type; no integration across multiple configurations.
SiO2Packed ColumnDEA40%Experimental onlyDid not consider nanoparticle concentration, surface functionalization, or ANN prediction.
NH2-GOEquilibrium CellWater19%Experimental onlyLimited to small-scale equilibrium studies; lacks generalizable modeling.
TiO2Bubble ColumnMDEA11.5%Experimental + Mechanistic ModelMechanistic models struggle to capture nonlinear interactions across diverse datasets.
CNTHollow Fiber MembraneWater32%Experimental onlyFocused on one nanoparticle; does not evaluate cross-system predictive capability.
Al2O3Wetted Wall ColumnDEA33%Experimental onlyLimited solvent variability; no integration of data-driven prediction methods.
This StudyIntegrated FrameworkMultiple (Water, Amine, Mixed)Up to 130%ANN-based meta-analysisFirst, to integrate 312 literature datasets across multiple reactor types, apply ANN predictive modeling, conduct feature importance analysis, and provide generalizable design insights.
Table 2. CO2 absorption in batch reactors.
Table 2. CO2 absorption in batch reactors.
Nanoparticle (NP)Base FluidOptimal NP LoadingEF (%)Ref
SiO2DW0.1 wt%7[32]
ZnODW0.1 wt%14[32]
Fe3O4−DW0.02 wt%25.07[31]
Fe3O4−DW0.1 wt%31.04[31]
Fe3O4@SiO2-NH2DW0.1 wt%34.23[31]
Fe3O4−MDEA0.02 wt%6.78[31]
Fe3O4−MDEA0.1 wt%12.13[31]
Table 3. CO2 absorption in bubble and tray columns.
Table 3. CO2 absorption in bubble and tray columns.
Nanoparticle
(NP)		Base Fluid
(BF)		DeviceOptimal NP LoadingEF (%)Ref
Al2O3MethanolTray Column0.05 vol%9.4[33]
SiO2MethanolTray Column0.05 vol%9.7[34]
ZnODEABubble column0.1 wt%33.3[35]
ZnOPiperazineBubble column0.1 wt%17[35]
TiO2MDEABubble column0.8 wt%11.54[36]
Al2O3DEAWetted wall column0.05 wt%33[37]
Al2O3NaClBubble Column0.01 vol%12.5[38]
SiO2DEAWetted wall column0.05 wt%40[37]
Table 4. CO2 absorption in hollow fiber membrane contactor (HFMC).
Table 4. CO2 absorption in hollow fiber membrane contactor (HFMC).
Nanoparticle Base FluidLFR (mL/min)Optimal NP LoadingEF (%)Ref
CNTDW166.70.05 wt%32[27]
SiO2DW166.70.05 wt%16[27]
Al2O3DW16000.2 wt%125[39]
Fe3O4DW116.70.15 wt%43.8[40]
CNTDW116.70.1 wt%38[40]
Al2O3DW116.70.05 wt%3[40]
SiO2DW116.70.05 wt%25.9[40]
ZnODW100.15 wt%130[41]
MWCNTDW100.15 wt%60[41]
TiO2DW100.15 wt%60[41]
Fe3O4DW100.025 wt%40.31[42]
Fe3O4@DW100.05 wt%84.45[42]
Table 5. CO2 absorption in Equilibrium cell.
Table 5. CO2 absorption in Equilibrium cell.
Nanoparticle Base FluidOptimal NP LoadingEF (%)Ref
GOMDEA0.2 wt%10.4[43]
PEI@GOMDEA0.1 wt%15[18]
NH2-GO reducedMDEA0.1 wt%16.2[43]
NH2-GO magneticMDEA0.1 wt%19[44]
PEI@HKUSTMDEA0.2 wt%16[45]
UiO66-NH2MDEA0.1 wt%10[46]
Table 6. Model Performance Comparison.
Table 6. Model Performance Comparison.
ModelR2RMSE (%)MAE (%)
ANN0.924.23.1
Random Forest0.885.64.3
Linear Regression0.767.96.5
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Ghasem, N. AI-Driven Predictive Modeling of Nanoparticle-Enhanced Solvent-Based CO2 Capture Systems: Comprehensive Review and ANN Analysis. Eng 2025, 6, 226. https://doi.org/10.3390/eng6090226

AMA Style

Ghasem N. AI-Driven Predictive Modeling of Nanoparticle-Enhanced Solvent-Based CO2 Capture Systems: Comprehensive Review and ANN Analysis. Eng. 2025; 6(9):226. https://doi.org/10.3390/eng6090226

Chicago/Turabian Style

Ghasem, Nayef. 2025. "AI-Driven Predictive Modeling of Nanoparticle-Enhanced Solvent-Based CO2 Capture Systems: Comprehensive Review and ANN Analysis" Eng 6, no. 9: 226. https://doi.org/10.3390/eng6090226

APA Style

Ghasem, N. (2025). AI-Driven Predictive Modeling of Nanoparticle-Enhanced Solvent-Based CO2 Capture Systems: Comprehensive Review and ANN Analysis. Eng, 6(9), 226. https://doi.org/10.3390/eng6090226

Article Metrics

Back to TopTop