Boosting CO₂ Uptake from Waste Concrete Powder Using Artificial Intelligence and the Marine Predators Algorithm

Abstract

Waste concrete powder (WCP) is emerging as a potential method of adoption for CO₂ sequestration due to its ability to chemically react with carbon dioxide and trap it within its structure. This study explores the application of artificial intelligence (AI) and the Marine Predators Algorithm (MPA) to maximize the absorption of CO₂ from waste concrete powder generated by recycling plants for building and demolition debris. Initially, a model is developed to assess CO₂ uptake according to carbonation time (CT) and water-to-solid ratio (WSR), utilizing the adaptive neuro-fuzzy inference system (ANFIS) modeling approach. Subsequently, the MPA is employed to estimate the optimal values for CT and WSR, thereby maximizing CO₂ uptake. A significant improvement in modeling accuracy is evident when the ANOVA method is replaced with ANFIS, leading to a substantial increase of approximately 19% in the coefficient of determination (R-squared) from 0.84, obtained through ANOVA, to an impressive 0.9999 obtained through the implementation of ANFIS; furthermore, the utilization of ANFIS yields a substantial reduction in the root mean square error (RMSE) from 1.96, as indicated by ANOVA, to an impressively low value of 0.0102 with ANFIS. The integration of ANFIS and MPA demonstrates impressive results, with a nearly 30% increase in the percentage value of CO₂ uptake. The highest CO₂ uptake of 3.86% was achieved when the carbonation time was 54.3 h, and the water-to-solid ratio was 0.27. This study highlights the potential of AI and the MPA as effective tools for optimizing CO₂ absorption from waste concrete powder, contributing to sustainable waste management practices in the construction industry.

1. Introduction

Currently, the optimization of CO₂ uptake has become a crucial focus in the pursuit of sustainable solutions to mitigate climate change. The atmospheric CO₂ concentration has been rapidly increasing at an exceptional rate of 2.2 ppm/year [1]. It is estimated that by approximately 2050, the atmospheric CO₂ concentration may reach around 450 ppm. Consequently, this increase in CO₂ levels is projected to cause a temperature rise of 2 °C to 3 °C on the Earth’s surface due to the greenhouse effect [2]. To tackle these challenges, numerous methods and techniques have been developed to capture and store CO₂, each offering its own advantages and potential applications [3,4,5]. Carbon capture and storage is a method that captures CO₂ emissions generated from industrial processes and securely stores them underground, ensuring that they do not escape into the atmosphere [6,7]. Carbon capture and utilization refers to the process of capturing CO₂ emissions and converting them into valuable commodities, such as chemicals, plastics, or fuels [8,9]. Biological carbon sequestration exploits the natural capacity of plants and trees to absorb CO₂ through photosynthesis [10]. Enhanced weathering accelerates natural processes to capture CO₂ by applying crushed minerals like basalt or olivine, which react with CO₂ to form stable carbonates [11]. Direct air capture uses specialized technologies to extract CO₂ directly from the ambient air [12]. The choice of approach relies on various factors, including the CO₂ emission source, implementation scale, economic feasibility, and environmental considerations.

The continuous growth of the construction industry has led to a rise in waste generation, particularly in the form of concrete by-products. Traditionally, waste concrete powder has been viewed as an environmental concern, typically leading to landfill disposal or pollution if not handled properly [13]. In the European Union, construction and demolition waste comprise approximately 35–46% of the total waste stream in the European Union [14]. Notably, end-of-life concrete makes up a significant portion of this construction and demolition waste, contributing around 12–40% to the overall waste volume [15]; however, recent studies highlighted the potential of waste concrete powder as a valuable resource for CO₂ sequestration [16]. By utilizing waste concrete powder in a controlled environment, we can enhance its ability to capture and store CO₂, turning it into a beneficial asset rather than a liability [17]. Studies have specifically explored the recycling of alkaline-rich waste, including filter-pressed concrete slurry waste, through accelerated carbonation techniques [18,19].

Waste concrete powder (WCP) exhibits promising potential as an option for carbon dioxide (CO₂) sequestration, owing to its capability to undergo chemical reactions with CO₂ and effectively store it within its structure. The CO₂ uptake capacity refers to the amount of carbon dioxide that can be absorbed and stored by WCP through a process known as mineral carbonation [20]. Mineral carbonation is a process that converts CO₂ into solid mineral forms, thereby effectively sequestering the carbon and preventing its release into the atmosphere [21]. The CO₂ uptake capacity of WCP is influenced by various factors such as the carbonation time (CT) (duration of the carbonation process) and the water-to-solid ratio (WSR), which evaluates the amount of water used in relation to the quantity of WCP [22]. Optimizing these variables is crucial for enhancing the CO₂ uptake capacity of WCP and establishing it as a sustainable solution for carbon capture. This entails selecting the appropriate WCP composition, identifying optimal CO₂ absorption conditions, and developing efficient implementation techniques [23]. Studies have shown that the ability of WCP to absorb CO₂ can be significantly enhanced by optimizing these factors [24]. For example, enhancing the fineness of the crushed concrete particles, which refers to reducing the particle size of WCP, can create more surface area for the chemical reaction to occur, leading to a higher CO₂ uptake capacity [25]. Kaliyavaradhan et al. [26] explored the dual functionality of concrete slurry waste as a CO₂ capture agent and a supplementary cementitious material. The study investigated the influence of WSR and reaction time on the CO₂ uptake capacity of concrete slurry waste using response surface methodology (RSM). The optimal conditions were found to be a w/s ratio of 0.25 and a reaction time of 72 h, resulting in a remarkable maximum CO₂ uptake of 20.4%.

Accelerated mineral carbonation has gained considerable attention as a promising technique in carbon capture and storage. This process allows for the long-term storage of CO₂ by employing a controlled chemical reaction between CO₂ and alkaline oxides, like calcium oxide (CaO) and magnesium oxide (MgO), commonly present in natural silicate rocks or industrial by-products. The outcome of this reaction is the formation of stable carbonate compounds, such as calcium carbonate (CaCO₃) and magnesium carbonate (MgCO₃), which find diverse practical applications. Pan et al. [27] estimated the global CO₂ mitigation potential of applying accelerated carbonation to various alkaline solid wastes. The results indicate that CO₂ mineralization and utilization can significantly reduce CO₂ emissions, achieving a 12.5% global reduction equivalent to 4.02 Gt per year.

A significant number of experimental investigations are dedicated to evaluating and optimizing CO₂ storage through accelerated carbonation processes. These studies aim to optimize operational parameters, encompassing temperature, pressure, gas humidity, liquid and gas flow rates, liquid-to-solid ratio, solid pre-treatment, and particle size [28,29,30]. Based on experimental results, the filter-pressed concrete slurry waste has exhibited an impressive CO₂ sequestration capacity, reaching up to 75% of the total CO₂ uptake in just a few hours. Over a duration of 144 h of carbonation, concrete slurry waste demonstrated the capability to capture 110 g of CO₂/kg of dry concrete slurry waste [31].

Due to the costly, limited, and time-intensive nature of conducting experimental studies on CO₂ uptake from waste concrete powder across different operational conditions to determine the highest CO₂ uptake, there is a need for alternative effective approaches to assess characteristics effectively. Employing optimization methods offers an advantageous approach to minimize the necessity for extensive experimental trials [32,33,34]. This study aims to bridge the existing gap in the application of AI techniques and the recent optimization method, specifically the Marine Predators Algorithm (MPA), for maximizing CO₂ absorption from WCP. Although WCP has shown promise as a potential CO₂ sequestration agent, there is a lack of comprehensive studies that explore the optimization of CO₂ uptake from WCP using advanced AI algorithms. Therefore, this study aims to bridge this gap by integrating the ANFIS modeling approach and the MPA to identify optimal values for CT and WSR, thus, enhancing CO₂ absorption from WCP. The primary point of this research is to enhance CO₂ uptake through the utilization of the MPA in combination with ANFIS modeling. The initial step involves the development of an ANFIS model based on experimental datasets to simulate the CO₂ uptake, considering the CT and WST as variables. Subsequently, the MPA is employed to identify the optimal values for the CT and WSR, aiming to maximize the CO₂ uptake. Furthermore, this study includes a comparative analysis between ANFIS modeling and the traditional ANOVA method in terms of accuracy and prediction capabilities for CO₂ uptake from WCP. This analysis provides insights into the superiority of ANFIS modeling, which exhibits a substantial increase in the R-squared and a significant reduction in the RMSE.

The contribution aspect of this study is observed in the integration of AI techniques, particularly ANFIS modeling and the MPA, to optimize CO₂ absorption from waste concrete powder. While previous studies have primarily focused on exploring the CO₂ sequestration potential of WCP, a limited number have employed advanced AI algorithms to enhance the absorption process. By utilizing ANFIS modeling, this research improves the accuracy and prediction capabilities for CO₂ uptake, leading to significant advancements in modeling accuracy; moreover, the integration of ANFIS and the MPA further enhances the percentage value of CO₂ uptake, demonstrating the effectiveness of this novel approach. The findings of this study contribute to sustainable waste management practices in the construction industry by providing valuable insights into optimizing CO₂ absorption from waste concrete powder using AI techniques.

The key contributions of the paper can be outlined as follows:

• Development of a robust ANFIS model for CO₂ uptake, considering carbonation time and water-to-solid ratio;
• Introduction of a novel implementation of MPA to enhance CO₂ uptake;
• Conducting a thorough analysis and comparison between the proposed Marine Predators Algorithm (MPA) and various featured algorithms such as Particle Swarm Optimization (PSO), Genetic Algorithm (GA), Harris Hawks Optimization (HHO), and Cuckoo Search (CS) to validate the efficacy of the MPA method; furthermore, employing statistical tests to ensure an equitable and objective assessment of the different metaheuristic algorithms performed.

2. Approach and Methodology

2.1. Dataset Description

In order to enhance the absorption of CO₂ from waste concrete powder, an initial step involved the development of an ANFIS model. This model was designed to simulate CO₂ absorption by considering the CT and the WSR. To facilitate this process, a dataset with the permission of the license number 5580660002812 was utilized [23]. The experimental trials, as depicted in

Table 1. Dataset description for training and testing the ANFIS model of CO₂ uptake.

CT (h)	WSR	CO2 Uptake (%)
120	0.7	2.44
72	0.1	1.74
168	0.4	2.94
120	0.4	2.87
24	0.25	2.47
72	0.7	0.89
72	0.4	2.71
168	0.1	2.15
1	0.1	0.65
120	0.25	2.56
1	0.1	0.65
120	0.1	2.07
72	0.4	2.71
168	0.4	2.94
168	0.7	2.91
1	0.7	0.03
24	0.55	1.56
1	0.7	0.03
1	0.4	0.81
168	0.7	2.91
168	0.1	2.15
72	0.55	2.67
72	0.4	2.71

, encompassed a range of WSR from 0.1 to 0.7 and CT from 1 to 168 h. In total, 23 experimental runs were performed to train and validate the ANFIS model for CO₂ uptake analysis.

2.2. ANFIS Model of CO₂ Uptake

The ANFIS, originating in the early 1990s, is a specialized form of an artificial neural network [35]. It combines the interpretability of fuzzy systems with the learning capabilities of neural networks, specifically based on the Takagi–Sugeno fuzzy inference system [36]. The fundamental principle of ANFIS revolves around mapping inputs to outputs through a sequence of fuzzy IF-THEN rules and learning algorithms. This feature enables the system to effectively handle nonlinear functions and dynamically adapt to evolving input/output data, making it well-suited for a diverse range of applications such as prediction, pattern recognition, control, and decision-making [37]. The ANFIS system is structured with three essential phases: the fuzzifier, inference engine, and defuzzifier, visually presented in Figure 1.

The initial operation in the ANFIS system involves the transformation (mapping) of crisp input values into fuzzy sets. In this phase, specific membership functions (MFs) are employed to achieve the task. These MFs are designed as convex functions, mapping the inputs to a restricted range within the interval of [0, 1]. The domain of discourse for the input can be divided into multiple MFs based on the context or application under consideration. Fuzzy membership functions (MFs) encompass various forms, including Gaussian, triangular, and trapezoidal, among others.

The rule set consists of fuzzy criteria that dictate the relationship between the inputs and outputs of the system.

These rules, regardless of their type, follow the IF-THEN format. There are two categories of fuzzy sets available, and the selection between them depends on the specific nature of the problem. The two classifications of fuzzy rules are referred to as Mamdani and Sugeno styles. In a system with two inputs (x and y) and one output (f), the rule structure is determined by the two types of fuzzy rules and can be represented as follows:

In a Sugeno-type fuzzy rule, IF x is A₁ and y is B₁ THEN f₁ = p₁x + q₁y + r₁

(1)

IF x is A₂ and y is B₁ THEN f₂ = p₂x + q₂y + r₂

(2)

where “A₁, A₂, B₁ are B₂ are fuzzy sets associated with premise input variables x and y, respectively, while, p₁, q₁, r₁ and p₂, q₂, r₂ are consequent part parameters”. The estimation of the final output is calculated as follows:

$$f={\tilde{\omega }}_1{f}_1+{\tilde{\omega }}_2{f}_2 (\mathrm{Output} \mathrm{Layer})$$

(3)

$$\mathrm{Defuzzification} \mathrm{Layer}: \mathrm{Evaluating} {\tilde{\omega }}_1{g}_1(x,y) and {\tilde{\omega }}_2{g}_2(x,y)$$

(4)

$${\tilde{\omega }}_1=\frac{{\omega }_1}{{\omega }_1+{\omega }_2} \mathrm{and} {\tilde{\omega }}_2=\frac{{\omega }_2}{{\omega }_1+{\omega }_2} \left(\mathrm{N}\mathrm{Layer}\right)$$

(5)

$${\omega }_1={\mu }_{A_1}\ast {\mu }_{B_1} \mathrm{and} {\omega }_2={\mu }_{A_2}\ast {\mu }_{B_2} (\pi \mathrm{Layer})$$

(6)

$$\mathrm{Fuzzification} \mathrm{Layer}: {\mu }_{A_1}, {\mu }_{A_2},{\mu }_{B_1} \mathrm{and} {\mu }_{B_2} \mathrm{are} \mathrm{the} \mathrm{MF} \mathrm{grades} \mathrm{of} \mathrm{the} \mathrm{two} \mathrm{inputs}$$

(7)

The generation of fuzzy rules is typically accomplished either through system experts or by employing a clustering algorithm on the dataset. In this phase, the IF-THEN statements are triggered to deduce their corresponding outputs based on a specific set of inputs.

In particular, the implication method is employed to determine the output of each rule, providing a range of techniques for its application. Among these techniques, the Min (Intersection) operation is the most commonly employed. Once the rules are activated, the outputs of all rules are combined or aggregated to yield a single output value. In the case of Mamdani-criterion fuzzy rules, in cases where the rule’s output is represented as a fuzzy value, the Max (Union) operation is utilized to determine the final outcome. Conversely, for Sugeno-criterion fuzzy rules, the final output is obtained by utilizing a weighted average.

2.3. Parameter Identification: Marine Predators Algorithm (MPA)

The MPA is an innovative metaheuristic search technique, inspired by the optimal foraging behavior and encounter rate policies exhibited by marine creatures in nature [38]. The MPA emulates the foraging strategies used by marine creatures in their pursuit of food through a three-phase approach [39]. Phase one is implemented during the initial third of the total iterations t_Max, and its modeling can be expressed as follows

$$\begin{gathered} D_i=R_B\otimes ({\mathit{Elite}}_i-R_B\otimes \mathit{Pr}e y_i) \\ \mathit{Pr}e y_{i+1}=\mathit{Pr}e y_i+0.5·R\otimes D_i \end{gathered}$$

(8)

where D_i denotes the step size of the ith predator; R_B is a vector generated based on the distribution of Brownian motion; R represents random numbers within the range [0, 1]; and the symbol ⨂ denotes entry-wise multiplications. When the iteration t falls within the second third of t_Max, phase two is initiated, which is further divided into two subphases. If t is less than half of t_Max, phase two can be mathematically expressed as follows:

$$\begin{gathered} D_i=R_L\otimes ({\mathit{Elite}}_i-R_L\otimes \mathit{Pr}e y_i) \\ \mathit{Pr}e y_{i+1}=\mathit{Pr}e y_i+0.5·R\otimes D_i \end{gathered}$$

(9)

where R_L represents a vector generated using the distribution of Lévy motion. If the iteration t is greater than half of t_Max, phase 2 can be mathematically expressed as follows:

$$\begin{gathered} D_i=R_B\otimes (R_B\otimes {\mathit{Elite}}_i-\mathit{Pr}e y_i) \\ \mathit{Pr}e y_{i+1}={\mathit{Elite}}_i+0.5·\mathit{CF}\otimes D_i \\ CF={[1-\left(\frac{t}{t_{Max}}\right)]}^{2t/t_{Max}} \end{gathered}$$

(10)

The final phase can be represented by the following equation:

$$\begin{gathered} D_i=R_L\otimes (R_L\otimes {Elite}_i-\mathit{Pr}e y_i) \\ \mathit{Pr}e y_{i+1}={Elite}_i+0.5·CF\otimes D_i \end{gathered}$$

(11)

Figure 2 illustrates the flowchart of the MPA algorithm.

The objective of the optimization phase is to acquire the optimal values of CT and WSR. In the MPA optimization process, these variables are treated as decision variables, while CO₂ uptake serves as the objective function to be maximized. The problem can be expressed as:

$$x=\underset{x\in R}{\arg }\max \left(y\right)$$

(12)

where x denotes the set of input variables, and y represents the output variable.

3. Results and Discussion

This study examined the CO₂ absorption from WCP, with a focus on assessing the impacts of two critical factors: carbonation time (CT) and water-to-solid ratio (WSR). The investigation harnessed the power of artificial intelligence (AI) and the Marine Predators Algorithm (MPA) to pinpoint optimal values for these parameters.

The impact of carbonation time (CT) on CO₂ uptake from waste concrete powder is a pivotal aspect in assessing the carbon capture potential of this material. Carbonation time refers to the duration of exposure of waste concrete powder to CO₂-rich environments, initiating chemical reactions between CO₂ and the calcium-containing compounds within the concrete. According to Table 1, longer CTs are associated with several notable effects on CO₂ uptake. With an increase in the CT, there is a tendency for enhanced CO₂ uptake driven due to enhanced penetration, leading to more extensive chemical reactions and calcium carbonate formation; however, identifying the optimal carbonation duration becomes crucial, as beyond a certain point, additional CO₂ exposure might yield diminishing returns in CO₂ absorption. Moreover, prolonged carbonation can induce physical changes in waste concrete powder, influencing properties like porosity, density, and strength, which impact its applicability.

Exploring the impact of the water-to-solid ratio (WSR) on CO₂ uptake from WCP constitutes a vital aspect of this study. The water-to-solid ratio denotes the proportion of water used in the carbonation process relative to the quantity of waste concrete powder. The changes in the WSR have the potential to influence CO₂ absorption capacity, potentially leading to enhancement due to improved interaction between CO₂ and the WCP. Additionally, at the beginning of the process, with a decrease in WSR or with an increase in the solid–liquid ratio, there is a corresponding elevation in the amount of CO₂ uptake. This implies a greater availability of calcium (Ca) to engage in CO₂ capture and the subsequent formation of calcium carbonate (CaCO₃). Additionally, the CO₂ uptake efficiency (measured as g-CO₂/g-concrete fines) remains relatively consistent, showing a slight increase with higher solid–liquid ratios, as displayed in Table 1. Furthermore, water distribution within the material plays a role in the uniformity of carbonation, and variations in the ratio can also lead to changes in physical properties. Therefore, determining the optimal WSR becomes crucial, similar to CT, in order to attain effective CO₂ capture.

3.1. Modeling Phase

The ANFIS framework was constructed based on a dataset of 23 experiments, with the dataset divided into two groups for training and testing purposes. The first group, comprising 18 points, was used for training the model, and the remaining points were designated for testing the model’s performance. The hybrid training method integrated least squares estimation in the forward path and utilized backpropagation for the backward direction to achieve an effective training process. A set of 10 fuzzy rules for the system was derived using the subtractive clustering method. The model was trained iteratively until a reduced RMSE was achieved.

Table 2. Statistical analysis of ANFIS model of CO₂ uptake.

RMSE			Coefficient of Determination
Training	Testing	All	Training	Testing	All
3.3124 × 10−6	0.022	0.0102	1.0	0.9994	0.9999

presents the statistical metrics obtained from the ANFIS model.

The ANFIS model for CO₂ uptake exhibited RMSE values of 3.3124 × 10⁻⁶ and 0.022 for the training and testing datasets, respectively, as displayed in Table 2. The R-Square for the training and testing stages are 1.0 and 0.9994, respectively, as presented in Table 2. Comparing these results with ANOVA [23], the R-Square has increased by approximately 19% from 0.84 using ANOVA to 0.9999 using ANFIS. Additionally, the RMSE has significantly decreased from 1.96 using ANOVA to 0.0102 using ANFIS, indicating the effectiveness of the ANFIS modeling phase. Figure 3 demonstrates the architecture of the ANFIS model with two inputs and one output, while Figure 4 displays the general contours of the Gaussian-form MFs.

Figure 5 presents a spatial representation from a 3-D perspective, illustrating the contours of the input–output function for each combination of inputs. The color gradient ranges from dark red, representing the highest output value of CO₂ uptake, to blue, indicating the lowest. The figure demonstrates the integration of ANFIS and MPA, highlighting the accomplishment of a remarkable CO₂ uptake rate of 3.86%. This optimization is achieved through the identification of the optimal carbonation time (CT) of 54.3 h and the selection of a suitable water-to-solid ratio (WSR) of 0.327.

Figure 6 displays the comparison between the predicted and measured data of the ANFIS model for CO₂ uptake. The plot demonstrates a strong alignment between the estimated and measured values. This alignment between the estimated and measured values is particularly robust, indicating a high level of accuracy and reliability in the ANFIS model’s predictive capabilities; furthermore, Figure 7 exhibits the prediction results for both the training and testing phases, depicting a close approximation to the line of 100% accuracy.

Table 3. Validation of ANFIS model of CO₂ uptake.

	WSR	CT	CO2 Uptake (%)	AE	% Error
Experimental [23]	0.32	72 h	2.83	-	-
ANOVA [23]	0.32	72 h	5.58	0.25	8.83
ANFIS and MPA	0.32	72 h	2.97	0.14	4.95

presents the validation results of the ANFIS model for CO₂ uptake in comparison to ANOVA. The absolute error values for CO₂ uptake are 0.25 and 0.14, respectively, when compared to the experimental data for ANOVA and ANFIS. The ANFIS model exhibits a 44% reduction in absolute error for CO₂ uptake compared to ANOVA. Furthermore, the percentage error values are 8.83% and 4.95% for ANOVA and ANFIS, respectively. These findings highlight the superior performance of the ANFIS model when compared to the ANOVA model.

3.2. Optimization Phase

In this section, the objective was to determine the optimal values of the water-to-solid ratio (WSR) and carbonation time (CT) that would result in a high CO₂ uptake percentage. To achieve this, the Marine Predators Algorithm (MPA) was utilized in combination with reliable ANFIS models of CO₂ uptake. In the process of optimization, the two parameters CT and WSR are utilized as decision variables, aiming to maximize the CO₂ uptake, which functions as the objective function. The optimized results obtained using the experimental, RSM, and recommended methods are displayed in

Table 4. Optimized results using experimental, RSM, and proposed methods.

	WSR	CT (h)	CO2 Uptake (%)
Experimental [23]	0.4	168	2.94
RSM [23]	0.4	90	2.8
ANFIS and MPA	0.27	54.3	3.86

. The integration of ANFIS and MPA led to a significant increase in the CO₂ uptake percentage by approximately 30% compared to the experimental data and RSM. The optimal values under these conditions were found to be 0.27 for the WSR and 54.3 h for the CT. Figure 8 provides a visual representation of the convergence process of particles concerning three key aspects: the objective function, WSR, and CT. The figure offers a clear trajectory of each particle’s iterative approach toward the optimal values of these parameters. It can be observed that all particles converged to the optimal values within 20 iterations. This concise and efficient convergence underscores the efficacy of the method employed, instilling confidence in the precision and reliability of the obtained optimal parameter values.

The Particle Swarm Optimization (PSO), Genetic Algorithm (GA), Harris Hawks Optimization (HHO), and Cuckoo Search (CS) outcomes are compared to validate the MPA suppository. To ensure consistent outcomes, each optimizer is executed 30 times, and a comprehensive statistical analysis is performed.

Table 5. Details of 30 runs using different optimizers.

No.	PSO	MPA	HHO	CS	GA	No.	PSO	MPA	HHO	CS	GA
1	3.864	3.866	3.834	3.865	3.845	16	3.866	3.866	3.748	3.866	3.821
2	3.866	3.866	3.833	3.866	3.752	17	3.866	3.866	3.79	3.861	3.798
3	3.865	3.866	3.862	3.866	3.865	18	3.863	3.86	3.857	3.866	3.632
4	3.864	3.866	3.851	3.866	3.835	19	3.866	3.866	3.797	3.866	3.521
5	3.865	3.866	3.675	3.866	3.803	20	3.865	3.866	3.855	3.866	3.84
6	3.865	3.865	3.866	3.866	3.863	21	3.865	3.866	3.862	3.865	3.849
7	3.866	3.866	3.81	3.863	3.851	22	3.864	3.866	3.829	2.987	3.815
8	3.866	3.866	3.857	3.863	3.858	23	3.864	3.866	3.866	3.855	3.862
9	3.866	3.866	3.866	3.865	3.865	24	3.866	3.866	3.86	3.866	3.844
10	3.863	3.866	3.863	3.865	3.843	25	3.866	3.866	3.86	3.814	3.772
11	3.866	3.866	3.85	3.372	3.864	26	3.866	3.857	3.862	3.866	3.801
12	3.866	3.866	3.659	3.86	3.653	27	3.865	3.866	3.79	3.866	3.86
13	3.866	3.866	3.863	3.748	3.527	28	3.864	3.866	3.864	3.866	3.848
14	2.993	3.864	3.846	3.866	3.467	29	3.864	3.866	3.815	3.866	3.836
15	3.864	3.866	3.842	3.552	3.7	30	3.866	3.866	3.72	3.865	3.167

and Figure 9 present the detailed results of 30 runs utilizing different optimizers. Statistical indicators such as the highest value, lowest value, average value, and standard deviation are computed and presented in

Table 6. Statistical evaluation for considered optimizers.

	PSO	MPA	HHO	CS	GA
Best	3.866	3.866	3.866	3.866	3.865
Worst	2.993	3.857	3.659	2.987	3.167
Mean	3.836	3.865	3.825	3.803	3.762
STD	0.157	0.002	0.055	0.184	0.156

. The average values of the cost function (based on the ANFIS model of CO₂ uptake) range from 3.762 to 3.857. The MPA achieved the highest average value (3.857), followed by PSO (3.836), and then GA (3.762) with the lowest average value. The standard deviation values range from 0.002 to 0.184, where MPA demonstrates the best standard deviation (STD) value of 0.002, followed by HHO (0.055), and CS (0.184). These results highlight the effectiveness of MPA in identifying optimal values that lead to the highest CO₂ uptake.

4. Conclusions

Exploring the enhancement of CO₂ uptake from waste concrete powder yields a range of advantages, including minimizing environmental impact by reducing building waste accumulation, fostering resource efficiency through waste utilization, and promoting sustainable construction practices. The percentage value of CO₂ uptake of WCP from a recycling plant for building and demolition debris is influenced by two primary factors: the WSR and the CT. The objective of this study is to identify the optimal values for the WSR and CT in order to maximize CO₂ uptake. The integration of artificial intelligence, specifically the Adaptive ANFIS model and the Marine Predators Algorithm (MPA), was employed for the modeling and parameter identification processes. During the modeling phase, the ANFIS model demonstrated excellent performance with low RMSE values of 3.3124 × 10⁻⁶ for the training data set and 0.022 for the testing data set. The coefficients of determination were also high, with values of 1.0 for training and 0.9994 for testing. Compared with the traditional ANOVA method, the ANFIS model achieved a significant improvement in the coefficient of determination from 0.84 to 0.9999, indicating a 19% increase. Additionally, the RMSE decreased from 1.96 to 0.0102, indicating the effectiveness of the ANFIS modeling phase. The integration of ANFIS and MPA led to a substantial increase in the percentage value of CO₂ uptake, improving it by approximately 30% in comparison to the experimental results and the RSM. Under these optimized conditions, the best values for the WSR and CT were found to be 0.27 and 54.3 h, respectively. Overall, this study demonstrates the effectiveness of the integrated ANFIS and MPA approach in improving the CO₂ uptake percentage for waste concrete powder. The findings provide valuable insights for optimizing the WSR and CT in the recycling process, leading to more sustainable and environmentally friendly practices in the construction industry.