Next Article in Journal
Piping Material Selection in Water Distribution Network Using an Improved Decision Support System
Previous Article in Journal
Strength Properties and Water-Blocking Stability of Hydrophobically Modified Silty Clay
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Water
Volume 17
Issue 3
10.3390/w17030341
water-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

Green Synthesis of nZVI-Modified Sludge Biochar for Cr(VI) Removal in Water: Fixed-Bed Experiments and Artificial Neural Network Model Prediction

by
Hao Zhao
1,
Fengfeng Ma
1,2,*,
Xuechang Ren
1,2,
Baowei Zhao
1,2,
Yufeng Jiang
1,2 and
Jian Zhang
1
1
School of Environmental and Municipal Engineering, Lanzhou Jiaotong University, Lanzhou 730070, China
2
Key Laboratory of Yellow River Water Environment in Gansu Province, Lanzhou Jiaotong University, Lanzhou 730070, China
*
Author to whom correspondence should be addressed.
Water 2025, 17(3), 341; https://doi.org/10.3390/w17030341
Submission received: 17 December 2024 / Revised: 16 January 2025 / Accepted: 23 January 2025 / Published: 25 January 2025
(This article belongs to the Section Wastewater Treatment and Reuse)
Browse Figures
Versions Notes

Abstract

:
The contamination of aquatic environments with hexavalent chromium (Cr(VI)) poses significant environmental and public health risks, necessitating the development of high-performance adsorbents for its efficient removal. This study evaluates the potential of green-synthesized nanoscale zero-valent iron-modified sludge biochar (TP-nZVI/BC) as an effective adsorbent for Cr(VI) removal through isothermal adsorption experiments, fixed-bed column studies, and artificial neural network (ANN) modeling. Fixed-bed experiments demonstrated that breakthrough time, exhaustion time, and unit adsorption capacity increased with bed height. Conversely, these parameters decreased with higher influent concentrations and flow rates. Breakthrough curve analysis revealed that the Thomas model provided the best fit for the experimental data (R2 = 0.992–0.998). An ANN model, developed using the Levenberg–Marquardt algorithm, employed a single hidden layer with six neurons and exhibited excellent predictive performance (R2 = 0.996, MSE = 0.520). The ANN model was validated for its ability to predict adsorption behavior under untested conditions, demonstrating its applicability for process optimization. This study highlights the superior performance of TP-nZVI/BC as an adsorbent for Cr(VI) and establishes a theoretical basis for optimizing and scaling up fixed-bed adsorption systems using ANN modeling. The findings provide valuable insights into the practical application of sustainable materials in environmental remediation.

1. Introduction

Chromium (Cr) is a transition metal element that is widely distributed in the Earth’s crust. Due to its extensive applications in industries such as electroplating, corrosion prevention, and leather tanning, the demand for chromium has significantly increased with industrialization, and the global annual production has reached 37.5 million tons [1,2]. However, the improper disposal and discharge of industrial waste have resulted in significant chromium contamination in aquatic environments, posing severe threats to ecosystems and human health [3]. Chromium mainly exists in two oxidation states, which are trivalent chromium (Cr(III)) and hexavalent chromium (Cr(VI)) [4,5]. While Cr(III) is relatively stable, has low solubility, and exhibits toxicity only at high concentrations [6,7], Cr(VI) is highly reactive, soluble, and mobile. These properties make Cr(VI) more bioavailable and hazardous, allowing it to penetrate biological membranes and accumulate in organisms through the food chain, where it induces genotoxic effects, including mutations and carcinogenesis [8,9,10]. Consequently, the increasing prevalence of Cr(VI) pollution necessitates the development of efficient technologies for its control and remediation to ensure water quality.
In recent years, biochar has gained significant attention as an adsorbent for Cr(VI) removal due to its high specific surface area and abundance of functional groups [11,12,13]. However, the reduction capacity of pristine biochar for Cr(VI) is limited, hindering its ability to effectively convert Cr(VI) into the less toxic Cr(III) [14,15]. Nano zero-valent iron (nZVI), a highly reductive and environmentally friendly nanomaterial, has demonstrated excellent performance in reducing Cr(VI) to Cr(III) [16]. Despite its potential, nZVI is prone to aggregation and oxidation during preparation, transport, and use, which reduces its electron transfer efficiency and limits its practical application [17]. To address these limitations, researchers have explored the integration of nZVI with biochar, leveraging the synergistic effects of these materials. The porous structure of biochar physically isolates nZVI particles, reducing aggregation, while nZVI enhances the reductive capacity of biochar, enabling combined adsorption and reduction [18,19]. Traditional chemical reduction methods for synthesizing nZVI often rely on expensive reductants such as borohydrides, which result in large particle sizes and low specific surface areas due to aggregation effects [20]. Alternatively, green synthesis methods using plant extracts, such as tea polyphenols (TPs), have attracted attention for their cost-effectiveness, environmental friendliness, and potential to produce nZVI with improved properties [21,22]. This study synthesized sludge biochar (BC) via a hydrothermal method and prepared a TP-nZVI/BC composite material using green synthesis techniques. Initial batch adsorption experiments confirmed the superior Cr(VI) removal performance of TP-nZVI/BC [23].
Currently, static adsorption experiments are mainly used for fundamental research on the properties of adsorbents, but they are limited in their ability to simulate the dynamic adsorption processes under actual industrial conditions [24,25]. Fixed-bed continuous adsorption experiments, as an important method for dynamic adsorption research, have the advantages of low energy consumption, ease of operation, and strong adaptability, as they allow for the flexible adjustment of operational parameters to achieve optimal removal performance [26,27,28]. By applying the TP-nZVI/BC composite material in a fixed-bed reactor, valuable reference information for the practical application of TP-nZVI/BC can be provided [29,30]. However, the application of fixed-bed reactors is influenced by multiple factors, such as bed height, flow rate, and influent concentration, and there are complex nonlinear interactions among these factors. Traditional mathematical models are often inadequate in fully describing this process.
Artificial neural network (ANN) models have been widely applied in the field of pollution control due to their strong nonlinear modeling and generalization capabilities [31,32]. This model does not require assumptions about complex physical mechanisms but instead uses statistical and computational algorithms to establish a mapping between input variables (such as operational parameters) and output results (such as adsorption efficiency), thereby enabling an accurate description of reactor performance [33,34,35]. Moreover, ANNs can not only predict the dynamic performance of fixed-bed reactors but also provide optimization recommendations for practical operations, reducing the reliance on experiments and improving research and application efficiency [36]. By integrating fixed-bed experiments with ANN modeling, it becomes possible to reveal nonlinear interactions between operational parameters, predict adsorption behavior under untested conditions, and optimize system design for industrial applications.
This study aims to evaluate the dynamic adsorption performance of TP-nZVI/BC for Cr(VI) removal under fixed-bed conditions and to develop an ANN-based optimization framework. The effects of bed height, influent concentration, and flow rate on adsorption efficiency were systematically investigated. Experimental results were fitted to traditional breakthrough models, including the Thomas, Yoon–Nelson, and Adams–Bohart models. Additionally, various ANN algorithms were compared with the Levenberg–Marquardt (LM) algorithm selected to construct a robust predictive model. The ANN model was validated through regression and importance analyses, highlighting its applicability for system optimization. This research not only advances the understanding of the dynamic adsorption behavior of TP-nZVI/BC but also provides a theoretical and technical foundation for the practical treatment of Cr(VI)-contaminated wastewater using optimized fixed-bed systems.

2. Materials and Methods

2.1. Materials

The sludge biochar (BC) used in this study was derived from the filter press dewatered wet sludge collected from the Qilihe Wastewater Treatment Plant in Lanzhou, China (103°44′53.27″ E, 36°5′38.09″ N). Tea polyphenols (TPs) were provided by Yueyuan Natural Company (Shenzhen, China). All solvents used in the experiments were deionized water.

2.2. Preparation of TP-nZVI/BC

To synthesize TP-nZVI/BC, a predetermined quantity of FeSO4·7H2O (Sinopharm Chemical Reagent Co., Ltd, Shanghai, China) was dissolved in 100 mL of deionized water. Subsequently, 0.2 g of the BC powder was added to the solution and stirred for 1 h (BC was prepared as described in Text S1). Tea polyphenols (1.2 g) were then introduced into the mixture, and stirring was continued for an additional hour. The solid and liquid components were separated via vacuum filtration, and the resulting solid was dried at 65 °C for 12 h, ground, and sieved through a 60-mesh screen to yield the TP-nZVI/BC composite material in powdered form.

2.3. Characterization of TP-nZVI/BC

The detailed characterization is given in the Supporting Information (Text S2).

2.4. Batch Adsorption Experiments

The adsorption properties of TP-nZVI/BC were explored by batch adsorption experiments following previously reported procedures [23]. The Supporting Information (Text S3) details the batch adsorption experimental procedure.

2.5. Fixed-Bed Experiments

Fixed-bed experiments were conducted using a transparent PVC column with an inner diameter of 25 mm and a height of 300 mm, as illustrated in Figure 1. To prepare the column, a layer of glass wool and quartz sand was first placed at the bottom to support the adsorbent. A specified amount of TP-nZVI/BC was then added and compacted, followed by an additional layer of quartz sand and glass wool to stabilize the adsorbent bed. The fixed-bed operating conditions were varied and effluent samples were collected periodically to monitor the concentration of Cr(VI) in order to construct breakthrough curves under various operating conditions.

2.6. Breakthrough Curve Modeling

The breakthrough curve of a fixed-bed adsorption system describes the variation in pollutant concentration (Ct) in the effluent over time, providing critical insights into adsorption performance. To analyze the mass transfer and kinetic behavior during Cr(VI) removal using the TP-nZVI/BC fixed-bed system, the experimental breakthrough curves were fitted using three commonly applied models, namely the Thomas model, the Yoon–Nelson model, and the Adams–Bohart model. The mathematical expressions of these models are provided in Table S2.

2.7. ANN Model

The ANN model was constructed using MATLAB R2022a software to predict the dynamic adsorption behavior of TP-nZVI/BC in fixed-bed systems based on experimental data. The dataset included 768 data points obtained from Cr(VI) removal experiments. Input variables for the ANN model included bed height, influent concentration, influent flow rate, and operation time, while the output variable was the normalized effluent concentration (Ct/C0).
The model was designed to ensure robustness and generalization by randomly splitting the data into training, validation, and testing sets. A maximum of 768 training cycles was applied. The ANN architecture incorporated connection weights, biases, activation functions, and summation nodes. Various algorithms were evaluated for optimal model construction, with the Levenberg–Marquardt (LM) algorithm ultimately selected due to its superior performance. The R2 and MSE are calculated in Text S4. The relative importance of input variables was assessed using the terminal weighted matrix of the ANN model. This method evaluates the overall contribution of each input variable by analyzing the absolute weights of connections between the input variables and the output layer across all neurons. The relative importance reflects the influence of each variable on the model’s predictions, offering insights into the dominant factors affecting the adsorption process.

3. Results and Discussion

3.1. Characterizations of TP-nZVI/BC

Figure 2 presents the SEM analysis results of BC and TP-nZVI/BC. As shown in Figure 2a, the surface of BC exhibits a rough and porous structure. In Figure 2b, after modification and loading, numerous nanoparticles are uniformly distributed on the surface and within the pores of TP-nZVI/BC. The nZVI particles display a clustered distribution, indicating successful loading of nZVI. Due to the natural organic compounds such as phenols, ketones, and aldehydes in TPs that coat the surface of nZVI, the aggregation and oxidation of nZVI particles are effectively inhibited. This coating exposes more active adsorption sites on the loaded nZVI, significantly enhancing the adsorption capacity of the material. The pore volume and pore size distribution diagram (Table 1 and Figure 2d) show the coexistence of micropores and mesopores in TP-nZVI/BC, with mesopores being predominant. This facilitates the physical adsorption properties of the material. Additionally, in the isothermal adsorption curve of TP-nZVI/BC (Figure 2d), the curve rises sharply when P/P0 > 0.8, indicating a strong adsorption of nitrogen by mesopores in TP-nZVI/BC, which is consistent with the pore size distribution.
As shown in Figure 2e, the peak near 2θ = 20–25° corresponds to an amorphous structure, while the peak near 2θ = 45.5° represents the characteristic peak of Fe. The appearance of these two peaks in the TP-nZVI/BC spectrum indicates that the active substances in TPs are coated on the surface of nZVI. After the reaction between TP-nZVI/BC and Cr(VI), the active substances dissolve, exposing the Fe0 [37]. Additionally, distinct diffraction peaks of FeCr2O4 and iron oxide were observed in the spectrum, suggesting that Cr(VI) was reduced to Cr(III) and adsorbed onto TP-nZVI/BC [18].
In the FTIR spectrum of TP-nZVI/BC shown in Figure 2f, absorption peaks are observed at 3409 cm−1 (-OH), 1138 cm−1 (C-O-C), 1637 cm−1 (C=O/C=C), 1695 cm−1 (-COOH), and 1397.5 cm−1 (C-H), indicating that the surface of TP-nZVI/BC contains abundant functional groups. These findings demonstrate the presence of surface functional groups and the coverage of organic molecules from TPs after loading. These functional groups, with their reductive and adsorptive properties, are beneficial for the removal of Cr(VI) [38]. Additionally, a prominent Fe-O peak is observed at 470 cm−1 in TP-nZVI/BC, confirming that nZVI was successfully loaded and uniformly distributed on BC, which enhances the Fe-O stretching vibration. After the reaction of TP-nZVI/BC with Cr(VI), the positions of the characteristic peaks remain largely unchanged, but their intensities decrease significantly, indicating that the functional groups participated in the removal of Cr(VI) and that nZVI reacted with Cr(VI). These results suggest that BC and nZVI in the composite material exhibit a synergistic effect in Cr(VI) removal.
The XPS spectra of TP-nZVI/BC before and after the reaction with Cr(VI) are shown in Figure 3. From the survey spectrum, Fe 2p peaks are observed both before and after the reaction, but the peak intensity decreases after the reaction. A Cr 2p peak appears near the binding energy of 577.0 eV, confirming that Fe⁰ was consumed during the reaction and Cr was adsorbed. The fitting results of the Cr 2p narrow spectrum indicate that the proportion of Cr(III) is as high as 80.94%, while Cr(VI) accounts for only 19.06%, demonstrating that the majority of Cr(VI) in the system was reduced to Cr(III). Three strong peaks are observed at binding energies of 576.8 eV, 577.7 eV, and 579.2 eV, indicating that Cr(III) is primarily adsorbed onto the material in the forms of Cr(OH)3 and Cr2O3. The C 1s spectrum is shown in Figure 3c, after the reaction, the relative content and binding energy of the C–O and O–C=O bonds decreased, while the relative content of the C–C/C=C bond increased, with the peak area ratio rising from 41.96% to 54.07%. These changes indicate that functional groups participated in redox reactions and ion exchange processes during the reaction, with H+ and Cr(VI) ion exchange likely playing a dominant role. Furthermore, Cr(VI) accounted for only 19.06% of the total chromium, suggesting that redox reactions between Cr(VI) and Fe0 or Fe(II) occurred.

3.2. Adsorption Isotherms

As shown in Figure 4, the adsorption capacity increases sharply with the equilibrium concentration of Cr(VI) at first, then gradually levels off. Table 2 indicates that the Freundlich model exhibits a higher fitting degree (R2 = 0.9915), suggesting that the adsorption process is primarily multilayer heterogeneous adsorption. In the fitting results of the Langmuir model, the Qm fitting value of TP-nZVI/BC (105.64 mg/g) deviates significantly from the actual value (73.69 mg/g), indicating that the adsorption process does not follow a monolayer adsorption mechanism [39]. In contrast, the R2 value of 0.9915 in the Freundlich model confirms the predominance of multilayer adsorption [23,40]. For the Dubinin–Radushkevich model fitting, the qm of TP-nZVI/BC for Cr(VI) is 17.11 mg/g, and the adsorption characteristic free energy (E) is 48.34 kJ/mol. These results suggest that the adsorption process involves strong intermolecular interactions, indicating a preference for chemisorption [41]. Moreover, TP-nZVI/BC exhibits high adsorption capacity for Cr(VI) even at low concentrations.

3.3. Effect of Operational Variables on Fixed-Bed Adsorption of Cr(VI)

3.3.1. Bed Height

The influence of bed height on fixed-bed performance is shown in Figure 5a. When the bed height increased from 4 cm to 10 cm, the tb extended from 2 h to 2.25 h and the te increased significantly from 10.5 h to 31 h. Additionally, the total amount of Cr(VI) removed by the fixed bed rose from 119.40 mg to 441.20 mg, and the Cr(VI) removal per unit mass of filler increased from 7.201 mg/g to 15.96 mg/g (see Table S3). These findings suggest that increasing the bed height significantly enhances the adsorption and reduction performance of the fixed bed. A greater bed height prolongs the mass transfer zone, effectively delaying both breakthrough and saturation. This increase in residence time provides more opportunities for Cr(VI) to undergo adsorption and reduction reactions. Consequently, the adsorbent utilization efficiency improves with increased bed height, aligning with similar observations in other studies [42,43].

3.3.2. Influent Concentration

The effect of varying influent concentrations on fixed-bed performance is shown in Figure 5b. When the C0 increased from 25 mg/L to 75 mg/L, both tb and te were significantly shortened. Meanwhile, the q decreased from 12.259 mg/g to 9.037 mg/g. The decline in q at higher influent concentrations can be attributed to the enhanced mass transfer driving force caused by the larger concentration gradient, which accelerates the diffusion of Cr(VI) to the adsorbent surface. However, the faster adsorption rate leads to earlier saturation of the fixed bed, reducing its removal efficiency per unit mass of filler. This behavior is consistent with the expectation that higher influent concentrations consume the adsorption and reduction capacities of TP-nZVI/BC more rapidly [44].

3.3.3. Flow Rate

Figure 5c illustrates the influence of different influent flow rates on fixed-bed performance. Increasing the flow rate caused tb to decrease from 12.25 h to 1.75 h and te to shorten from 32.75 h to 9.75 h. Simultaneously, the Cr(VI) removal per unit mass of filler decreased from 250.35 mg to 188.55 mg. Although a higher influent flow rate enhances the mass transfer driving force, it reduces the contact time between the solution and the adsorbent. This limited contact time prevents sufficient interaction between Cr(VI) molecules and the active sites on the TP-nZVI/BC surface, thereby reducing the adsorption efficiency. These results suggest that achieving optimal fixed-bed performance requires balancing the influent flow rate with the contact time to maximize Cr(VI) removal while avoiding premature saturation.

3.4. Breakthrough Curve Modeling

3.4.1. Thomas Model

The Thomas model fitting results, presented in Figure 6, demonstrate excellent agreement with experimental data, as evidenced by R2 values ranging from 0.992 to 0.998 (Table 3). An increase in Z led to a decrease in the kTH but an increase in the theoretical maximum adsorption capacity (Q0), which reached 16.437 mg/g. This behavior can be attributed to the higher adsorbent mass providing more adsorption sites and extending the contact time, thus enhancing the overall adsorption performance. Higher C0 values resulted in decreases in both kTH and Q0. This decline is likely due to the rapid approach of effluent concentrations to the exhaustion point at higher C0, which reduces the effective mass transfer driving force and lowers adsorbent utilization efficiency [26,45]. As the flow rate increases, kTh gradually rises while Q0 decreases. Although a higher flow rate enhances the mass transfer driving force, it reduces the solid–liquid contact time, leading to incomplete adsorption and lower capacity.

3.4.2. Yoon–Nelson Model

The fitting results of the model are shown in Figure 7. As Z increased, the kYN decreased, reflecting a slower adsorption process due to a prolonged retention time for Cr(VI) molecules in the fixed bed. The τ aligned well with experimental values, further confirming that increased Z enhances adsorption stability [46,47]. As the C0 increases, kYN increases, suggesting that a higher C0 enhances the mass transfer driving force, thereby accelerating the adsorption rate of TP-nZVI/BC in the fixed bed. However, faster adsorption results in earlier exhaustion of the fixed bed. Increasing the Q also leads to an increase in kYN, indicating that higher flow rates promote the mass transfer rate of Cr(VI) molecules to the surface of the adsorbent. Nevertheless, higher flow rates reduce the solid–liquid contact time, causing the fixed bed to exhaust more quickly due to incomplete adsorption.

3.4.3. Adams–Bohart Model

The fitting results of the Adams–Bohart model are shown in Figure 8. The R2 ranges from 0.897 to 0.956, indicating that the model accurately describes the early-stage adsorption behavior of the fixed bed, but its overall fitting accuracy is slightly lower than that of the previous two models. As Z increased, the kAB decreased, indicating a slower adsorption rate, while the N0 increased. This suggests that greater bed heights enhance adsorption by providing more active sites and extending the adsorption zone. Higher C0 values increased kAB, signifying enhanced mass transfer rates due to larger concentration gradients. However, N0 decreased, likely due to rapid site saturation at higher concentrations, leading to earlier bed exhaustion. Increasing Q raised kAB, reducing liquid-phase diffusion resistance and accelerating mass transfer. However, the reduced contact time at higher flow rates decreased N0, limiting adsorption efficiency [48,49].
The Thomas and Yoon–Nelson models accurately describe the Cr(VI) adsorption process in the TP-nZVI/BC fixed bed, making them suitable for designing and optimizing fixed-bed systems. The Adams–Bohart model, while less effective at capturing later breakthrough stages, provides critical insights into the initial adsorption phase, including the dynamics of mass transfer and site saturation. Optimizing key operational parameters such as bed height, influent concentration, and flow rate can significantly enhance adsorption performance. Increasing Z improves adsorption efficiency by offering more active sites and extending the contact time. The proper adjustment of C0 and Q ensures a balance between mass transfer driving forces and solid–liquid interaction, maximizing the adsorbent’s utilization and efficiency. While the Adams–Bohart model offers a foundational understanding of early adsorption kinetics, the Thomas and Yoon–Nelson models are more comprehensive in describing the overall breakthrough behavior.

3.5. ANN Model

3.5.1. ANN Model Optimization

The optimization results are shown in Figure 9. The GDA algorithm updates the parameters in the direction of the negative gradient, gradually approaching the local or global minimum of the loss function. The GDM algorithm introduces a momentum term based on the GDA algorithm, allowing it to reach the minimum more quickly and stably. However, the models built using the GDA and GDM algorithms both resulted in negative R2 values, indicating poor fitting performance and an inability to accurately describe the dynamic process of Cr(VI) removal by the TP-nZVI/BC fixed bed. This could be due to the slow response of the GDA and GDM algorithms to gradient changes, causing them to fall into local optima or suffer from slow convergence. The RBA algorithm accelerates convergence and avoids local minima by adaptively changing the update step size for each weight, but it ignores the size of the gradient. The SCG algorithm can overcome the limitations of local minima, requiring fewer iterations and providing higher computational accuracy. The BR algorithm uses Bayesian methods to optimize the regularization term of the loss function, which helps improve the generalization of complex networks. The LM algorithm combines the advantages of gradient descent and the Gauss–Newton method, enabling rapid convergence while avoiding local optima and thus performs exceptionally well in modeling complex nonlinear problems [50]. The ANN models trained with the LM, BR, RBA, and SCG algorithms all exhibited R2 values greater than 0.97, indicating good fitting performance. In particular, the model trained with the LM algorithm had the highest R2 and MSE, demonstrating its excellent performance in simulating dynamic processes. Based on the comparative analysis of different optimization algorithms, the LM algorithm was selected for optimizing the ANN model due to its higher fitting accuracy and lower error. Future research will use the LM algorithm to construct the ANN model for simulating the dynamic process of Cr(VI) removal by the TP-nZVI/BC fixed bed. This optimization significantly improves the predictive ability of the model and provides a reliable tool for optimizing fixed-bed adsorption performance and industrial applications.
The study also investigated the influence of hidden neuron counts (ranging from 2 to 20) on model performance. As shown in Figure 10, both R2 and MSE values stabilized when the number of hidden neurons reached six, demonstrating an optimal balance between computational efficiency and predictive accuracy. Therefore, as shown in Figure 10c, the final ANN structure selected consisted of four input neurons, six hidden neurons, and one output neuron. This structure achieved a good balance between computational efficiency and fitting performance. The weight matrix of the ANN model is presented in Table 4. The influent concentration of Cr(VI) was the most significant factor, contributing 49.33% to the output variability. Other influential factors include flow rate, time, and bed height, underscoring the dominant role of influent concentration in driving adsorption dynamics.

3.5.2. ANN Model Prediction Performance Analysis

The training process of the ANN model (Figure 11) illustrates the model’s optimization and validation performance. The gradient measures the magnitude of the effect of the error change on the weight adjustment, which is an important indicator of the optimization of the ANN model [51]. As the number of rounds of training, validation, and test data increases, the MSE value gradually decreases, indicating that the fit of the model continues to improve. The optimal validation performance of the model is determined by calculating the MSE for each training cycle. At the 26th iteration, the MSE value reaches the lowest value of 0.52 (Figure 11b), at which time the best validation performance is achieved, indicating that the model achieves high accuracy while avoiding overfitting [52,53]. Figure 11c shows the error distribution between the predicted values of the ANN model and the actual experimental values, and the errors are concentrated in the range of −0.354 to 0.379 with a small amplitude, indicating that the model has a high prediction accuracy for the adsorption of Cr(VI). In addition, the symmetry and concentration of the errors indicated that the model had a strong fitting ability and did not show systematic bias.
The effectiveness of the ANN model is determined by the R2. The performance was evaluated through regression analysis on the training, cross-validation, test, and overall datasets, with the results shown in Figure 12. The R2 value for the training set was 0.99484, indicating that the model fitted the training data well and demonstrated strong learning capability. The R2 value for the cross-validation set was 0.99750, suggesting that the model has good generalization ability and can effectively avoid overfitting. The R2 value for the test set was 0.99734, which indicates that the model also showed high prediction accuracy on previously unseen data. The overall R2 value for all datasets was 0.99537, further confirming the reliability of the ANN model in predicting the Cr(VI) removal performance of the TP-nZVI/BC fixed bed under various operating conditions.
The ANN model was used to predict the Cr(VI) concentration in the effluent of the TP-nZVI/BC fixed bed under 30 different operating conditions, and the predicted values were compared with experimental data. Figure 13b shows the distribution of residuals from the ANN model, with the maximum positive residual being 2.269 and the maximum negative residual being −1.591, indicating that the prediction error is small and there is no significant systematic bias. Figure 13c describes the deviation magnitude of the predicted data, with the maximum positive deviation being 5.488% and the maximum negative deviation being −4.596%, further confirming the high accuracy of the ANN model predictions. Overall, the combination of high regression values, small residuals, and minimal deviation further proves the outstanding capability of the ANN model in predicting the complex Cr(VI) removal behavior of the TP-nZVI/BC fixed bed. This model can effectively support the optimization and scaling-up of the fixed bed system and provides valuable theoretical and technical support for real-world wastewater treatment applications.

4. Conclusions

This study integrated fixed-bed experiments and ANN modeling to comprehensively evaluate the adsorption performance of green-synthesized nZVI modified sludge biochar (TP-nZVI/BC) for Cr(VI) removal. The findings demonstrate that TP-nZVI/BC exhibits remarkable adsorption capacity across diverse operating conditions, including bed height, influent concentration, and flow rate. Optimizing these parameters significantly enhanced adsorption performance. With an R2 of 0.996 and an MSE of 0.520, the ANN model provided high-precision theoretical support for optimizing the TP-nZVI/BC adsorption process. Under optimal operating conditions, the TP-nZVI/BC fixed bed achieved a maximum adsorption capacity of 16.437 mg/g, showcasing its potential for efficient Cr(VI) removal. Relative importance analysis identified influent concentration as the most critical factor (49.33%), followed by flow rate, time, and bed height. The ANN model demonstrated robust predictive performance, accurately forecasting adsorption behavior under untested conditions and highlighting its versatility for broader applications. In the kinetic analysis of the fixed-bed adsorption process, the Thomas model provided the best fit for the breakthrough curve (R2: 0.992–0.998), offering reliable insights into adsorption kinetics. This study elucidates the key factors influencing TP-nZVI/BC adsorption performance and identifies pathways for optimization, thereby providing valuable references for designing and operating Cr(VI)-contaminated wastewater treatment systems. Moreover, the successful implementation of ANN modeling underscores the potential of data-driven technologies in pollution control. Future research should expand investigations into the adsorption performance of TP-nZVI/BC for various pollutants and explore the application of ANNs in multi-objective optimization. This could facilitate the large-scale deployment and iterative advancement of adsorption materials, contributing to more efficient and intelligent solutions for environmental pollution control.

Supplementary Materials

The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/w17030341/s1, Text S1. Preparation of BC; Text S2. Characterization of TP-nZVI/BC; Text S3. Batch adsorption experiments; Text S4. ANN model; Table S1. List of the adsorption kinetic and isotherm models; Table S2. List of breakout curve models used in the study; Table S3. TP-nZVI/BC fixed-bed experimental parameters and experimental results.

Author Contributions

Conceptualization, F.M.; validation, F.M. and H.Z.; investigation, J.Z.; writing—original draft preparation, H.Z.; writing—review and editing, F.M.; visualization, B.Z.; supervision, X.R.; project administration, Y.J. All authors have read and agreed to the published version of the manuscript.

Funding

The authors gratefully acknowledge financial support from the Tianyou Youth Talent Lift Program of Lanzhou Jiaotong University, the Natural Science Foundation of Gansu Province (23JRRA874), and the Lanzhou Jiaotong University-Tianjin University Innovation Fund Project (2022069).

Data Availability Statement

Data will be made available upon request.

Conflicts of Interest

The authors declare that they have no know competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Abbreviations

The following abbreviations are used in this manuscript:
BCSludge Biochar
nZVINano Zero-Valent Iron
TPTea Polyphenol
TP-nZVI/BCGreen-Synthesized Nano Zero-Valent Iron-Modified Sludge Biochar
ANNArtificial Neural Network

References

  1. Li, C.; Wang, H.; Liao, X.; Xiao, R.; Liu, K.; Bai, J.; Li, B.; He, Q. Heavy metal pollution in coastal wetlands: A systematic review of studies globally over the past three decades. J. Hazard. Mater. 2022, 424, 127312. [Google Scholar] [CrossRef]
  2. Iyer, M.; Anand, U.; Thiruvenkataswamy, S.; Babu, H.W.S.; Narayanasamy, A.; Prajapati, V.K.; Tiwari, C.K.; Gopalakrishnan, A.V.; Bontempi, E.; Sonne, C.; et al. A review of chromium (Cr) epigenetic toxicity and health hazards. Sci. Total Environ. 2023, 882, 163483. [Google Scholar] [CrossRef] [PubMed]
  3. Rong, K.; Li, X.; Yang, Q.; Liu, Z.; Yao, Q.; Zhang, Z.; Li, R.; Zhao, L.; Zheng, H. Removal of aqueous Cr(VI) by green synthesized sulfide iron nanoparticles loaded corn straw biochar: Performance, mechanism, and DFT calculations. Appl. Surf. Sci. 2024, 670, 160729. [Google Scholar] [CrossRef]
  4. Ma, L.; Liu, T.; Li, J.; Yang, Q. Interaction characteristics and mechanism of Cr(VI)/Cr(III) with microplastics: Influence factor experiment and DFT calculation. J. Hazard. Mater. 2024, 476, 134957. [Google Scholar] [CrossRef] [PubMed]
  5. Monga, A.; Fulke, A.B.; Dasgupta, D. Recent developments in essentiality of trivalent chromium and toxicity of hexavalent chromium: Implications on human health and remediation strategies. J. Hazard. Mater. Adv. 2022, 7, 100113. [Google Scholar] [CrossRef]
  6. Sadeghi, P.; Kazerouni, F.; Savari, A.; Movahedinia, A.; Safahieh, A.; Ajdari, D. Application of biomarkers in Epaulet grouper (Epinephelus stoliczkae) to assess chromium pollution in the Chabahar Bay and Gulf of Oman. Sci. Total Environ. 2015, 518–519, 554–561. [Google Scholar] [CrossRef]
  7. Mallik, A.K.; Moktadir, M.A.; Rahman, M.A.; Shahruzzaman, M.; Rahman, M.M. Progress in surface-modified silicas for Cr(VI) adsorption: A review. J. Hazard. Mater. 2022, 423, 127041. [Google Scholar] [CrossRef]
  8. Wise, J.P.; Young, J.L.; Cai, J.; Cai, L. Current understanding of hexavalent chromium [Cr(VI)] neurotoxicity and new perspectives. Environ. Int. 2022, 158, 106877. [Google Scholar] [CrossRef] [PubMed]
  9. Pushkar, B.; Sevak, P.; Parab, S.; Nilkanth, N. Chromium pollution and its bioremediation mechanisms in bacteria: A review. J. Environ. Manag. 2021, 287, 112279. [Google Scholar] [CrossRef] [PubMed]
  10. Liang, J.; Huang, X.; Yan, J.; Li, Y.; Zhao, Z.; Liu, Y.; Ye, J.; Wei, Y. A review of the formation of Cr(VI) via Cr(III) oxidation in soils and groundwater. Sci. Total Environ. 2021, 774, 145762. [Google Scholar] [CrossRef]
  11. Aihemaiti, A.; Chen, J.; Hua, Y.; Dong, C.; Wei, X.; Yan, F.; Zhang, Z. Effect of ferrous sulfate modified sludge biochar on the mobility, speciation, fractionation and bioaccumulation of vanadium in contaminated soil from a mining area. J. Hazard. Mater. 2022, 437, 129405. [Google Scholar] [CrossRef] [PubMed]
  12. Fu, D.; Kurniawan, T.A.; Wang, Y.; Zhou, Z.; Wei, Q.; Hu, Y.; Hafiz Dzarfan Othman, M.; Wayne Chew, K.; Hwang Goh, H.; Gui, H. Applicability of magnetic biochar derived from Fe-enriched sewage sludge for chromate removal from aqueous solution. Chem. Eng. Sci. 2023, 281, 119145. [Google Scholar] [CrossRef]
  13. Mishra, R.K.; Mohanty, K. A review of the next-generation biochar production from waste biomass for material applications. Sci. Total Environ. 2023, 904, 167171. [Google Scholar] [CrossRef] [PubMed]
  14. Shen, C.; Gu, L.; Chen, S.; Jiang, Y.; Huang, P.; Li, H.; Yu, H.; Xia, D. Sewage sludge derived FeCl3-activated biochars as efficient adsorbents for the treatment of toxic As(III) and Cr(VI) wastewater. J. Environ. Chem. Eng. 2022, 10, 108575. [Google Scholar] [CrossRef]
  15. Ambika, S.; Kumar, M.; Pisharody, L.; Malhotra, M.; Kumar, G.; Sreedharan, V.; Singh, L.; Nidheesh, P.V.; Bhatnagar, A. Modified biochar as a green adsorbent for removal of hexavalent chromium from various environmental matrices: Mechanisms, methods, and prospects. Chem. Eng. J. 2022, 439, 135716. [Google Scholar] [CrossRef]
  16. Wang, P.; Hu, J.; Liu, T.; Han, G.; Ma, W.-m.; Li, J. New insights into ball-milled zero-valent iron composites for pollution remediation: An overview. J. Clean. Prod. 2023, 385, 135513. [Google Scholar] [CrossRef]
  17. Zhang, X.; He, Y.; Gao, G.; Zhang, X.; Han, R.; Xu, S.; Shi, X.; Xu, C.; Yan, G.; Wang, Z.; et al. Novel core-shell structural nZVI@Fe2P for sulfadiazine (SDZ) degradation: Accelerated Fe(III)/Fe(II) dynamic cycling, boosted Fenton performance and stability. Sep. Purif. Technol. 2024, 347, 127705. [Google Scholar] [CrossRef]
  18. Hu, Y.-b.; Du, T.; Ma, L.; Feng, X.; Xie, Y.; Fan, X.; Fu, M.-L.; Yuan, B.; Li, X.-y. Insights into the mechanisms of aqueous Cd(II) reduction and adsorption by nanoscale zerovalent iron under different atmosphere conditions. J. Hazard. Mater. 2022, 440, 129766. [Google Scholar] [CrossRef]
  19. Xi, M.; Zhang, X.; Chen, G.; Zhang, L.; Jiang, Z.; Zheng, H. Synergizing carbon sequestration mechanisms during the remediation of Cr(VI) by nano zero-valent iron loaded biochar (nZVI-BC). J. Environ. Chem. Eng. 2024, 12, 114781. [Google Scholar] [CrossRef]
  20. Deng, J.; Yoon, S.; Pasturel, M.; Bae, S.; Hanna, K. Interactions between nanoscale zerovalent iron (NZVI) and silver nanoparticles alter the NZVI reactivity in aqueous environments. Chem. Eng. J. 2022, 450, 138406. [Google Scholar] [CrossRef]
  21. Wang, X.; Zheng, Y.; Ning, P.; Lynch, I.; Guo, Z.; Zhang, P.; Wu, L. Synergetic effect of green synthesized NZVI@Chitin-modified ZSM-5 for efficient oxidative degradation of tetracycline. Environ. Res. 2024, 258, 119360. [Google Scholar] [CrossRef] [PubMed]
  22. Xu, C.; Zhou, S.; Song, H.; Hu, H.; Yang, Y.; Zhang, X.; Ma, S.; Feng, X.; Pan, Y.; Gong, S.; et al. Green tea polyphenols-derived hybrid materials in manufacturing, environment, food and healthcare. Nano Today 2023, 52, 101990. [Google Scholar] [CrossRef]
  23. Ma, F.; Zhao, H.; Zheng, X.; Zhang, J.; Ding, W.; Jiao, Y.; Li, Q.; Kang, H. Green synthesis of nZVI-modified biochar significantly enhanced the removal of Cr(VI) from aqueous solution. Environ. Sci. Pollut. Res. 2024, 31, 33993–34009. [Google Scholar] [CrossRef] [PubMed]
  24. Patel, M.; Karamalidis, A.K. Fixed-bed column adsorption of Ge(IV) using catechol-based adsorbents and aqueous complexation modeling to understand Ge(IV) selectivity. Sep. Purif. Technol. 2024, 351, 128106. [Google Scholar] [CrossRef]
  25. Nieto-Sandoval, J.; Morabet, F.E.; Munoz, M.; Lopez-Arago, N.; de Pedro, Z.M.; Casas, J.A. In-situ regeneration of a novel Fe3O4/GAC adsorbent for micropollutants removal in a continuous fixed-bed. J. Hazard. Mater. Adv. 2023, 10, 100267. [Google Scholar] [CrossRef]
  26. Bai, S.; Chen, S.; Li, J.; Ya, R.; Ao, N.; Wang, J. Effect of silicon on the mathematical model prediction and adsorption mechanism of boron removal by a fixed-bed column. J. Water Process Eng. 2022, 49, 103194. [Google Scholar] [CrossRef]
  27. Chatterjee, R.; Majumder, C. Application of modified graphene oxide-chitosan composite for the removal of 2-methylpyridine using fixed bed adsorption and subsequent regeneration of the adsorbent by UV photolysis. J. Water Process Eng. 2023, 53, 103654. [Google Scholar] [CrossRef]
  28. Diniz, V.; Rath, S. Adsorption of aqueous phase contaminants of emerging concern by activated carbon: Comparative fixed-bed column study and in situ regeneration methods. J. Hazard. Mater. 2023, 459, 132197. [Google Scholar] [CrossRef] [PubMed]
  29. Dalhat, M.A.; Mu’azu, N.D.; Essa, M.H. Generalized decay and artificial neural network models for fixed-Bed phenolic compounds adsorption onto activated date palm biochar. J. Environ. Chem. Eng. 2021, 9, 104711. [Google Scholar] [CrossRef]
  30. Vakili, M.; Mojiri, A.; Kindaichi, T.; Cagnetta, G.; Yuan, J.; Wang, B.; Giwa, A.S. Cross-linked chitosan/zeolite as a fixed-bed column for organic micropollutants removal from aqueous solution, optimization with RSM and artificial neural network. J. Environ. Manag. 2019, 250, 109434. [Google Scholar] [CrossRef] [PubMed]
  31. Gordanshekan, A.; Arabian, S.; Solaimany Nazar, A.R.; Farhadian, M.; Tangestaninejad, S. A comprehensive comparison of green Bi2WO6/g-C3N4 and Bi2WO6/TiO2 S-scheme heterojunctions for photocatalytic adsorption/degradation of Cefixime: Artificial neural network, degradation pathway, and toxicity estimation. Chem. Eng. J. 2023, 451, 139067. [Google Scholar] [CrossRef]
  32. Pauletto, P.S.; Lütke, S.F.; Dotto, G.L.; Salau, N.P.G. Forecasting the multicomponent adsorption of nimesulide and paracetamol through artificial neural network. Chem. Eng. J. 2021, 412, 127527. [Google Scholar] [CrossRef]
  33. Dragović, S. Artificial neural network modeling in environmental radioactivity studies—A review. Sci. Total Environ. 2022, 847, 157526. [Google Scholar] [CrossRef] [PubMed]
  34. Praveen, S.; Jegan, J.; Pushpa, T.B.; Gokulan, R. Artificial neural network modelling for biodecolorization of Basic Violet 03 from aqueous solution by biochar derived from agro-bio waste of groundnut hull: Kinetics and thermodynamics. Chemosphere 2021, 276, 130191. [Google Scholar] [CrossRef]
  35. Ahmed, Y.; Siddiqua Maya, A.A.; Akhtar, P.; Alam, M.S.; AlMohamadi, H.; Islam, M.N.; Alharbi, O.A.; Rahman, S.M. A novel interpretable machine learning and metaheuristic-based protocol to predict and optimize ciprofloxacin antibiotic adsorption with nano-adsorbent. J. Environ. Manag. 2024, 370, 122614. [Google Scholar] [CrossRef] [PubMed]
  36. Khan, A.A.; Naqvi, S.R.; Ali, I.; Arshad, M.; AlMohamadi, H.; Sikandar, U. Algal-derived biochar as an efficient adsorbent for removal of Cr (VI) in textile industry wastewater: Non-linear isotherm, kinetics and ANN studies. Chemosphere 2023, 316, 137826. [Google Scholar] [CrossRef] [PubMed]
  37. Huang, X.; Niu, X.; Zhang, D.; Li, X.; Li, H.; Wang, Z.; Lin, Z.; Fu, M. Fate and mechanistic insights into nanoscale zerovalent iron (nZVI) activation of sludge derived biochar reacted with Cr(VI). J. Environ. Manag. 2022, 319, 115771. [Google Scholar] [CrossRef] [PubMed]
  38. Zhou, Y.; Li, X. Green synthesis of modified polyethylene packing supported tea polyphenols-NZVI for nitrate removal from wastewater: Characterization and mechanisms. Sci. Total Environ. 2022, 806, 150596. [Google Scholar] [CrossRef]
  39. Yang, K.; Wang, X.; Lynch, I.; Guo, Z.; Zhang, P.; Wu, L. Green construction of MBI corrosion-resistant interfaces modified NZVI@MOFs-regulated 3D PAN cryogel film to enhance Cr(VI) removal. Sep. Purif. Technol. 2024, 333, 125902. [Google Scholar] [CrossRef]
  40. Zhao, N.; Zhao, C.; Liu, K.; Zhang, W.; Tsang, D.C.W.; Yang, Z.; Yang, X.; Yan, B.; Morel, J.L.; Qiu, R. Experimental and DFT investigation on N-functionalized biochars for enhanced removal of Cr(VI). Environ. Pollut. 2021, 291, 118244. [Google Scholar] [CrossRef]
  41. Puccia, V.; Avena, M.J. On the use of the Dubinin-Radushkevich equation to distinguish between physical and chemical adsorption at the solid-water interface. Colloid Interface Sci. Commun. 2021, 41, 100376. [Google Scholar] [CrossRef]
  42. Ordonez, D.; Podder, A.; Valencia, A.; Sadmani, A.H.M.A.; Reinhart, D.; Chang, N.-B. Continuous fixed-bed column adsorption of perfluorooctane sulfonic acid (PFOS) and perfluorooctanoic acid (PFOA) from canal water using zero-valent Iron-based filtration media. Sep. Purif. Technol. 2022, 299, 121800. [Google Scholar] [CrossRef]
  43. Bacelo, H.; Santos, S.C.R.; Ribeiro, A.; Boaventura, R.A.R.; Botelho, C.M.S. Antimony removal from water by pine bark tannin resin: Batch and fixed-bed adsorption. J. Environ. Manag. 2022, 302, 114100. [Google Scholar] [CrossRef]
  44. Albayati, T.M.; Kalash, K.R. Polycyclic aromatic hydrocarbons adsorption from wastewater using different types of prepared mesoporous materials MCM-41in batch and fixed bed column. Process Saf. Environ. Prot. 2020, 133, 124–136. [Google Scholar] [CrossRef]
  45. Feizi, F.; Sarmah, A.K.; Rangsivek, R. Adsorption of pharmaceuticals in a fixed-bed column using tyre-based activated carbon: Experimental investigations and numerical modelling. J. Hazard. Mater. 2021, 417, 126010. [Google Scholar] [CrossRef]
  46. de Araujo, C.M.B.; Ghislandi, M.G.; Rios, A.G.; da Costa, G.R.B.; do Nascimento, B.F.; Ferreira, A.F.P.; da Motta Sobrinho, M.A.; Rodrigues, A.E. Wastewater treatment using recyclable agar-graphene oxide biocomposite hydrogel in batch and fixed-bed adsorption column: Bench experiments and modeling for the selective removal of organics. Colloids Surf. A Physicochem. Eng. Asp. 2022, 639, 128357. [Google Scholar] [CrossRef]
  47. Chu, K.H.; Hashim, M.A. Removal of antibiotics through fixed bed adsorption: Comparison of different breakthrough curve models. J. Water Process Eng. 2023, 56, 104512. [Google Scholar] [CrossRef]
  48. Juela, D.; Vera, M.; Cruzat, C.; Astudillo, A.; Vanegas, E. A new approach for scaling up fixed-bed adsorption columns for aqueous systems: A case of antibiotic removal on natural adsorbent. Process Saf. Environ. Prot. 2022, 159, 953–963. [Google Scholar] [CrossRef]
  49. Zhang, D.; Zhang, K.; Hu, X.; He, Q.; Yan, J.; Xue, Y. Cadmium removal by MgCl2 modified biochar derived from crayfish shell waste: Batch adsorption, response surface analysis and fixed bed filtration. J. Hazard. Mater. 2021, 408, 124860. [Google Scholar] [CrossRef] [PubMed]
  50. Ahmad, T.; Chen, H. A review on machine learning forecasting growth trends and their real-time applications in different energy systems. Sustain. Cities Soc. 2020, 54, 102010. [Google Scholar] [CrossRef]
  51. Kavitha, B.; Deepa, R.; Sivakumar, S. Evolvulus alsinoides plant mediated synthesis of Ag2O nanoparticles for the removal of Cr(VI) ions from aqueous solution: Modeling of experimental data using artificial neural network. Mater. Today Sustain. 2022, 18, 100124. [Google Scholar] [CrossRef]
  52. Hou, J.; Bao, W.; Zhang, J.; Yu, J.; Chen, L.; Di, G.; Zhou, Q.; Li, X. Characteristics and mechanisms of sulfamethoxazole adsorption onto modified biochars with hierarchical pore structures: Batch, predictions using artificial neural network and fixed bed column studies. J. Water Process Eng. 2023, 54, 103975. [Google Scholar] [CrossRef]
  53. Yang, C.; Liu, K.; Yang, S.; Zhu, W.; Tong, L.; Shi, J.; Wang, Y. Prediction of metformin adsorption on subsurface sediments based on quantitative experiment and artificial neural network modeling. Sci. Total Environ. 2023, 899, 165666. [Google Scholar] [CrossRef]
Water 17 00341 g001
Figure 1. Fixed-bed experimental setup.
Figure 1. Fixed-bed experimental setup.
Water 17 00341 g001
Water 17 00341 g002
Figure 2. (a,b) SEM micrograph; (c,d) BET micrograph; (e) XRD images of TP-nZVI/BC; (f) FTIR images of TP-nZVI/BC.
Figure 2. (a,b) SEM micrograph; (c,d) BET micrograph; (e) XRD images of TP-nZVI/BC; (f) FTIR images of TP-nZVI/BC.
Water 17 00341 g002
Water 17 00341 g003
Figure 3. XPS spectra of TP-nZVI/BC: (a) the total absorption spectra; (b) Fe 2p; (c) C 1s; (d) O 1s; (e) Cr 2p.
Figure 3. XPS spectra of TP-nZVI/BC: (a) the total absorption spectra; (b) Fe 2p; (c) C 1s; (d) O 1s; (e) Cr 2p.
Water 17 00341 g003
Water 17 00341 g004
Figure 4. (a) Isothermal adsorption fitting curve of Cr(VI) adsorption on TP-nZVI/BC; (b) Dubinin–Radushkevich.
Figure 4. (a) Isothermal adsorption fitting curve of Cr(VI) adsorption on TP-nZVI/BC; (b) Dubinin–Radushkevich.
Water 17 00341 g004
Water 17 00341 g005
Figure 5. Effect of TP-nZVI/BC fixed-bed parameters on Cr(VI) removal: (a) bed height, (b) influent concentration, and (c) influent flow rate.
Figure 5. Effect of TP-nZVI/BC fixed-bed parameters on Cr(VI) removal: (a) bed height, (b) influent concentration, and (c) influent flow rate.
Water 17 00341 g005
Water 17 00341 g006
Figure 6. Thomas model fitting curves for TP-nZVI/BC fixed-bed operational data: (a) bed height; (b) influent concentration; (c) influent flow rate.
Figure 6. Thomas model fitting curves for TP-nZVI/BC fixed-bed operational data: (a) bed height; (b) influent concentration; (c) influent flow rate.
Water 17 00341 g006
Water 17 00341 g007
Figure 7. Fitted curves for the Yoon–Nelson model TP-nZVI/BC fixed-bed operational data: (a) bed height; (b) influent concentration; (c) influent flow rate.
Figure 7. Fitted curves for the Yoon–Nelson model TP-nZVI/BC fixed-bed operational data: (a) bed height; (b) influent concentration; (c) influent flow rate.
Water 17 00341 g007
Water 17 00341 g008
Figure 8. Results of the Adams–Bohart model fit to data from TP-nZVI/BC fixed-bed operation under different operating conditions: (a) bed height; (b) influent concentration; (c) influent flow rate.
Figure 8. Results of the Adams–Bohart model fit to data from TP-nZVI/BC fixed-bed operation under different operating conditions: (a) bed height; (b) influent concentration; (c) influent flow rate.
Water 17 00341 g008
Water 17 00341 g009
Figure 9. Test set data and model fitting results for different training algorithms. (a) LM, (b) BR, (c) RBA, (d) SCG, (e) GDA, (f) GDM.
Figure 9. Test set data and model fitting results for different training algorithms. (a) LM, (b) BR, (c) RBA, (d) SCG, (e) GDA, (f) GDM.
Water 17 00341 g009
Water 17 00341 g010
Figure 10. Number of hidden layer neurons versus MSE: (a) MSE, (b) R2; (c) model structure; (d) relative importance of each parameter in the TP-nZVI/BC fixed bed.
Figure 10. Number of hidden layer neurons versus MSE: (a) MSE, (b) R2; (c) model structure; (d) relative importance of each parameter in the TP-nZVI/BC fixed bed.
Water 17 00341 g010
Water 17 00341 g011
Figure 11. ANN analysis: (a) output target analysis, (b) MSE analysis, (c) error histograms.
Figure 11. ANN analysis: (a) output target analysis, (b) MSE analysis, (c) error histograms.
Water 17 00341 g011
Water 17 00341 g012
Figure 12. Regressivity of ANN models for TP-nZVI/BC fixed beds for (a) training set, (b) validation set, (c) test set, and (d) all datasets.
Figure 12. Regressivity of ANN models for TP-nZVI/BC fixed beds for (a) training set, (b) validation set, (c) test set, and (d) all datasets.
Water 17 00341 g012
Water 17 00341 g013
Figure 13. ANN-predicted Cr(VI) concentrations in fixed-bed effluent: (a) experimental versus predicted values, (b) residuals of predicted values, and (c) magnitude of prediction deviation.
Figure 13. ANN-predicted Cr(VI) concentrations in fixed-bed effluent: (a) experimental versus predicted values, (b) residuals of predicted values, and (c) magnitude of prediction deviation.
Water 17 00341 g013
Table 1. Specific surface area and porosity parameters of BC and TP-nZVI/BC.
Table 1. Specific surface area and porosity parameters of BC and TP-nZVI/BC.
Type Specific Surface Area
(m2/g)		Porosity
(cm3/g)		Aperture
(nm)
BC0.880.00321.50
TP-nZVI/BC4.870.01515.20
Table 2. Adsorption isothermal model fitting parameters for TP-nZVI/BC adsorption of Cr(VI).
Table 2. Adsorption isothermal model fitting parameters for TP-nZVI/BC adsorption of Cr(VI).
ModelParameter 1Parameter 2R2
LangmuirKL = 0.004Qm = 105.640.9834
FreundlichKF = 2.97N = 1.910.9915
TemkinA = 19.78Kt = 0.070.9326
Dubinin Radushkevichqm = 17.11E = 48.340.8070
Table 3. Results of fitting different models to TP-nZVI/BC fixed-bed operational data.
Table 3. Results of fitting different models to TP-nZVI/BC fixed-bed operational data.
ZC0QThomasYoon–NelsonAdams–Bohart
kTHQ0R2kYNτR2kABN0R2
cmmg/LmL/minmL/mg·minmg/g mL/mg·minmin mL/mg·minmg/L
650100.00997.2280.9980.536.230.9970.003170.740.897
850100.009410.8810.9960.4211.260.9900.0027110.270.922
1050100.004716.4370.9970.2323.170.9920.0017201.180.956
825100.010312.0350.9970.2622.160.9970.0018199.20.942
850100.009410.8810.9960.4211.260.9900.0027110.270.922
875100.00849.4340.9920.706.60.9930.004664.460.933
85050.004611.7440.9970.2321.730.9960.0014213.250.915
850100.009210.8810.9960.4211.260.9900.0027110.270.922
850150.01098.2260.9970.556.010.9970.003567.210.922
Table 4. ANN model weight matrix.
Table 4. ANN model weight matrix.
n1n2n3n4n5n6
Bed height (cm)−3.7112−5.11460.23231.1503−2.9781−5.1545
Influent concentration (mg/L)2.9673−22.55898.9515−1.94465.40003.1561
Inlet flow rate (mL/min)3.72023.60713.4719−3.10301.17742.4647
Running time (h)−7.4896−10.64341.7755−3.55413.18301.5486
Hidden layer to output layer weights−0.3011−0.5723−0.2447−0.29160.36410.1648
Hide layer bias items−8.500716.5916−6.9992−1.02951.11033.2168
Output layer bias term−0.2876
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

Zhao, H.; Ma, F.; Ren, X.; Zhao, B.; Jiang, Y.; Zhang, J. Green Synthesis of nZVI-Modified Sludge Biochar for Cr(VI) Removal in Water: Fixed-Bed Experiments and Artificial Neural Network Model Prediction. Water 2025, 17, 341. https://doi.org/10.3390/w17030341

AMA Style

Zhao H, Ma F, Ren X, Zhao B, Jiang Y, Zhang J. Green Synthesis of nZVI-Modified Sludge Biochar for Cr(VI) Removal in Water: Fixed-Bed Experiments and Artificial Neural Network Model Prediction. Water. 2025; 17(3):341. https://doi.org/10.3390/w17030341

Chicago/Turabian Style

Zhao, Hao, Fengfeng Ma, Xuechang Ren, Baowei Zhao, Yufeng Jiang, and Jian Zhang. 2025. "Green Synthesis of nZVI-Modified Sludge Biochar for Cr(VI) Removal in Water: Fixed-Bed Experiments and Artificial Neural Network Model Prediction" Water 17, no. 3: 341. https://doi.org/10.3390/w17030341

APA Style

Zhao, H., Ma, F., Ren, X., Zhao, B., Jiang, Y., & Zhang, J. (2025). Green Synthesis of nZVI-Modified Sludge Biochar for Cr(VI) Removal in Water: Fixed-Bed Experiments and Artificial Neural Network Model Prediction. Water, 17(3), 341. https://doi.org/10.3390/w17030341

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop