Abstract
Innovative technologies enable businesses to stay competitive in marketplace while also increasing profits in manufacturing of nanocomposites for water treatment. Aforementioned driving factors resulted in adoption of a number of innovative technologies, and no other trend has had a greater impact in recent years than Industry 4.0. Industry 4.0 is a comprehensive term that encompasses data management, manufacturing competitiveness, manufacturing processes, and efficiency. The term “Industry 4.0” refers to a group of key enabler technologies, such as cyber physical models, Internet of Things (IoT), artificial intelligence (AI), and big data analytics, including embedded devices which are all significant components to the mechanized and digitized industrial environments. AI approaches have been used for water treatment processes as well as desalination in recent years for optimizing the process along with providing realistic answers to water scarcity and water pollution-related issues. AI applications have been used to predict and minimize water treatment process operational costs by lowering costs and optimizing chemical utilization. Several AI models are successful and accurate in predicting effectiveness of various adsorbents used in the removal process of a variety of contaminants from water. To identify the current level of research and future development prospects of smart manufacturing, this study uses a comprehensive literature review technique for manufacturing sustainability of nanocomposite fabrication for water treatment applications. The model provided will help to create a baseline for AI and hybrid models in the water treatment and wastewater management sectors, allowing for the increased performance and innovative growth. It will serve to provide the framework and give guidance for researchers interested in creating superior nanocomposites for waste and fresh water treatment and management using Industry 4.0. This study looks at a variety of AI approaches as well as how they may be used in water treatment, with an emphasis on pollutant adsorption. This assessment also identified certain obstacles and research gaps in the field of AI applications in water treatment processes.
1. Introduction
Industry 4.0, the fourth industrial revolution, has had a huge impact on academia, government policymaking, and the industrial sector. Industry 4.0 is being studied extensively for its potential to improve business models, product quality, employee skills, communications, and supply chains [1]. As per the definition of Industry 4.0, it is “a name for the current trend of automation and data exchange in industrial technologies, spanning cyber-physical systems, the Internet of things, cloud computing and cognitive computing, and the construction of the smart factory” [2, 3]. Businesses, regardless of industry, will need to transition to cloud-based technology, which also will play a vital part in the development of the Fourth Industrial Revolution. Cloud computing enables businesses of all sizes to develop and adapt to the changes more quickly, providing them access to all of the benefits that Industry 4.0 will provide. Organizations may harness the power of technologically advanced applications and solutions to expand their operations by utilizing cloud-based technologies. The goal is to enable autonomous decision-making processes, real-time asset and process monitoring, and real-time linked value creation platforms through early stakeholder engagement and vertical and horizontal connectivity [4, 5]. Many technologies are related with the fabrication of nanocomposites and thin-film membranes for water treatment and wastewater cleaning purposes in the framework of Industry 4.0 [6]. Water pollution is exceedingly difficult to reverse. Natural water filtration procedures can take years, decades, or even centuries to eliminate all hazardous substances from the water, and even the most costly contemporary treatments can take years [7, 8]. There are two kinds of pollutants to be handled. Water can be contaminated by insecticides and herbicides, food processing waste, toxins from livestock operations, volatile organic compounds (VOCs), heavy metals, chemical waste, and other substances. Nutrient pollution (nitrogen, phosphates, etc.) causes an abundance of poisonous algae, which is ingested by other aquatic species and can be lethal; it can also cause fish disease outbreaks. To avoid additional contamination, the source of the water pollution must first be eradicated. Pesticides, cleaners, and home chemicals may not appear to be a major concern. However, a variety of chemicals, including bleach, paints, paint thinner, and ammonia, are becoming a major concern. Organic foods can be processed with chemicals; however, they are normally made with little synthetic chemicals. The use of organic foods reduces the amount of chemical pollution that reaches the water supply. Cities may enhance their wastewater treatment systems in order to reduce water pollution [9, 10]. Because the toxic substances created in wastewater streams have such a negative influence on human existence, the separation along with purification of water as well as wastewater streams has been of critical relevance for several countries globally. Moreover, toxic substances produced by chemical operations can cause catastrophic harm to water bodies. Ecotoxicological effects of textile wastewater containing dyes on the environment and living things are quite harmful. When azo dyes are present in the water, they prevent light from penetrating, which has an impact on the growth of algae and aquatic plants. The absorption and reflection of sunlight by dyes into water is a serious environmental issue. Light absorption has a huge influence on the food chain because it reduces the photosynthetic activity of algae, which impacts all creatures above them because algae are at the bottom of the food chain. To remove pollutants from water and wastewater streams, efficient processes and techniques must be applied from an environmental standpoint [11]. Many contaminants can be found in water as well as wastewater streams but also effluents, depending on the quality of water as well as region. Organic materials along with pharmaceutical waste and dyes and greasy waste, and heavy metals are all examples of pollutants. Thin-film nanocomposite (TFN) membranes found extensive use in water treatment processes like reverse osmosis (RO) as well as nanofiltration (NF) processes because they provide high flux as well as rejection along with high mechanical stability [12, 13]. The reverse osmosis (RO) and nanofiltration (NF) membrane filtration technologies are intended to remove solute dissolved substances from a pretreated liquid stream. Well-designed RO/NF systems are compact and need minimal maintenance, making them a feasible replacement to standard treatment trains [14, 15]. Fouling occurs when contaminants accumulate on the surface of a filter membrane and hinder the passage of water through the membrane’s pores [16, 17]. Various parameters like thickness, hydrophilicity, porosity distribution, density, and surface zeta potential are all altered when such nanoparticles are introduced into the PA layer. Membranes can gain positive features such as high adsorption, improved antibacterial activity, and higher photocatalytic activity, along with good antifouling properties by using the right nanoparticles. It is critical to discover the appropriate production parameters for TFN membranes that enhance water permeability and salt rejection before they can be used on a large basis [18]. TFN membrane development entails a slew of time-consuming and expensive laboratory procedures involving poor separation performance and extensive surface chemical analysis, along with degrading membrane stability. Estimation of membrane performance with time is required for process analysis and optimization, as well as the design of new membrane processes. Artificial intelligence has recently been employed and shown the great accuracy for membrane better separation, adsorption, higher fluid dynamic reaction, and separation applications. Artificial neural networks (ANNs) are the most prevalent artificial intelligence modelling approach, and they have been widely utilized to simulate chemical and physical processes. The adsorption equilibrium concentration of pollutants and the adsorption capacity of the adsorbent as more than just a function of adsorption variables may very well be simulated through AI technology such as solution pH or a solid dose, solute starting concentration, and medium temperature [19]. The optimal weights for an optimization strategy are used to find an ANN model that minimizes a performance index during training. The universal approximation theorem states that a multiple-layer-based feedforward neural network framework integrates a single hidden layer and capable to approximate any continuous and nonlinear mapping function. The number of neurons of the hidden layer with initial weights and biases and optimization strategy along with other parameters should be chosen carefully because they have significant impact on ANN models’ working. Machine learning (ML) approaches are generally helpful tools, which have been increasingly employed in several systems with complicated nature, given a huge number of factors and complexity of membrane processes [20, 21]. ANNs have been used in the manufacture and separation of various membranes, including NF, UF, RO, and MF. Nonlinear functions can be used to detect complex correlations between input and output data using machine learning approaches. Artificial neural networks (ANNs) are increasingly being used in water treatment applications due to their ability to represent complex processes. The current work proposed an approach for developing and verifying ANNs for applications requiring drinking water quality and water distribution systems. ANNs are not dependent on biological, pharmacological, or physical processes. Instead, these empirical models recognize that, given certain inputs, some processing conditions may be modified to achieve a specific goal based on “learning” from earlier training. However, the performance of the ML method’s estimation and generalization capabilities is dependent on the training method along with architecture used. Through a training phase, ANN link input and output data have been using a set of weights among neurons along with biases and linear and nonlinear functions. The other important applications of AI and ML in Industry 4.0 for manufacturing of nanocomposites are shown in Figure 1.
A lot of research has been done in the field of manufacturing of nanocomposites by Industry 4.0 and the concern of making Earth a safer and cleaner place to live for us a future generation has constantly been raised and discussed [22]. The manufacturing of nanocomposite-based membrane for removing wastewater and chemical in Industry 4.0 using ML and DL has been reported by many researchers [23, 24]. However, a systematic review of the efficiency of the proposed ML DL techniques has not been discussed in the recent times. The work presented here offers a promising mesoporous material with a short contact time and high efficacy for treating wastewater contaminated with high concentrations of unitary and binary dyes. Additionally, the ANN technique may hold great a promise for applications involving the removal of pollutants. The material has a great deal of promise for usage in wastewater treatment facilities. Additionally, even though it requires operators to analyze complex data, AI may aid operators by making sophisticated intelligent decisions that enhance the accuracy and dependability of the treatment system.
2. State-of-the-Art Review of Industry 4.0-Mediated Nanocomposite Manufacturing for Wastewater Treatment and Management
2.1. Machine Learning-Based Frameworks
Mostly in adsorption process for aqueous solution treatment, numerous parameters may be employed to control the mass transfer rate as well as separation. The affinity between molecules and adsorbent surface is normally regulated by functional groups linked to the adsorbent surface and is the critical parameter which determines adsorption rate. As a consequence, LI et al. [25] applied the genetic algorithm to optimize these values. Indeed, in this study, the authors employed several ML models for forecasting adsorption data for separation process which removes water impurities. The data required for modelling purposes is derived from literature which has been utilized for training as well as validation purpose in order to develop a robust process model/framework.
The authors present three basic regression models which have been strengthened in the presented study using AdaBoost method [26, 27]. One of the most basic regression techniques is linear regression. Bayesian ridge regression uses Bayesian knowledge to handle the problem of highly correlated independent variables in the linear regression model along with calculating regression coefficients as well as selecting variables. To achieve optimum parameters in the Bayesian ridge regression model, the GA algorithm has been applied with a number of different generations. According to the present results, all the optimal parameters remain the same or even stable after 30 generations, producing the same precision. AdaBoost (also known as adaptive boosting) is an assembling strategy used for improving weak estimators for building a stronger as well as more accurate regressor for predicting or forecasting water treatment-related processes. This procedure begins by fitting a regression model to the raw data values and then fitting a replica of the regressor to the same data, although with the weights of the instances varying relying on the forecast’s accuracy. Figure 2 compares the created models’ final outcomes with several criteria of regression accuracy. It has been observe that Huber regression found to be the best suited model among other three AdaBoost-enhanced models. The most commonly used approach with AdaBoost is decision trees with one level or decisions trees with only one split. These trees are also known as Choice Stumps. On the classification front, these models provide results that are just marginally better than random chance. Decision trees with one level operate well with AdaBoost and are hence the most commonly used approach. A set of the machine learning models were used to estimate the adsorption capacity of a proposed composition of a nanocomposite material. One of the case studies involves adsorption of a specific pollutant present in water under various settings achieved by varying pH of the solution along with adsorbent dose. The genetic algorithm has shown to produce models which has output having adequate generality of the scenario and using models with minimal complexity as well as able to pick a fitting function relayed on no overfitting models. The evolving generational cycle is used by the genetic algorithm to deliver great responses. These algorithms use a variety of procedures to enhance or replace the population in order to achieve a better fit outcome. It is used in machine learning to solve optimization problems. It is an important algorithm since it allows for the speedy and efficient resolution of difficult problems.
2.2. AI for Nanocomposite Membrane for Filtration
Fetanat et al. [18] present a unique machine learning technique for estimating permeate flow as well as foulant rejection by membranes made of nanocomposite filtration material. Artificial neural networks have been provided with nine independent variables (ANNs). The suggested technique was tested for 2 datasets which includes training, validation, and testing of datasets, even for unknown dataset. To identify the optimum ANN models, 2250 various starting weights as well as a number of neurons present in the hidden layer in the suggested ANN models have been evaluated and compared.
Thin-film nanocomposite (TFN) membranes have been frequently employed in reverse osmosis (RO) and nanofiltration (NF) of water and wastewater treatment processes because they provide high flux along with higher rejection while being mechanically stable. TFN membranes have been generally created by generating single polyamide layer (PA) on the top of substrate made of porous polymer by using an interfacial polymerization (IP) technique. This results in polycondensation reaction between one monomer present in form of aqueous solution and another monomer present in organic solution [28, 29]. Thin-film composite membranes and thin layer nanocomposite membranes are two types of membranes used for desalination and other applications (TFNs). TFNs are superior than TFCs generated previously using interfacial polymerization (IP). The alteration comprises adding nanoparticle to a thin polyamide (PA) thick layer on top of the membrane in order to improve the performance of the TFC membrane. This enhancement might take numerous forms, including greater water porosity and solute rejection [30, 31].
For TFN membranes to be widely used on industrial sizes, it is critical to discover the optimal manufacturing parameters for maximizing water permeability coupled with higher salt rejection. Currently, formulation of TFN membranes entails a slew of time-consuming and expensive laboratory research involving separation performance testing, surface chemical analyses, and membrane stability. Estimating the results of membrane processes in general is required for critical process analysis as well as optimization along with the designing of novel membrane manufacturing processes. In support of this context, theoretical modelling tools allow us to simulate membrane composition manufacturing processes to aid in the designing more efficient and cost-effective processes. The goal of this study is to provide a technique for formulating permeate flux along with higher foulant rejection in nanocomposite filtration membranes which shall be supported by experimental factors, as well as to examine the impacts of individual experimental condition on performance of nanocomposite filtration membranes.
The goal of the author is to estimate permeable flow along with foulant rejection using data available from genuine experiments described in the literature thereby reducing time as well as expenditures associated with experiments. The author suggested an ANN architecture that included input along with output variables, and the input, hidden, and output layers were coupled within the ANN models. As the ANN model outputs, nine input variables have been chosen for predicting the permeable flow as well as foulant rejection of nanocomposite filtering membranes. The following parameters were fed into the ANN model: support followed by a particle concentration and then organic phase trimesoyl chloride (TMC) in n-hexane (TMC in n-hexane) including operation pressure and contact angle measurement along with thin layer thickness coupled with location of the NPs within the filtration membrane, posttreatment temperature, and posttreatment duration. The ANN model weights coupled with biases have been chosen randomly, and the backpropagation training procedure was used to update the ANN weights and biases to minimize output error.
The neurons in the hidden layer have been set between 5 and 50, and 80%, 10%, and 10% of the datasets are chosen at random for training, validation, and testing, respectively. The technique was performed 50 times for training the ANN models with varying beginning weights. BR training algorithm has been used for training approach in all of the runs. The activation functions for the hidden layer were set to hyperbolic tangent sigmoid as well as linear for the output function, while the normalization technique has been set to min and max displays of the MSE (Figure 3) of the ANN models’ training, validation, and testing of datasets for estimating permeate flow and foulant rejection dependent on the number of neurons present in the hidden layer of the framework.
The results shown in Figure 4 present that the ANN models are effective approaches for estimating permeate flow and foulant rejection with good accuracy followed by generalizability on training, validation, testing of unknown datasets.
The most relevant input factors were discovered to be posttreatment temperature and contact angle. The suggested technique may be utilized for calculating permeable flux and least foulant rejection, as well as assessing the impacts of independent experimental condition for nanocomposite filtration membrane performance without conducting time-consuming and expensive real-world tests.
2.3. AI Simulation of Water Treatment Using Nanocomposite
Syah et al. [28] created an artificial intelligence simulation approach for predicting adsorption processes for ion removal of water using an ordered manufactured nanostructured adsorbent. The constructed artificial intelligence model was used to mimic the separation of two ions, including Ni and Hg followed by prediction of adsorbent equilibrium concentration as well as adsorption capacity of the membrane in the elimination of Hg and Ni from aqueous solutions. The dataset has been collected in preparation for the adsorption capacity of nanostructured adsorbent for these two ions. The suggested ANN model with two hidden layers having two inputs includes ion type and starting ion concentration. The models predict the equilibrium concentration () in the solution and the adsorption capacity () of the adsorbent as shown in Figure 5.
(a)
(b)
The training and validation results of the ANN model are shown in Figure 5 [32]. In addition, the statistical data which have been used for the fitting are shown in Figure 5 for the simulation of and . The simulated data and measured values were found to have a higher coefficient of determination (). It was also found that the computed for all situations is more than 0.999 thereby proving model’s validity in simulation of adsorption isotherm and adsorption equilibrium. The findings showed that the proposed machine learning approach can create adsorption isotherms with significantly higher accuracy than the well-known Langmuir model which is an adsorption isotherm model.
Pollutant separation followed by removal from water/wastewater is a critical process which provides environmental protection. Environmental protection can be accomplished using a variety of extraction techniques such as by using adsorption membranes, use of an enhanced oxidation, and biological degradation techniques. Several water impurities such as heavy metals and inorganic and organic contaminants shall be reduced to a safe level. Wei et al. [29] implemented the machine learning technique for a comprehensive simulation strategy for predicting adsorption kinetic data for filtration process such as ion separation from aqueous media. A ML model has been implemented using ANN for predicting adsorption kinetics of Pb and Cd ions from water for ion removal process. The adsorbent under investigation was found to be a metal organic framework coupled with composite having layered double hydroxide function groups.
The ANN model has ion type of Pb and Cd ions as an input component, coupled with absorption duration, and adsorption capacity as the major output (). The simulations were statistically analyzed, and consequently, the results for training as well as validation are presented in Figure 6.
As can be observed, 11 data points were utilized for training and 5 data points were used for validation. The high value of in an indication of the well-constructed model accurately fits the adsorption kinetic data for Pb and Cd ions. It is clear from the results that the created ANN model found to be capable of forecasting adsorption kinetic data and even can be used to accurately predict adsorption kinetics as well as separation end point temporal data. The presented neural model outperforms typical empirical models that have been used for kinetic data fitting, which includes first-order as well as pseudo-first-order models. The findings showed that the ANN model has been capable in accurately forecasting kinetic data which can be utilized to determine the final point of adsorption separation.
Environmental issues stemming from dye wastewater contamination have become more apparent in recent years. Toxicity and bioaccumulation are common in dye stuffs having complex aromatic benzene ring structures, making them difficult to breakdown under natural circumstances. Dye wastewater contamination is regarded as one of humanity’s greatest issues because of its high chromaticity and limited biodegradability along with high toxicity to human, plants, and animals. When compared to ion exchange technology and other technologies such as coagulation and advanced oxidation along with electrochemical and membrane separation technologies, adsorption is found to be a successful approach in the treatment of wastewater filled with dye colors in terms of simplicity, cheap cost, ease of operation, and many other benefits.
With the use of AI technology, the removal of contaminants from the aquatic environment may be considerably enhanced. Several techniques of AI including backpropagation (BP) technique have been used to create a forecasting mathematical model having four variables, namely, dye concentration, pH of solution, time required for reaction, and dosage, in this study (Figure 7). The triple layered network has been featured with an input, a hidden, and an output layer. The hidden input layer tends to have a tangential S-shaped transfer function, where the hidden output layer is coupled to a linear transfer function.
The experimental data was optimized and forecasted using ANN-PSO, where the anticipated conditions were confirmed. A radial basis network (RBF) is a type of multilayer forward neural network. The four key parameters, including reaction time, pH, dose, and dye concentration, were assessed as input data using Autorr, and the percentage of carmine elimination was calculated using an RBF as output data.
The response surface BBD model projected a maximum percentage of decontamination of 95.8%, with a comparable experimental result of 93.28 percent. The ANN-PSO model was found to be effective in predicting carmine decontamination occurred due to nanohybrids of GO/Fe/Cu ions. As a result, the material has shown a high level of carmine decontamination efficiency shown in Figure 8.
This study generated a promising mesoporous material with a small contact time along with high effectiveness for treating wastewater with high concentration contamination of unitary as well as binary dyes, and the ANN technique might have a lot of promises for pollutant removal applications. The material has a lot of potential in terms of being used in wastewater treatment systems.
The temporal diversity of Th4+ adsorption was modelled using an artificial neural network (ANN) by Broujeni et al. The adsorbent weight and contact duration and pH values have been taken into account as operating condition parameters. 144 experiments were used to build, train, validate, and test various ANN topologies. Based on neurons present in the hidden layer, the training procedure, activation function, statistical errors, and the ideal three-layer MLP structure for predicting Th4+ elimination have been found after determining the optimized structure of MLP network in predicting final Th4+ ions. A genetic algorithm has been implemented to gain global efficiency. The optimization process has been carried out using various initial datasets, generation sizes, multiple crossover, and mutation along with selection functions for the purpose of gaining global efficiency. According to the findings, the optimum network’s Levenberg-Marquardt (LM) training method comprises four neurons present in the hidden layer along with logsis found to be the best activation function.
The model tends to reduce MSE with an increasing number of epochs up to 12 epochs. The MSE for the experimental, training, and validation datasets is stabilized using 12 epochs. This emphasizes the proposed network’s training process’s utilization of 12 epochs. The statistical constraints show that variation between the observed and projected data is inconsiderable, showing that the model has a lot of potential. Following that, the GA has been used to determine the best conditions required to achieve maximal Th4+ ion removal, which projected to be 98% removal at pH value of 6.5 and adsorbent weight of 0.2 g/L, along with a contact time of 64 minutes. Furthermore, the adsorption kinetic is found to be well matched to the kinetic model of pseudo-second-order, indicating that chemical processes are involved in adsorption.
Recently, mesoporous silica materials (MSM) for environmental applications have shown dramatic development. Several mesoporous silica-based materials are produced as well as tested for the process of adsorption which includes KCC-1 having a fibrous structure and a large area of surface for pollutant removal from aqueous solution. Computational studies may be utilized to anticipate the behavior of the proposed adsorbents based on empirical and semiempirical formulas for a deeper comprehension and development of adsorption synthetic materials. Artificial intelligence techniques may be utilized in adsorption studies to examine the impact of a dependent variable on removal efficiency, in addition to empirical and molecular-level modelling tactics. This approach entails utilizing measured adsorption data to train a network and then using the learned model to forecast the process. The approach has a high degree of accuracy, but it requires measurable data to train.
Saravanan et al. [30] modelled the adsorption capacity of a composite nanocomposite in the removal of Cd(II) ions for water treatment process, a simulation approach based on ANN. The focus has been on developing a composite consisting of KCC-1 silica and polyamide 6 (COOH-KCC-1/PA6) as a functional group. The simulations were run in order to forecast the equilibrium adsorption capacity (). Adsorption capacity () was shown to have a relationship with pH and adsorbent dose (). The data on adsorption was gathered from the literature, and for batch studies, the framework is created with double hidden layers having one input layer, with the other as the output layer to study adsorption capacity of functional groups in removing Cd(II) ions from the solution. The input and output layers, as well as the hidden layers, are linked together. Adsorption dose () and solution pH are two inputs, whereas adsorbent capacity, , is the output anticipated by the constructed network. This ANN topology’s function combination was determined by a trial-error process. The presented topology suggested best performance with fewest variations available.
Table 1 shows the results of the training as well as validation phases. As previously stated, excellent arrangement was found for the training and validation model as well as high more than 0.99 along with SSE below than 0.5. The constructed neural network model has been verified and found to be valid for simulating and fitting the experimental data of Cd(II) ions adsorption. The impact of solution pH found to be more significant when compared to adsorbent dose, according to the model. It was discovered that the best adsorption values have been obtained at a pH of (6), indicating the greatest . This might be due to exchanges between adsorbent surface and solutes resulted from the changing pH of the solution. A change in pH solution results in variation of electrostatic forces of the system’s components. The model revealed that the pH of the solution had a greater impact on the removal of Cd(II) ion with the use of composite adsorbent, causing the highest adsorption occurring at pH 6.
3. Discussion
In comparison to traditional mathematical modelling, AI techniques offer a number of advantages: utilized to forecast the efficacy of various water treatment processes, lowering the cost of testing. There are certain constraints that have prevented these approaches from being widely used in industrial applications. Operational dataset from actual facilities of water treatment has been utilized as input for the AI models to improve accuracy of pollution removal predictions. These latest advances in AI technology are benchmarks and have a key role in a long term for treatment of wastewater, resulting in considerable cost savings while also protecting the environment.
Artificial intelligence-based technologies, particularly artificial neural networks (ANNs), can successfully discover the most significant parameters that impact nanocomposite production and function. Predicting system performance can also help in averting incidents and lowering the health and safety concerns associated with any potential system breakdown. AI may help operators by making advanced intelligent judgments that improve the accuracy as well as reliability of the treatment system despite requiring the operators to analyze complicated data. The existing literature has clearly demonstrated that applying AI to such systems may improve their performance and, as a result, lead to their speedy commercialization.
4. Conclusion
Due to their capacity to represent complicated processes, artificial neural networks (ANNs) are being employed more and more in water treatment applications. For applications needing drinking water quality and water distribution systems, the current study suggested a method for creating and validating ANNs. For use in wastewater treatment plants, the material holds out a lot of potential. The water sector can use artificial intelligence to improve and oversee water monitoring and management. AI may also assist operators by making sophisticated, intelligent decision-making that improves the accuracy and dependability of the treatment system, even if it still requires operators to analyze complicated data. To deliver intelligent judgments, new AI-based algorithms are required to handle specific challenges in water management as well as treatment which include water quality, leak detection, and water process optimization. AI has the ability to completely alter the wastewater treatment process. The key AI techniques used in the treatment of wastewater and water for absorption of several contaminants were covered in this paper. The evaluation of various adsorbents in terms of color removal, metals, organic chemicals, medicines, medications, pesticides, and PCP removals from water has been effectively predicted using the AI models mentioned. This will aid in the establishment of a benchmark for the AI and hybrid models in the water treatment and wastewater management industries, allowing for improved performance and creative development. The current effort is an attempt to create the groundwork and provide direction for researchers who are interested in developing better nanocomposites for wastewater and fresh water treatment and management through the use of Industry 4.0.
Data Availability
The data are available from the corresponding author upon request.
Conflicts of Interest
The authors declare that there is no conflict of interest.