Advanced control of membrane fouling in filtration systems using artificial intelligence and machine learning techniques: A critical review

https://doi.org/10.1016/j.psep.2019.01.013Get rights and content

Highlights

  • An advanced approach is introduced to automatically control membrane fouling.

  • Artificial intelligence and machine learning techniques are used in the control system.

  • Intelligent techniques are capable for the modeling and optimization of membrane fouling.

  • Membrane fouling can be monitored and appropriate decisions can be made at the right time.

Abstract

This paper critically reviews all artificial intelligence (AI) and machine learning (ML) techniques for the better control of membrane fouling in filtration processes, with the focus on water and wastewater treatment systems. Artificial neural networks (ANNs), fuzzy logic, genetic programming and model trees were found to be four successfully employed modeling techniques. The results show that well-known ANNs such as multilayer perceptron and radial basis function can predict membrane fouling with an R2 equal to 0.99 and an error approaching zero. Genetic algorithm (GA) and particle swarm optimization (PSO) are optimization methods successfully applied to optimize parameters related to membrane fouling. These optimization techniques indicated high capabilities in tuning various parameters such as transmembrane pressure, crossflow velocity, feed temperature, and feed pH. The results of this survey demonstrate that hybrid intelligent models utilizing intelligent optimization methods such as GA and PSO for adjusting their weights and functions perform better than single models. Clustering analysis, image recognition, and feature selection are other employed intelligent techniques with positive role in the control of membrane fouling. The application of AI and ML techniques in an advanced control system can reduce the costs of treatment by monitoring of membrane fouling, and taking the best action when necessary.

Introduction

Membrane bioreactor (MBR) is a capable filtration system, which is superior to the traditional separation systems (Le-Clech et al., 2006). For example, application of MBRs for water and wastewater treatment has several advantages such as simple flow configuration, stable and high effluent quality, high volumetric loading, lower surplus sludge production, and high biomass concentration without sludge settling problems over conventional treatment systems (Bagheri et al., 2016b). However, membrane fouling is a major problem for more widespread and large scale applications of MBRs since it increases operating costs due to the needs for more frequent membrane cleaning and replacement (Bagheri and Mirbagheri, 2018; Le-Clech et al., 2006).

To effectively deal with the problem of membrane fouling, a number of studies have been conducted to shed light on the mechanisms of membrane fouling in filtration system (Chen et al., 2016; Qu et al., 2018; Teng et al., 2018a, b; Teng et al., 2019; Zhang et al., 2018). The main introduced fouling mechanisms are based adhesion and filtration theories. The adhesion behaviors of foulants can be described by the extended Derjaguin-Landau-Verwey-Overbeek (XDLVO) theory (Brant and Childress, 2002). This theory is the most notable approach for the quantitative analysis of membrane fouling. However, it is mainly applied for single foulants, cannot plausibly explain the extremely high adhesion, and rarely is used for complicated organic mixture fouling (Sun et al., 2018; Teng et al., 2019). The exact mechanism underlying the typical fouling behaviors of soluble microbial products (SMPs) as predominate foulants, which are highly adhesive has been described using a unified thermodynamic mechanism (Teng et al., 2019). There are also other recently published studies numerically examined interfacial interactions, which govern adhesion/deposition of various foulants on membrane surface (Qu et al., 2018; Teng et al., 2018a). The mechanisms of membrane fouling also have been described using classical filtration theories such as Kozeny-Carman equation, which can be used to calculate cake resistance in filtration systems (Jepsen et al., 2018). However, it should be taken into account that the Carman–Kozeny equation is appropriate for granular media bed with particles of millimeter in size and with a consideration of the flow drag (Bai and Leow, 2002). High filtration resistance of gel layer has been investigated to elucidate the underlying mechanisms of fouling caused by the gel layer (Chen et al., 2016). Using a proposed mechanism based on Flory-Huggins theory (Flory, 1942), it was revealed that chemical potential mechanism related with foulant/cake layer filtration is mainly responsible for the filtration resistance (Zhang et al., 2018). Flory-Huggins and Flory-Rehner theories also have been used to investigate the mechanisms and roles of gel elasticity in extremely specific filtration resistance (Teng et al., 2018b). Pore blockage, foulant layer formation and effects of hydrodynamic conditions are other available mechanisms to describe membrane fouling. The survey of previous studies on the mechanisms of membrane fouling shows that a considerable number of models have been developed based on filtration resistance and pore blocking laws.

Several antifouling strategies have been applied to mitigate reversible and irreversible membrane fouling in filtrations systems. These antifouling strategies are very diverse and vary from optimizing operational conditions to applying chemical agents. The more improved strategies employ new and novel technologies such as nanotechnology to disrupt fouling mechanism through aff ;ecting fouling causing bacteria. Application of nanoparticles, cell entrapment, biologically- and electrically-based strategies are the latest antifouling eff ;orts (Bagheri and Mirbagheri, 2018; Meng et al., 2017). The application of antifouling strategies in an advanced fouling control system has received great attention in recent years. The purpose of utilizing advanced control systems is to monitor membrane fouling and making appropriate decisions at the appropriate time. Optimizing operating conditions from on-line data, and finding the right time for membrane cleaning are two of several benefits of using the advanced control system.

Artificial intelligence (AI) and machine learning (ML) are two intelligent approaches, which have demonstrated high performance in solving environmental engineering problems (Bagheri et al., 2016a; Mirbagheri et al., 2015c). The survey paper by Gernaey et al. (2004), refers to the combination of AI methods and mechanistic models for activated sludge wastewater treatment plants to achieve and improved control of the treatment process. To support plant operators, supervisory control systems have been developed with the goal of improving plant performance and increasing operational reliability through automation in which both AI techniques and mechanistic models are important elements. In another review, Fan et al. (2018) have summarized the application of AI methods in experimental design for pollutants removal in water treatment. The AI methods for the modeling and optimization of pollutants removal in water treatment have been used to generate optimal operational variables. The results of conducted studies confirm the successful application of hybrid models in water treatment with satisfactory accuracies. The AI and ML techniques have been successfully applied for better control and management of membrane fouling in a number of studies. Schmitt and Do (2017) reviewed modeling efforts employed artificial neural networks (ANNs) as AI techniques for the prediction of membrane fouling in MBRs treating wastewater. Schmitt and Do (2017) concluded that ANNs are efficient modeling tools to predict membrane fouling in filtration systems and particularly MBR plants. They also suggested investigating the optimization of experimental database and ANN architecture along with the utility of considering operational time as a parameter in order to apply membrane fouling prediction using ANNs to large-scale MBRs.

The current research was an effort to critically discuss all AI and ML techniques that can be useful for the better control and management of membrane fouling in filtrations systems, with the focus on water and wastewater treatment systems. This review also summarized up-to-date studies concerning the application of AI and ML techniques such as single models, optimization methods, hybrid models, cluster analysis, and image recognition for the better control of membrane fouling. Finally, a typical procedure to achieve an advanced control and management system for membrane fouling using AI and ML techniques was discussed by considering the results of previous studies. In this paper, we first discuss the most important factors contributing to membrane fouling because successful application of AI and ML techniques requires the understanding of the working parameters. In the next section, the general concepts behind AI and ML techniques are explained, and AI and ML techniques for the control of membrane fouling are classified. We discuss the fundamentals and applications of intelligent modeling techniques including ANNs, fuzzy logic (FL) and genetic programming (GP). Then the basic concepts and applications of optimizing techniques such as genetic algorithm (GA) and particle swarm optimization (PSO) are discussed in details. Following the modeling and optimization techniques, we discuss the applications and advantages of hybrid models consisting of single models and optimization techniques. In addition to these intelligent techniques, other useful techniques such as various clustering algorithms, image recognition, and feature selection are reviewed to show their importance on the control of membrane fouling. In the last step, we propose an intelligent approach for the better control of membrane fouling based on the discussed AI and ML techniques.

Section snippets

Contributing factors to membrane fouling

Successful modeling of and optimization of membrane fouling in MBRs can be achieved only when there is understanding about important contributing factors to membrane fouling. The contributing factors are normally considered as inputs of intelligent models and optimization approaches to better control membrane fouling. The important contributing factors to membrane fouling have been classified into three main categories including operating conditions, biomass characteristics, and membrane

Mechanistic models for membrane fouling

Providing a tool to predict flux decline not only is helpful for scaling up and down of filtration system but also improve our understanding of membrane fouling phenomenon (Zheng et al., 2018). Mathematical modeling of flux decline based pore blocking mechanisms (Hermia’s model) are the early attempts to provide predictive tools and improve the understanding of the fouling occurrence during filtration. Most of the published studies have used one of the four classical fouling models including

Artificial intelligence and machine learning

AI is a branch of computer science that deals with the simulation of intelligent behavior in computers. With the advances in computer power, large amounts of data, and theoretical understanding, AI techniques have received high attention and have become an essential part of many research studies to solve challenging problems in various fields of science and engineering (Shi, 2011). AI is the intelligence displayed by machines to perceive the environment by them and take actions that maximize

Modeling approaches for membrane fouling

A number of studies regarding membrane fouling have used intelligent models for the prediction of various fouling indices such as TMP or membrane permeate flux. The main goal of employing the AI and ML based models were to achieve higher accuracy than mechanistic models and also avoid the problems such as model calibration. The ANNs are the first intelligent models that were used for prediction of membrane fouling following by FL models. GP is another intelligent approach, which has

Optimization approaches for membrane fouling

The optimization of filtration processes is of great importance both in terms of membrane fouling mitigation and reduce the cost of treatment. Intelligent methods have received great attention for optimizing the effective parameter related to membrane fouling. GA and PSO are two intelligent techniques that have shown high capabilities for the optimization of filtration processes with the purpose of mitigating membrane fouling. Fig. 5 illustrates the typical approach through which the effective

Hybrid modeling approaches for membrane fouling

The typical hybrid models are AI and ML based models that utilize intelligent optimization techniques to adjust their weights, thresholds and functions with the objective of achieving higher accuracy. The intelligent models also have been applied together with mechanistic models for membrane fouling in order to improve our understanding of the mechanism of membrane fouling. The analysis of conducted studies indicate that while the accuracy of single intelligent models is acceptable for

Clustering approaches for the control of membrane fouling

Cluster analysis is an ML technique to group a set of objects in the same cluster based on the similarity of objects (Bezdek, 1981). Cluster analysis has been reported as a capable technique, which can be employed alone or with other AI and ML techniques to better control and manage membrane fouling. Cluster analysis can be applied to solve many problems such as finding various groups of bacteria with different potential for causing membrane fouling, and grouping different wastewaters from

Image recognition and membrane fouling

Image recognition is an intelligent technique to identify and detect an object or a feature in a digital image. Image recognition in the context of machine vision is used in many applications such as smart photo libraries, systems for factory automation, toll booth monitoring, and security surveillance. Typical image recognition algorithms are employed for optical character recognition, pattern matching, face recognition, license plate matching, scene identification, and scene change detection (

Feature selection or variable selection

Feature selection or variable selection is the automatic selection of attributes in the data that are most relevant to the predictive modeling problems. Feature selection is a useful tool for building an accurate predictive model by selecting variables that give better accuracy whilst requiring less data. Feature selection is employed to identify and remove irrelevant and redundant variables from data that do not contribute to the accuracy of predictive models and even decrease the accuracy of

Prospects of employing AI and ML for fouling mitigation

Almost all of employed AI and ML techniques in the conducted studies regarding membrane fouling have shown high performance and capabilities in performing their assigned tasks. The intelligent modeling techniques have significantly improved the accuracy of prediction as comparing with mechanistic models. The application of the intelligent models together with mechanistic models seems to be interesting in terms of practical applications and also improving our understanding of the problem. Some

Conclusions

Comparison between the results of single models and hybrid models indicate that both approaches have high performance for the prediction of membrane fouling. However, the hybrid models achieve to intended results faster and do not need high expertise to tune the weights and functions of the modeling algorithms. Optimization algorithms are other intelligent techniques highly advantageous for optimizing the effective parameters related to membrane fouling. The findings of this study demonstrate

Acknowledgments

The authors declare that they have no competing interests.

References (178)

  • M. Barello et al.

    Neural network based correlation for estimating water permeability constant in RO desalination process under fouling

    Desalination

    (2014)
  • Ś.F. Boerlage et al.

    Modified fouling Indexultrafiltration to compare pretreatment processes of reverse osmosis feedwater

    Desalination

    (2000)
  • E.H. Bouhabila et al.

    Fouling characterisation in membrane bioreactors

    Sep. Purif. Technol.

    (2001)
  • J.A. Brant et al.

    Assessing short-range membrane–colloid interactions using surface energetics

    J. Membr. Sci.

    (2002)
  • E. Brauns et al.

    A new method of measuring and presenting the membrane fouling potential

    Desalination

    (2002)
  • G. Chandrashekar et al.

    A survey on feature selection methods

    Comput. Electr. Eng.

    (2014)
  • H. Chen et al.

    Prediction of permeate flux decline in crossflow membrane filtration of colloidal suspension: a radial basis function neural network approach

    Desalination

    (2006)
  • J. Chen et al.

    Application of wavelet analysis and decision tree in UTDR data for diagnosis of membrane filtration

    Chemometr. Intell. Lab. Syst.

    (2012)
  • J. Chen et al.

    Membrane fouling in a membrane bioreactor: high filtration resistance of gel layer and its underlying mechanism

    Water Res.

    (2016)
  • C.M. Chew et al.

    A practical hybrid modelling approach for the prediction of potential fouling parameters in ultrafiltration membrane water treatment plant

    J. Ind. Eng. Chem.

    (2017)
  • S. Curcio et al.

    Reduction and control of flux decline in cross-flow membrane processes modeled by artificial neural networks

    J. Membr. Sci.

    (2006)
  • M. Dalmau et al.

    Comparison of a deterministic and a data driven model to describe MBR fouling

    Chem. Eng. J.

    (2015)
  • J. De Bruijn et al.

    Membrane blocking in ultrafiltration: a new approach to fouling

    Food Bioprod. Process.

    (2005)
  • N. Delgrange et al.

    Neural networks for prediction of ultrafiltration transmembrane pressure–application to drinking water production

    J. Membr. Sci.

    (1998)
  • N. Delgrange-Vincent et al.

    Neural networks for long term prediction of fouling and backwash efficiency in ultrafiltration for drinking water production

    Desalination

    (2000)
  • N. Dhanachandra et al.

    Image segmentation using K-means clustering algorithm and subtractive clustering algorithm

    Proced. Comput. Sci.

    (2015)
  • M. Dornier et al.

    Dynamic modeling of crossflow microfiltration using neural networks

    J. Membr. Sci.

    (1995)
  • A. Drews et al.

    Improving the efficiency of membrane bioreactors by a novel model-based control of membrane filtration

    Comput. Aided Chem. Eng.

    (2007)
  • A. Drews et al.

    Model-based recognition of fouling mechanisms in membrane bioreactors

    Desalination

    (2009)
  • M. Fan et al.

    A review on experimental design for pollutants removal in water treatment with the aid of artificial intelligence

    Chemosphere

    (2018)
  • A. Fouladitajar et al.

    Membrane fouling in microfiltration of oil-in-water emulsions; a comparison between constant pressure blocking laws and genetic programming (GP) model

    Desalination

    (2013)
  • R.-Q. Fu et al.

    Modelling of the adsorption of bovine serum albumin on porous polyethylene membrane by back-propagation artificial neural network

    J. Membr. Sci.

    (2005)
  • D.-w. Gao et al.

    Linking microbial community structure to membrane biofouling associated with varying dissolved oxygen concentrations

    Bioresour. Technol.

    (2011)
  • W. Gao et al.

    Influence of temperature and temperature shock on sludge properties, cake layer structure, and membrane fouling in a submerged anaerobic membrane bioreactor

    J. Membr. Sci.

    (2012)
  • K.V. Gernaey et al.

    Activated sludge wastewater treatment plant modelling and simulation: state of the art

    Environ. Model. Softw.

    (2004)
  • S. Ghandehari et al.

    A comparison between semi-theoretical and empirical modeling of cross-flow microfiltration using ANN

    Desalination

    (2011)
  • B. Grosman et al.

    Automated nonlinear model predictive control using genetic programming

    Comput. Chem. Eng.

    (2002)
  • A. Guadix et al.

    Predicting the flux decline in milk cross-flow ceramic ultrafiltration by artificial neural networks

    Desalination

    (2010)
  • H. Hazrati et al.

    The influence of hydraulic retention time on cake layer specifications in the membrane bioreactor: experimental and artificial neural network modeling

    J. Environ. Chem. Eng.

    (2017)
  • K. Hong et al.

    Assessment of various membrane fouling indexes under seawater conditions

    Desalination

    (2009)
  • C. Huyskens et al.

    Validation of a supervisory control system for energy savings in membrane bioreactors

    Water Res.

    (2011)
  • J. Izquierdo et al.

    Design optimization of wastewater collection networks by PSO

    Comput. Math. Appl.

    (2008)
  • C.H. Koo et al.

    Review of the effect of selected physicochemical factors on membrane fouling propensity based on fouling indices

    Desalination

    (2012)
  • P. Lasch et al.

    Imaging of colorectal adenocarcinoma using FT-IR microspectroscopy and cluster analysis

    Biochim. Biophys. Acta Mol. Basis Dis.

    (2004)
  • P. Le-Clech et al.

    MBR focus: the operators’ perspective

    Filtr. Sep.

    (2005)
  • P. Le-Clech et al.

    Fouling in membrane bioreactors used in wastewater treatment

    J. Membr. Sci.

    (2006)
  • T.-M. Lee et al.

    Prediction of membrane fouling in the pilot-scale microfiltration system using genetic programming

    Desalination

    (2009)
  • C. Li et al.

    The application and research of the GA-BP neural network algorithm in the MBR membrane fouling

    Abstr. Appl. Anal.

    (2014)
  • Q.-F. Liu et al.

    Evaluation of membrane fouling models based on bench-scale experiments: a comparison between constant flowrate blocking laws and artificial neural network (ANNs) model

    J. Membr. Sci.

    (2008)
  • H. Abdi et al.

    Principal component analysis

    Wiley Interdiscip. Rev. Comput. Stat.

    (2010)
  • Cited by (139)

    View all citing articles on Scopus
    View full text