Microsoft Word - 476hernandez.docx


CHEMICAL ENGINEERINGTRANSACTIONS 
 

VOL. 43, 2015 

A publication of 

The Italian Association 
of Chemical Engineering 
Online at www.aidic.it/cet 

Chief Editors:SauroPierucci, JiříJ. Klemeš 
Copyright © 2015, AIDIC ServiziS.r.l., 
ISBN 978-88-95608-34-1; ISSN 2283-9216                                                                               

 

About the Development of Advanced Membrane Process 
Control Systems 

Marco Stoller 

University of Rome “La Sapienza”, Dept. Of Chemical Materials Environmental Engineering, Via Eudossiana 18, 00184 
 Rome, Italy 
marco.stoller@uniroma1.it 

This paper focuses on the design and development of advanced control systems to use on either batch or 
continuous, new or existing membrane process plants. 
In the last decade, membrane technologies resulted to be very appealing and shows positive market trends. 
One main drawback is membrane fouling, which affects productivity, selectivity and longevity of the process, 
which leads to both technical and economical failures: proper membrane process design and control is a 
difficult task to accomplish. This leads to overdesign the plant capacities by process engineers, making this 
technology less reliable and convenient. 
Nowadays membrane processes are controlled by a constant permeate flow rate or constant applied 
operating pressure. These simple control strategy approaches are sufficient to operate the processes, but do 
not distinguish different fouling operating regions, and therefore do not avoid process failures due to fouling. 
Fouling may be described by the boundary flux theory in a convenient way, separating low-fouling operations 
from high-fouling ones. 
The paper reports about the development of an advanced membrane process control system based on the 
boundary flux concept. The developed advanced control strategy by the use of a simulation software, capable 
to predict boundary flux values by measurement of some key parameters, provides suitable set-point values to 
the feedback controllers in order to work at or below the boundary flux. As a consequence, the membrane 
process is always operated far from irreversible fouling issues. The developed approach was then 
successfully validated by experiments on lab scale. 

1. Introduction 

In the last decade membrane technologies gain market and are used as stand-alone, integrated or substitutive 
processes. The technology is very appealing, both from technical and economic point of view, and has many 
advantages if compared to conventional technologies. On the other side, one main drawback is membrane 
fouling, which represents a main constraint holding membrane technologies away from a definitive maturation. 
Nowadays, proper membrane process design can be a difficult task to accomplish when fouling is present and 
must be considered. The designer normally should consider the project variables concerning productivity and 
selectivity and follow these targets; in the presence of fouling, additional parameters, in particular the longevity 
of the membrane modules and the constancy of the permeate fluxes as a function of time must be considered. 
Fouling leads to a reduction of the permeate flux rate and parallel to this, may lead to sensibly shorten the life 
time of membrane modules. The presence of fouling, and the consequent reduction of permeate fluxes as a 
function of time, forces the designer to over-design the membrane plant in order to guarantee sufficient 
operating autonomy to conduct the process for a certain period of time at or above the permeate project 
values (Saad, 2005). In most cases the over-design is performed a forfeit or by past experience of the 
designer, starting only from the knowledge of the permeate project value, without considering in detail the 
entity and nature of fouling. Some examples of existing overdesigned membrane plants are reported in 

                                

 
 

 

 
   

                                                  
DOI: 10.3303/CET1543233 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Please cite this article as: Stoller M., 2015, About the development of advanced membrane process control systems, Chemical Engineering 
Transactions, 43, 1393-1398  DOI: 10.3303/CET1543233

1393



bibliography (Hassan et al.; Pirokova, 2006). In other cases, even worse, engineers underdesign the 
membrane plant, depending on higher operating conditions, which are not sustainable for long period of time. 
One method to increase the reliability of a process is to adopt stable control systems. In the case of most 
membrane processes, this is performed by simple control strategies, that do not include knowledge and 
control of fouling. Most proposed systems relays on the timed control of washing procedures (Ruano et al., 
2013), when fouling just triggered, or the control of anti-fouling procedures (Xu et al., 2014), such as the 
correct dosage of chemicals or air bubbles. Those systems do react on fouling issues, but do not include 
active strategies to avoid them. The complete lack of advanced control systems in membrane technologies, 
capable to avoid fouling issues, limits the reliability of the technology and represent one key problem to be 
solved in order to permit further maturation of membrane applications. 
In order to achieve this result, the fouling behaviour of the system must be defined a priori. For liquid-liquid 
separation processes, Field et al. introduced the concept of critical flux for microfiltration, stating that there is a 
permeate flux below which fouling is not promptly observed (Field et al., 1995). Afterwards, it was possible to 
identify critical flux values on ultrafiltration (”UF”) and nanofiltration (“NF”) membranes systems, too (Mänttäri 
and Nystörm, 2000). Nowadays, the critical flux concept is well accepted by both scientists and engineers as a 
powerful membrane process optimization tool as long critical fluxes apply [7]. In case of most real waste water 
streams Le Clech et al. noticed that operations below the critical flux may not be sufficient in order to have 
zero fouling rates (Le Clech et al., 2006). Moreover, the measurement of critical fluxes was often not possible, 
and in order to overcome this problem, the identification of "apparent" critical points were used. Therefore it 
appears that membrane systems treating real waste water streams do not exhibit a critical flux in strict way. To 
overcome this limitation in the definition of critical flux, in a recent paper, Field and Pearce introduced for the 
first time the concept of the threshold flux  (Field and Pierce, 2011). Summarizing briefly the concept, the 
threshold flux is the flux that divides a low fouling region, characterized by a nearly constant rate of fouling, 
from a high fouling region, where flux dependant high fouling rates can be observed.  
The main drawback of this concept is that the determination of threshold flux values cannot be theoretically 
predicted, but only experimentally determined by time consuming experiments at a certain time moment. The 
unavoidable development of fouling will irremediably change the relevant threshold flux value, too.  
The Author actively performed much research in the last years by applying both concepts on different 
systems, such as olive mill, tannery and tomato industry wastewater streams. Finally, both critical and 
threshold flux concepts, that share many common aspects, were merged by the principal investigator into a 
new concept, that is the boundary flux, and was introduced in year 2014 (Stoller and Ochando-Pulido, 2014). 
The introduction of the new boundary flux concept does not extend by addition of new theory or knowledge the 
critical and threshold flux concepts. The important point is that the knowledge of real-time boundary flux 
values is a key factor to design a stable control system for membrane processes, since operation in sub-
boundary flux conditions avoids irreversible fouling and thus premature technical (and economical) failures.  
The boundary flux values are sensibly influenced by those parameters affecting the critical and threshold flux, 
that is: hydrodynamics, temperature, membrane properties, time and feedstock characteristics. Concerning 
the first three points much has been reported in literature and the effects are nowadays well known. Moreover, 
it is relatively simple to maintain these parameters constant during membrane operation once fixed. 
Concerning time, the boundary flux values are as a function of this parameter, changing mainly due to the 
variable feedstock characteristics and the membrane actual health. This factor sensibly increases the difficulty 
to follow the boundary curve throughout operation, since theoretically boundary flux measurements must be 
frequently performed in order to accurately define the actual status of the process. Concerning the feedstock 
characteristics, Stoller and Ochando-Pulido performed some research on this topic by using olive mill 
wastewater streams, both exiting the 2-phase with (Ochando-Pulido et al., 2014) and without (Ochando-Pulido 
and Stoller, 2014) and the 3-phase olive oil production processes, again with (Stoller et al., 2014) and without 
(Stoller and Ochando-Pulido, 2012) pretreatment processes. A good design of the pretreatment processes 
(so-called pretreatment tailoring) beforehand the membrane section appears to sensibly increase the value of 
Jb, such as using photocatalysis with magnetic nanocore (Ruzmanova et al., 2013a) or doped nanotitania 
(Ruzmanova et al., 2013b) and coagulation at different dosages (Stoller et al., 2013ab). Moreover, for control 
purposes, it appears mandatory that the key parameters were measured on-line, in-situ and real-time. 
Together with operating conditions measurement, an advanced control system, based on a simulation 
software capable to predict the actual boundary flux value, is capable to control the process in sub-boundary 
conditions. This advanced control system is put behind conventional control systems, and defines real-time 
proper set-points to optimize the process with insight of productivity, selectivity and longevity by avoiding 
irreversible fouling. 

1394



2. Basic control systems of membrane processes 

Mainly two main membrane processes control strategies are adopted worldwide, that is controlling the 
permeate flow rate by changing the applied pressure value by a regulation valve on the retentate line, or 
controlling directly the constancy of the applied pressure.  
The advantage of the constant permeate flow rate control strategy is the possibility to maintain the productivity 
of the membrane plant always in line with the project values. Moreover, the application requires the use of a 
simple proportional-integral (PI) type control system. The problem of this strategy is to define correctly the 
permeate flow rate set-point of the controlled system. Since boundary flux values will change as a function of 
time and of the pollutant concentration in the feedstock, it is not correct to assume as set-point value the 
boundary flux rate value at beginning of the operation. In particular, this problem would have higher impact 
when applied to batch membrane processes. In order to overcome this problem, a simulation software is 
required to estimate the boundary flux value at the end of the operation.  
The advantage of the constant operating pressure value control strategy is the simplicity, but the productivity 
of the membrane plant may not be in line with the expected project values. The problem of this strategy is to 
define correctly the operating pressure set-point PP of the controlled system: the osmotic pressure will change 
as a function of time and of the pollutant concentration in the feedstock, and needs to be added to the TMP 
value; again, a simulation software is required to follow the changes during operation. In particular, this 
problem would have higher impact when applied to batch membrane processes.  
The advantages and disadvantages of applying the control of the permeate flow rate (FC) or of the operating 
pressure (PC) to both continuous and batch membrane processes are hereafter briefly listed, partly in 
accordance with previous observations from Vyas et al. (2002): the membrane area requirements in case of 
batch processes are lower for PC strategies, in case of constant feedstock, if compared to FC strategies. The 
gap reduces sensibly as soon as KP changes accordingly to feedstock concentration variations. By the same 
terms, the advantage still holds for continuous processes. The control of the process is performed in both 
cases by a PI type controller. In case of changing feedstock at the start of operation or batch processes, 
advanced control is necessary: PC strategies requires the indirect measurement of the initial osmotic pressure 
of the feedstock, and this is in most cases not an easy task to accomplish; FC strategies requires the 
simulation software to calculate a new value for FF. In both cases the purpose it to give a different set-point to 
the PI type control systems.  
In Figure 1, the two basic control strategies are reported and the relevant block diagram is shown. 
 

                       
          

                                                       

       
 
Figure 1: Scheme and block diagrams for the basic control systems of FC (left) and PC (right) strategies. 

1395



3. The boundary flux equations 

The boundary flux divides the operation of membranes in two different regions: a lower one, where no or a 
small, constant amount of fouling triggers, and a higher one, where fouling builds up very quickly. The relevant 
equations may be written as: 
 
dm/dt = - α; Jp(t) ≤ Jb          (1) 
dm/dt = - α + β ( Jp(t) - Jb ); Jp(t) > Jb          (2)  
 
where α, expressed in L h-2 m-2 bar-1, is the sub-boundary fouling rate index and β, expressed in h-1 m-2 bar-1, 
is the super-boundary fouling rate index.  
The boundary flux value itself is a function of many parameters, mainly affected by hydrodynamics (Re 
number), temperature (T) and the solute concentration (KP). In this work the key parameter KP will be the 
COD value of the feedstock. If the other parameters remain constant, the relationships can be summarized 
separately in the following set of equations: 
 
Jb (Re) = Jb (Re→∞) (1 – e A Re)  (3) 
Jb (T) = AT T2 + BT T + CT        (4) 
Jb(KP,t) = w Pb(0) - α t Pb(0) - [ w p1 - α p1 t + m1 Pb(0) ] KP + m1 p1 KP2              (5) 
 
With A, AT, BT and CT fitting parameters. It is interesting to notice that only the last one is a function of time as 
well, even if KP remain constant during operation. On contrary, by proper control systems, both Re and T can 
be maintained constant by fixing their set-point. In eq.(5) w, m1 and p1 are fitting parameters of the respective 
equations fitting the membrane permeability m and the osmotic pressure π to KP: 
  
m(KP,t) = w – m1 KP           (6) 
π(KP) = p1 KP       (7) 
 
and where Pb(0) is the applied operating pressure at the boundary flux conditions at the start of operation, in 
detail: 
 
Pb(0) = TMPb + π(KP)|t=0       (8) 
 
The pure water permeability, that is w, may be a function of time depending from the amount of irreversible 
fouling formed over on the membrane.  
The definition of a value for Jb permits to determine the range of validity of eq.(1), here of interest for 
membrane process design purposes. 
 
The model is completed by two equations concerning permeate fluxes and selectivity, that is: 
 
Jp(t) = Jp(0) - α TMP(t) t; Jp(0÷t) ≤ Jb  (9) 
R(KP) = σ TMP ( TMP + γ )-1  (10) 
 
where σ is the reflection coefficient and γ a fitting parameter. 

4. Advanced control system  

Eqs(3-10) define the value of Jb once values of KP, T, t, Re are measured and the other fitting parameters are 
known. In the hypothesis to have all the available data to run the model, the proposed equations are suitable 
for advanced control.  
If the membrane area A and the total operating time τ is fixed, the set-point of the two control strategies may 
be defined as follows: 
 
For FC strategy: FpSP = 0.9 Jb(τ) A  (11) 
For PC strategy: PSP = 0.9 Pb  (12) 
 
where 0.9 is an applied safety factor. In order to evaluate Jb(τ), following relationship can be used if the final 
value of the desired recovery factor Y* is fixed: 
 

1396



KP(τ) = KP(0) ( 1 - Y* + Y* R(TMPb) ) ( 1 - Y* )-1  (13) 
 
Figure 2 shows the relevant block diagrams. 
 
 

 
 
Figure 2: Block diagram of the advanced control system for the FC (left) and PC (right) strategies. 

5. Results and discussion  

The two control strategies were applied to a batch membrane system well known in terms of data and 
parameters and reported elsewhere, and hereafter briefly summarized in Table 1 (Stoller, 2013). The 
feedstock was a pretreated olive mill wastewater stream (COD equal to 20000 mg/l), the used ultrafiltration 
membrane was supplied by GE Water (Model GM). 
In Figure 3 the results were shown. 
 

 
 
Figure 3: Control of a batch membrane process by FC (left) and PC (right) control strategy. 
 
In case of both strategies, the beginning polarization of the membrane is not taken into account by the model, 
thus permeability and pressure data points were slightly off curve before their stabilization.  
Batch membrane processes are not easy to control since as a function of time and operation, KP changes 
accordingly due to the concentrate recycle and increase of solute concentration. From these preliminary 
results, a good control of the batch membrane process at the fixed FpSP or PSP values, respectively, was 
observed and the suggested approach in this paper appears to be validated.  
Both processes were operated to the end, and after this no fouling was observed. 

6. Conclusions 

Basic control system for membrane processes may not prove to be sufficient in those cases severe fouling 
issues are triggered, as a consequence of a manual setting of the set-point values to those of the project. The 
problem is that the boundary flux values will change as a function of time, and fixed set-point values may 
overcome boundary flux conditions during operation. Advanced control systems are desirable and should 
provide suitable set-point values to the control system in order to perform the operation always below the 
actual boundary flux value, avoiding irreversible fouling operating conditions. 
The model reported in the paper was capable to provide suitable set-point values for both constant permeate 
flux and constant TMP control strategies leading to the end of operation of a wastewater batch membrane 
treatment process, without formation of irreversible fouling. The success of the method relies on the 
determination of some parameters and the continuous measurement of others. If properly set up, the 
proposed approach may fit to other membrane batch systems, providing advanced process control with the 
target to avoid irreversible fouling formation.  

1397



References 

Field R.W., Pearce G.K., Critical, sustainable and threshold fluxes for membrane filtration with water industry 
applications, 2011, Advances in Colloid and Interface Science, vol. 164, no. 1-2, 38-44. 

Field R.W., Wu D., Howell J.A., Gupta B.B., Critical flux concept for microfiltration fouling, 1995, Journal of 
Membrane Science, vol. 100, 259-272. 

Hassan M., Jamaluddin T., Saeed O., Al-Rubaian A., Al-Reweli A., Al-Birk A., SWRO plant operation with 
toyobo membrane in train 200 instead of Dupont b-10 membrane, 2010, SWCC research paper 2010. 

Le-Clech P., Chen V., Fane T.A.G., Fouling in membrane bioreactors used in wastewater treatment, 2006, 
Journal of Membrane Science, vol. 284, no. 1-2, 17-53. 

Mänttäri M., Nystörm M., Critical flux in NF of high molar mass polysaccharides and effluents from the paper 
industry, 2000, Journal of Membrane Science, vol. 170, 257-273. 

Ochando-Pulido J.M., Stoller M., Boundary flux optimization of a nanofiltration membrane module used for the 
treatment of olive mill wastewater from a two-phase extraction process, 2014, Separation and Purification 
Technology, 130, 124-131. 

Ochando-Pulido J.M., Stoller M., Di Palma L., Martinez-Ferez A., Threshold performance of a spiral-wound 
reverse osmosis membrane in the treatment of olive mill effluents from two-phase and three-phase 
extraction processes, 2014, Chemical Engineering and Processing: Process Intensification, 83, 64-70.   

Pikorová T., Two years of the operation of a domestic MBR wastewater treatment plant, 2006, Slovak J. Civ. 
Eng. VXX(2), 28-36. 

Ruano A., Ribes M.V., Ferrer J., Advanced control system for optimal filtration in submerged anaerobic MBRs 
 (SAnMBRs), 2013, Journal of Membrane Science, 430, 330-341. 
Ruzmanova Y., Stoller M., Chianese A., Photocatalytic treatment of olive mill wastewater by magnetic core 

titanium dioxide nanoparticles, 2013, Chemical Engineering Transactions, 32, 2269-2274. 
Ruzmanova Y., Ustundas M., Stoller M., Chianese A., Photocatalytic treatment of olive mill wastewater by n-

doped titanium dioxide nanoparticles under visible light, 2013, Chemical Engineering Transactions, 32, 
2233-2238. 

Saad M.A., Membrane desalination for the Arab World: Overview and Outlook, 2005, Arab Water World, 
1/2005, 10-14. 

Stoller M., A three year long experience of effective fouling inhibition by threshold flux based optimization 
methods on a nf membrane module for olive mill wastewater treatment, 2013, Chemical Engineering 
Transactions, 32, 37-42. 

Stoller M., Ochando-Pulido J.M., Di Palma L., On the relationship between suspended solids of different size, 
the observed boundary flux and rejection values for membranes treating a civil wastewater stream, 2014,  
Membranes, 4 (3), 414-423. 

Stoller M., Ochando-Pulido J.M., Going from a critical flux concept to a threshold flux concept on membrane 
processes treating olive mill wastewater streams, 2012, Procedia Engineering, 44, 607-608. 

Stoller M., Ochando-Pulido J.M., About merging threshold and critical flux concepts into a single one: The 
boundary flux, 2014, The Scientific World Journal, Article ID: 656101. 

Stoller M., Ochando-Pulido J.M., Chianese, A., Comparison of critical and threshold fluxes on ultrafiltration 
and nanofiltration by treating 2-phase or 3-phase olive mill wastewater, 2013a, Chemical Engineering 
Transactions, 32, 397-402. 

Stoller, M., Sacco, O., Sannino, D., Chianese, A., Successful integration of membrane technologies in a 
conventional purification process of tannery wastewater streams, 2013b, Membranes, 3 (3), 126-135. 

Vyas H.K., Bennett R.J., Marshall A.D., Performance of cross flow MF during constant TMP and constant flux    
 operations, 2002, International Diary Journal, vol. 12, 473-479. 
Xu H., Chen W., Xiao H., Hu X., Stability of an ultrafiltration system for drinking water treatment, using chlorine 
 for fouling control, 2014, Desalination, 336 (1), 187-195. 
   
 

1398