CHEMICAL ENGINEERING TRANSACTIONS  
 

VOL. 57, 2017 

A publication of 

 
The Italian Association 

of Chemical Engineering 
Online at www.aidic.it/cet 

Guest Editors: Sauro Pierucci, Jiří Jaromír Klemeš, Laura Piazza, Serafim Bakalis 
Copyright © 2017, AIDIC Servizi S.r.l. 

ISBN 978-88-95608- 48-8; ISSN 2283-9216 

Optimal Design of Membrane Processes: a Problem of 
Choices Between Process Layout, Operating Conditions and 

Adopted Control System 
Marco Stoller*a, Javier M. Ochando Pulidob 
aUniversity of Rome La Sapienza, Dept.of Chemical Materials Environmental Engineering, Via Eudossiana 18, 00184 
Rome, Italy  
bUniversity of Granada, Dept. of Chemical Engineering, Avenida de la Fuente Nueva S/N, 18071 Granada, Spain  
marco.stoller@uniroma1.it 

The development of membrane processes as a technology for environmental treatment applications and in 
particular for the purification of wastewater streams has significantly increased in the last decades. Fouling on 
membranes appears to be one of the main technical limit of this technology. This phenomenon causes the 
unavoidable deposition of particles on the membrane surface, building a resistive growing layer to 
permeability. Sensible fouling of the membrane leads to a significant reduction of the performances, a 
decrease of the operating life and, as a consequence, the increase of the operational costs due to the 
replacement or cleaning of the exhausted membrane modules. The presence of the fouling phenomena 
makes the proper design and control of membrane systems a difficult task. 
Optimal design of the membrane processes will be here discussed. The procedure requires to determine the 
optimal process layout given the input data and target requirements. At the end, the required membrane area 
is calculated. This latter property is strictly dependant of the adopted operating conditions, most importantly by 
the adopted value of transmembrane pressure (TMP). Moreover, it depends if the value of TMP remain fixed 
as a function of time or is variable (as in case of fixed permeate flow rates). Therefore, the optimal design of 
the system may occur only if the adopted control strategy is defined a priori. As a consequence, design 
choices of the membrane process layout, operating condition and adopted control system are strictly 
dependant, and connections between these different aspects should not be neglected during the engineering 
and P&I development stage of membrane systems. 
This paper will start from the theory of the boundary flux, in order to describe a novel design approach to 
membrane systems. Parallel to this, the development of an advanced control system, that allows to limit 
fouling formation during operation, is presented. The advanced control system relies on a suitable simulation 
software capable to predict the boundary flux, that changes the controller's set-points accordingly. Finally, the 
paper will merge all elements together, and report about the optimal design of membrane processes equipped 
with the advanced membrane process control system; validation of the proposed approach will be based on 
the use of a custom simulation model in ASPEN HYSYS and by experiments on lab scale. 

1. Introduction 

Membrane technologies have gained great importance in water and wastewater treatment applications due to 
its wide range of operation, ease to scale-up and high versatility. Membranes exhibits both high productivity 
and selectivity values towards pollutants, and therefore high efficiencies of water treatment. 
Environmental hazardous effluents may be treated to an aqueous stream, reaching the requirements for a 
municipal sewer system discharge, superficial aquifer release or industrial reuse. As stated by Schroetter and  
Boskaya-Scroetter, in 2010 about 60 Mm3 of wastewater are treated worldwide every day, and almost 50% by 
membrane processes such as microfiltration (MF), ultrafiltration (UF), nanofiltration (NF) and reverse osmosis 
(RO). In the next 10 years, an annual growth of 10% of this capacity is foreseen, and in respect to this, most 
probably membrane technology will gain more and more in importance. 

                               
 
 

 

 
   

                                                 
DOI: 10.3303/CET1757182

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Please cite this article as: Stoller M., Ochando-Pulido J.M., 2017, Optimal design of membrane processes: a problem of choices between 
process layout, operating conditions and adopted control system, Chemical Engineering Transactions, 57, 1087-1092   
DOI: 10.3303/CET1757182 

1087



The main drawback of membranes appears to be membrane fouling, that requires to be avoided or at least 
inhibited. This phenomenon does not permit to assure the performances of membranes over a long period of 
time once it starts to form. Starting with the last decade, research and expertise is focused on understanding 
and avoiding membrane fouling, and covers modelling (Ortiz Jerez et al., 2008), reporting (Daufin et al., 2001), 
development of new membrane materials (Lu et al., 2002), suggestion of proper operating conditions (Stoller 
and Chianese, 2007), management and control of the process (Stoller, 2016), proper pretreatment processes 
Ochando Pulido et al., 2016) and cleaning procedures (Ochando-Pulido, 2015). The result of actual know-how 
is that fouling cannot be avoided at all and will exist to a different extend in every membrane operation. If the 
fouling formation is small, it might be neglected when compared to other phenomena, such as aging. 
On contrary, is sensible fouling occurs, the phenomena must be considered and might represent a bottle neck 
to overcome, in order to guarantee a reliable separation process for long term operations. The latter case is 
most common in case of wastewater treatment. 
In this case, for a given system, the input parameters are: 
 

1. Capacity, that is the flow rate of water to be purified 
2. Recovery, that is the amount of water to recover as purified stream from the initial volume 
3. Selectivity, that is the amount of pollutants to be removed in order to obtain a suitable purified water 

stream 
 
It is possible to observe that fouling do not enter as an initial input parameter, at least not directly. In the past, 
many times this aspect was completely neglected or, in best case, this problem was overcomed by over-
design a forfeit or based on the experience of the process designers. Despite the adoption of these 
techniques, failures were not avoided.  
Therefore, an additional input parameter must be introduced in the design of the process, that is: 
 

4. Membrane longevity, which is a minimum number of operation cycles where membranes are capable 
to exhibit performances that guarantees the input targets 1 to 3. 

 
In this insight, it appears mandatory to perform design checks on membrane fouling. Options to inhibit fouling 
are the modelling and suggestion of proper operating conditions: the boundary flux concept and its 
determination seems to be one of the best method to identify process conditions capable to limit the formation 
of irreversible fouling (Stoller and Ochando Pulido, 2015). On the other hand, reversible and semi-reversible 
fouling will always trigger, leading to a performance loss as a function of time between washing periods. This 
latter phenomenon has a great importance on the definition of the process capacity, since the minimum 
required capacity must be guaranteed at the end of operation. 
Moreover, many times, boundary flux values are very low and not very attractive from an economic point of 
view. Therefore, proper pretreatment processes are mandatory: since operation is limited by the boundary flux 
values, one possible strategy may involve the sensible increase of its value. This can be performed by proper 
pretreatment of the feed stream (pretreatment tailoring). 
Membrane processes may be controlled mainly by maintaining fixed a pressure (TMP) or a permeate flow rate 
value. The first case is of easy implementation, but the ever changing permeate flux values during operation 
makes this approach unattractive. The second case requires pressure adjustments in order to be 
accomplished, and represent the common control scheme for membrane processes. In this paper, the control 
system will rely on this latter control strategy (permeate flow rate set-point). 
Management and control of the process appears to be key to reliability of this technology. Once the boundary 
flux value is determined and the relevant changes as a function of time are well defined, a control system 
should be implemented in order to operate the system correctly.  
The development of advanced control systems for membrane processes based on the boundary flux concept 
and relevant equations were introduced in the framework of previous membrane technology works, and here 
only limited to those used and briefly summarized (Stoller and Ochando Pulido, 2015): 
 
-dm/dt = α if Jp<Jb   (1) 
 
Jp(t) = m(t) TMP   (2) 
 
Fp(t) = A Jp(t)   (3) 
 
where Jp is the permeate flux, TMP the transmembrane pressure, Jb the boundary flux, m is the membrane 
permeability, α the sub-boundary fouling rate index, ∆w% the cleaning efficiency (and thus indication of 

1088



membrane aging), Y the recovery value, Fp the permeate flow rate. All other relationships may be derived or 
integrated starting from these. 

2. Process design and control in absence of fouling 

In this first case study, fouling will be neglected or simply is neglectable. There are very few processes where 
this approximation may apply, mostly in case of critical flux based systems. As ease of an example, olive mill 
wash wastewater is one of those systems (Stoller and Chianese, 2007). 
In this paper, typical input parameters in case of the treatment of olive mill wastewaters are assumed as in 
Table 1: 

Table 1:  Process design input parameters 

Design parameter  Value 

Capacity, F* 1 m3 h-1 
Recovery, Y* 90% 
Selectivity (on COD), R* 60% 
Longevity, N* 1000 d 
 
Washing is performed at the end of day, lasting a time T equal to 18h of operation. 
On the other hand, the membrane choice will affect together with the feedstock, the performances of the 
system. The main constraint to consider in a proper membrane type identification process is selectivity. In our 
case, the Desal Osmonics DK - SW nanofiltration membrane is capable to achieve the desired purification 
result. Therefore, we will assume that this design parameter results fully accomplished. 
Typical input values for the boundary flux equation set for this membrane are reported in Table 2, obtained 
adopting as a pretreatment ultrafiltration: 

Table 2:  Membrane properties 

Membrane characteristics  Value 

Pure water permeability 1,07 l h-1 m-2 bar-1 

Osmotic pressure 0 bar 
Selectivity (on COD) 70% 
Boundary flux 12,18 l h-1 m-2 

Boundary TMP 7 bar 
α 130 10-4 l h-2 m-2 bar-1 
∆w% 0,0005 % h-1 
 
Since no fouling will occur, the constraint on the desired longevity of the membranes is apparently useless. In 
this case, the value of α is assumed to be equal to 0. On the other hand, even with absence of fouling, 
membrane aging lowers the membrane performances as a function of time to a small extend, and this must be 
considered (∆w%). 
At this point, it is possible to perform the calculations for our (quite simple) example (referred as case study A) 
to evaluate the required membrane area A to fulfil the process targets, as reported in Table 3: 

Table 3:  Calculations for case study A 

Step Calculation  Equation Result 

1 No fouling  α = 0 
2 Permeate flux loss due to fouling -dm/dt = 0 m = const 
3 Permeate flux loss due to aging ∆W% = 1/100 ∆w% T N*  ∆W% = 0,09 
4 Project capacity due to flux losses F** = F* Y* (1 + ∆W%) F** = 981 l h-1 

5 Determination of permeate flux Jp = Jb Jp = 12,18 l h
-1 m-2 

6 Determination of membrane area A = F** Jp
-1 A = 80,54 m2 

 

1089



As noticed, calculations are very easy and quickly performed. Once finished, many times an over-design 
coefficient of 10-15% was applied in order to be safe from process failures. But again, this approach was not 
sufficient in many cases where fouling triggers. 

3. Process design and control in presence of fouling 

In this case study, called “B”, the same process design input parameters and membrane as in the case study 

A will be used (Table 1 and Table 2). 
In this case, α cannot be assumed equal to 0 and must be considered during calculations. In order to perform 
the calculations without a simulation model (although strongly suggested), two assumptions will be made: 
 

 Since both α and ∆w% will reduce the permeate flux values as a function of time, in order to 
guarantee the project targets, the suggested approach is to look at the last operation cycle, that is for 
N equal to 1000. If this latter cycle meets the requirements, all previous cycles automatically will 
reach them. 

 The boundary flux value is assumed to be constant as a function of COD. This may appear to be a 
strong condition, but is as much true if the given value is referred to the typical end of operation COD 
value (and this is the case of the data reported in Table 2).  

 
Again, we may perform the required calculations as reported in Table 4: 

Table 4:  Calculations for case study B 

Step Calculation  Equation Result 

1 Permeate flux loss due to fouling (last cycle) ∆Jp = α T TMPb ∆Jp = 1,63 l h
-1 m-2 

2 Permeate flux loss due to aging (all cycles) ∆W% = 1/100 ∆w% T N* ∆W% = 0,09 

3 Project capacity due to flux losses (all cycles) F** = F* Y* (1 + ∆W%) F** = 981 l h-1 

4 Determination of permeate flux (last cycle) Jp(T) = Jb Jp(T) = 12,18 l h
-1 m-2 

5 Determination of membrane area (last cycle) A = F** (Jp
-1(T) - ∆Jp) A = 92,98 m

2 
 
It can be noticed that the required membrane area is about 16% higher than that calculated in case study A, at 
the limit of the overdesign applied to the process a forfeit. Given that both processes are properly controlled, 
that is to work below boundary flux conditions all the time. 

4. The control system 

The previous calculations were possible since many parameters remain constant, the most important being α 
and ∆w%, respectively.  
In a previous work it was shown that these parameters will change sensibly and irreversibly as soon as 
superboundary conditions were used (Stoller, 2011). Even operating 20 minutes the system in superboundary 
conditions will lead to irreversible fouling, and a consequent increase about 30% of these parameters. 
A suitable control system capable to check operation in subboundary flux conditions is therefore mandatory 
and part of the overall high longevity strategy of the used membrane modules. Without this approach, 
membranes must be substituted more frequently, thus increasing the operating cost to such a level where the 
overall process results, form an economic point of view, not attractive and advantageous. This is especially 
true in wastewater treatment applications. 
The proposed control system is here shortly reported, and consists in a simulation model capable to evaluate 
and predict the boundary flux value profiles as a function of time (and COD) implemented in Hysys 
(AspenTech).  
In Figure 1, the scheme of the simulation model as shown in HYSYS GUI is pictured.  
 

1090



 

Figure 1: Simulation of the membrane process and the advanced control system  

The advanced control system is capable to: 
 

 Determine the operating parameters to reach end of operation in sub-boundary flux conditions given 
a desired permeate flow rate set-point by adopting given initial input values. 

 Check the constancy of the input parameters values given initially to the model during operation. If 
confirmed, the control system will follow the set-point settings. If the check fails, the model will 
evaluate again the operating conditions, and if these are not compatible to the given set-point, 
override the setting to best possible performances.  

 
For example, this happens in the case reported below. The missing operating time was about 8,5h, and the 
check on the model parameters (in particular on α) failed. The simulation model calculated a new boundary 
flux plot as a function of time, and estimated that a safe operation was possible only below 9,00 l h-1 m-2 and 
not at 12,18 l h-1 m-2 anymore.  
Moreover, the simulation code calculated that in this case, an additional operating time of 3h was required to 
reach the desired Y* value: therefore, end of operation was shifted to 11,5h.  
Without all these adjustments, it might be noticed that operating at previous boundary flux values or at 
adjusted ones without considering the additional amount of operating time, that is at 12,18 l h -1 m-2 (previous 
value) and 9,15 l h-1 m-2 (value at 8,5h), respectively, would lead to super-boundary flux conditions and, as a 
consequence, to severe irreversible performance loss of the process. 

 

Figure 2: Determination of the new set-point after a failed check on the model parameters 

1091



5. Conclusions 

In this paper, design concepts of membrane processes affected by fouling issues are given and highlighted.  
First of all, proper design requires to know, beside boundary flux data, at least the value of α. In many works, 
this important data is missing, since this parameter will affect directly the required overdesign to overcome 
fouling constraints in the long run.  
Finally, the importance of an advanced control system is mandatory, since the application of such a controller 
is the only way to guarantee the constancy of important values such as α and ∆w%. The approach relies on a 
guarantee to never overcome boundary flux conditions, but wants also to optimize operating conditions as 
high as possible in order to use all the available membrane area. The control system should therefore by of 
inferring and predictive type, to be capable to determine by real time measurable parameters the actual 
situation and to calculate the outcome of the operation till end. 
The developed control system in HYSYS appears to have all these properties and capabilities.   

References 

Daufin G., Escudier J.P., Carrere H., Berot S., Fillaudeau L., Decloux M., 2001, Recent and emerging 
applications of membrane processes in the food and dairy industry, Trans. IChemE 79, pp. 89–102. 

Lu X., Bian X., Shi L., 2002, Preparation and characterization of NF composite membrane, J. Membr. Sci. 210, 
pp. 3–11. 

Ochando-Pulido J.M., 2015, On the cleaning procedure of a hydrophilic reverse osmosis membrane fouled by 
secondary-treated olive mill wastewater, Chem. Eng. J. 260, pp. 142–151. 

Ochando-Pulido J.M., Stoller M., Di Palma L., Martínez-Ferez A., 2015, On the optimization of a flocculation 
process as fouling inhibiting pretreatment on an ultrafiltration membrane during olive mill effluents 
treatment, Desalination, DOI: 10.1016/j.desal.2015.12.021, accepted in press 

Ortiz Jerez M.J., Velez Pasos C.A., Mejıa E.F., 2008, Mathematical models of membrane fouling in crossflow 
microfiltration, Rev. Ingeniera Invest. 28 (1), pp. 123–132. 

Schroetter J., Boskaya-Scroetter B., 2010, Current and emerging processes for water treatment, Membranes 
for Water Treatment, pp. 53–72. 

Stoller M., Chianese A., 2007, Influence of the adopted pretreatment process on the critical flux value of batch 
membrane processes, Ind. Eng. Chem. Res. 46 (8), pp. 2249–2253. 

Stoller M., Ochando Pulido J.M., 2015, The Boundary Flux Handbook: A Comprehensive Database of Critical 
and Threshold Flux Values for Membrane Practitioners, ISBN 9780128015896. 

Stoller M., 2016, About the Validation of Advanced Membrane Process Control Systems in Wastewater 
Treatment Applications, CET 47, pp. 385-391. 

Stoller M., 2011, Effective fouling inhibition by critical flux based optimization methods on a NF membrane 
module for olive mill wastewater treatment, Chemical Engineering Journal 168, pp. 1140–1148. 

 
 

1092