Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
SUPERVISORY CONTROL SYSTEM AND METHOD FOR MEMBRANE CLEANING
Document Type and Number:
WIPO Patent Application WO/2008/132186
Kind Code:
A1
Abstract:
A method of controlling membrane cleaning actions of a membrane filtration system comprises measuring the rate of permeate produced by a permeate production facility comprising a membrane in function of time in a first measuring step and iteratively performing the steps of cleaning the membrane of said permeate production facility by at least one mechanical cleaning action, followed by measuring the rate of permeate produced by said permeate production facility in function of time. The method further comprises steps of computing a reversible fouling propensity of the feed based on the rate of permeate which is recovered after each cleaning step, computing an irreversible fouling propensity of said feed based on the rate of permeate which remains unrecovered after each cleaning step and calculating the value of at least one first parameter related to a mechanical membrane cleaning action based on the reversible fouling propensity. The method can further comprise a step of calculating the value of at least one second parameter related to a chemical membrane cleaning action based on the irreversible fouling propensity. A membrane filtration system comprising a supervisory control system implementing the method is also disclosed.

Inventors:
BRAUNS, Etienne (Lemmensblok 2, Mol, B-2400, BE)
VAN HOOF, Erwin (Horzelend 42, Retie, B-2470, BE)
Application Number:
EP2008/055113
Publication Date:
November 06, 2008
Filing Date:
April 25, 2008
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
VLAAMSE INSTELLING VOOR TECHNOLOGISCH ONDERZOEK (VITO) (Boeretang 200, Mol, B-2400, BE)
BRAUNS, Etienne (Lemmensblok 2, Mol, B-2400, BE)
VAN HOOF, Erwin (Horzelend 42, Retie, B-2470, BE)
International Classes:
B01D61/12; B01D61/22; B01D65/02; B01D65/08
Domestic Patent References:
2002-06-20
Foreign References:
US5198116A1993-03-30
EP1300186A12003-04-09
Other References:
CABASSUD M ET AL: "Neural networks: a tool to improve UF plant productivity" DESALINATION, ELSEVIER, AMSTERDAM, NL, vol. 145, no. 1-3, 10 September 2002 (2002-09-10), pages 223-231, XP004386295 ISSN: 0011-9164
DELGRANGE-VINCENT N ET AL: "Neural networks for long term prediction of fouling and backwash efficiency in ultrafiltration for drinking water production" DESALINATION, ELSEVIER, AMSTERDAM, NL, vol. 131, no. 1-3, 20 December 2000 (2000-12-20), pages 353-362, XP004306366 ISSN: 0011-9164
BRAUNS, E.: "Could fuzzy logic be the key to membrane fouling control?" INTERNATIONAL DESALINATION AND WATER REUSE QUARTERLY, vol. 13, no. 2, August 2003 (2003-08), - September 2003 (2003-09) pages 18-24, XP008085301
BRAUNS E ET AL: "A new method of measuring and presenting the membrane fouling potential" DESALINATION, ELSEVIER, AMSTERDAM, NL, vol. 150, no. 1, 10 October 2002 (2002-10-10), pages 31-43, XP004393524 ISSN: 0011-9164 cited in the application
Attorney, Agent or Firm:
PRONOVEM - OFFICE VAN MALDEREN (Avenue Josse Goffin, 158, Brussels, B-1082, BE)
Download PDF:
Claims:

CLAIMS

1. A method of controlling membrane cleaning actions of a membrane filtration system, the method comprising the steps of: — providing a device for measuring a fouling propensity of a feed applied to the membrane filtration system, wherein the device comprises a permeate production facility comprising a membrane,

- feeding at least a part of the feed to the membrane of said permeate production facility,

- applying a trans-membrane-pressure across the membrane of said permeate production facility for producing permeate in cross-flow,

- measuring the rate of permeate produced by said permeate production facility in function of time in a first measuring step,

- iteratively performing steps of cleaning the membrane of said permeate production facility by at least one mechanical cleaning action, followed by measuring the rate of permeate produced by said permeate production facility in function of time,

- computing a reversible fouling propensity of the feed based on the rate of permeate which is recovered after each cleaning step, - computing an irreversible fouling propensity of said feed based on the rate of permeate which remains unrecovered after each cleaning step, and

- calculating the value of at least one first parameter related to a membrane cleaning action based on the reversible fouling propensity.

2. The method according to claim 1, further comprising the step of calculating the value of at least

one second parameter related to a membrane cleaning action based on the irreversible fouling propensity.

3. The method according to claim 1 or 2, wherein the step of feeding at least a part of the feed to the membrane of the permeate production facility comprises providing a flow of air bubbles along the membrane surface for transporting the feed.

4. The method according to claim 2, wherein in the iteratively performed measuring steps a larger flow of air bubbles is used compared to the first measuring step, preferably larger by at least 50%.

5. The method according to any one of the claims 1 to 4, wherein during the measuring steps the trans membrane pressure is held constant. 6. The method according to any one of the preceding claims, wherein the cleaning and the measuring steps are performed for at least 10 iterations.

7. The method according to any one of the preceding claims, wherein the first measuring step is performed until the rate of permeate has decreased to below a predetermined threshold.

8. The method according to any one of the claims 1 to 7, wherein in each iteration, the measuring step is performed for a predetermined period of time. 9. The method according to any one of the claims 1 to 8, wherein the cleaning steps comprise providing a flow of air bubbles along the membrane surface.

10. The method according to any one of the claims 1 to 9, wherein the cleaning steps comprise a relaxation step.

11. The method according to any one of the preceding claims, wherein said first parameter is related to a mechanical membrane cleaning action.

12. The method according to claim 11, wherein said mechanical membrane cleaning action is selected from the group consisting of: back-wash, back-pulse, aeration, relaxation, dosing of a floe modifying agent and the dosing of a coagulant.

13. The method according to any one of the claims 2 to 12, wherein said second parameter is related to a chemical membrane cleaning action.

14. The method according to claim 13, wherein said chemical membrane cleaning action is selected from the group consisting of: maintenance chemical cleaning, dosing of a floe modifying agent and the dosing of a coagulant.

15. The method according to any one of the preceding claims, comprising the step of using fuzzy set logic for calculating the value of said at least one first parameter and preferably also said at least one second parameter .

16. The method according to any one of the preceding claims, wherein the reversible and the irreversible fouling propensity are multi-value indexes, preferably in function of permeate volume.

17. A membrane filtration system, comprising:

- at least one membrane and a lower-level control system implementing a set of parameters arranged for actuating cleaning actions of said at least one membrane,

— a supervisory control system and a device for measuring a membrane fouling propensity of a feed applied to the membrane filtration system, the device for measuring coupled to the supervisory control system, wherein the supervisory control system and the device for measuring comprise means for carrying out the steps according to any one of the preceding claims.

18. The membrane filtration system according to claim 17, wherein the supervisory control system comprises a first output arranged for setting one or more first parameters of said set of parameters, the one or more first parameters related to mechanical cleaning actions, and wherein the supervisory control system is arranged for calculating the value of said one or more first parameters based on the reversible fouling propensity.

19. The membrane filtration system according to claim 18, wherein the supervisory control system comprises a second output arranged for setting one or more second parameters of said set of parameters, the one or more second parameters related to chemical cleaning actions and wherein the supervisory control system is arranged for calculating the value of said one or more second parameters based on the irreversible fouling propensity.

20. The membrane filtration system according to claim 18 or 19, wherein said mechanical cleaning actions are selected from the group consisting of: back-wash, back- pulse, aeration, relaxation, dosing of a floe modifying agent and the dosing of a coagulant.

21. The membrane filtration system according to any one of claims 17 to 20, wherein said chemical cleaning actions are selected from the group consisting of: maintenance chemical cleaning, dosing of a floe modifying agent and the dosing of a coagulant.

Description:

SUPERVISORY CONTROL SYSTEM AND METHOD FOR MEMBRANE CLEANING

Field of the Invention

[0001] The present invention is related to membrane filtration systems and to the cleaning of the membranes of said systems. More particularly, the present invention is related to pressure-driven membrane filtration systems.

State of the Art [0002] Membranes are used in filtration processes in order to clean liquids from contaminants and/or solid particles. During pressure-driven membrane filtration, a feed is separated into a "clean" filtrate fraction (i.e. the permeate) and a concentrated fraction as a result of a pressure difference across the membrane surface. The separation process is mechanical since the pores within the membrane act as a physical barrier towards constituents which are larger than the pores. As a result, at least all particles which are larger than those pores are withheld from the feed that is passing through the membrane. Due to a build-up of retained compounds at the membrane surface during filtration, deposition onto the membrane and obstruction of the membrane pores will take place to various extents. These processes of membrane fouling are inevitably linked to all pressure-driven membrane filtration processes and result in an increase in hydraulic resistance. As a consequence, either permeate production will decrease when operating at constant trans-membrane pressure (TMP) or the TMP has to be increased to maintain the desired permeate production.

[0003] As an example, the cleaning of waste water can be done in a very efficient way in a membrane bioreactor (MBR) at a high sludge concentration (called mixed liquid suspended solids or MLSS) while removing effluent through membrane filtration. The MBR filtration produces a very high quality effluent.

[0004] In a membrane bioreactor, the production of the MBR effluent is done by using a membrane. The membrane permeate constitutes the MBR effluent while the concentrate is remixed within the biomass (sludge) itself.

[0005] However, the efficiency of the filtration process largely depends on the membrane fouling by the sludge. Particles that are withheld by the membrane tend to deposit onto the membrane surface and obstruct the membrane pores. As a result, the hydraulic resistance increases, which would need a pressure difference increase in order to keep the permeate production constant. Membrane fouling hence reduces the permeate production rate significantly. [0006] Membrane fouling consists of extremely complex phenomena. Different types of fouling mechanisms act in parallel since there are numerous fouling constituents, e.g. particles, dissolved inorganics, dissolved organics, bacteria and floes, colloids and coagulated constituents. They interact with the membrane surface and membrane pores in very complex ways. The fouling constituents also interact with one another. [0007] In order to counteract the fouling of the membranes, a number of techniques have been developed in order to clean the membranes at defined intervals. Cleaning can be performed by mechanical actions or chemically. In the former case, a specific "mechanical" action is exercised on the fouling layer in order to disrupt it or to make it as thin as possible. Some "sticking" type of fouling can not be removed by mechanical action, but can

only be countered by chemically cleaning the membrane surface .

[0008] Membrane filtration systems of the prior art are equipped with a lower-level control system (e.g. PLC) in which specific parameter values, which are related to the mechanical and/or the chemical cleaning, are set according to the experience of a human programmer. Generally, these parameter values are only adjusted on an occasional basis. Hence, the operation of the membrane filtration system may not be optimized in cases in which the feed has a varying fouling propensity (which is the case in e.g. municipal waste water systems) . In a large number of cases the cleaning results to be performed either too often or too seldom and hence does not take into account the actual fouling propensity of the treated sludge. This results in membrane filtration systems which operate in a less optimized and less efficient way. [0009] A few measurement devices have been developed, which may give an indication of the fouling propensity of a sludge that may be applied as a feed to a membrane filtration system. These measurement devices however, either are not suitable to be used in an on-line or continuous measurement, or measure only a specific kind of fouling constituents present in the sludge, or they measure the total fouling propensity of the sludge.

[0010] One of these measurement devices is disclosed in "A new method of measuring and presenting the membrane fouling potential", E. Brauns et al . , Desalination 150 (2002), pp. 31-43. In order to carry out a measurement with this device, a part of the sludge has to be separated from the feed and pressurized. Thereafter, the pressurized feed is fed to a dead-end measurement membrane, and the permeate is measured in function of time. The measurement interval of this sensor is typically restricted to a small period

because of the fouling of the measurement membrane. For a new measurement to be carried out, the membrane has to be exchanged.

[0011] The latter fouling measurement method is hence restricted to the dead-end measurement of the total fouling propensity of a sludge. An indication on the total fouling propensity of a sludge is however of limited practical use for steering the membrane cleaning actions, as these would need a distinction to be made between mechanical and chemical cleaning, which are related to distinct fouling types.

Aims of the Invention [0012] The present invention aims to optimize the efficiency of a membrane filtration system. It is an aim of the invention to provide an automated and optimized control of membrane cleaning actions in a membrane filtration system, thereby obviating the shortcomings of the prior art. It is an aim of the invention to separately and individually optimize the control of the mechanical and chemical cleaning actions in a membrane filtration system. [0013] It is a further aim of the present invention to provide a membrane filtration system which operates at higher efficiency compared to membrane filtration systems of the prior art. The present invention aims to provide a system for controlling the mechanical and/or chemical membrane cleaning actions separately and automatically.

Summary of the Invention [0014] The present invention achieves the aims outlined above by providing a system and a method, as set out in the appended claims, for controlling one or more parameters governing one or more actions for cleaning membranes in a membrane filtration system.

[0015] According to a first aspect, there is provided a method of controlling membrane cleaning actions of a membrane filtration system. The method comprises a step of providing a device for measuring a fouling propensity of a feed applied to (membranes of) the membrane filtration system, wherein the device comprises a permeate production facility comprising a membrane. The method further comprises a step of feeding at least a part of the feed to the membrane of said permeate production facility. The method further comprises a step of applying a trans- membrane-pressure across the membrane of said permeate production facility for producing permeate in cross-flow. The method further comprises a step of measuring the (volume) rate of permeate produced by said permeate production facility in function of time in a first measuring step. The method further comprises a step of iteratively performing the steps of cleaning the membrane of said permeate production facility by at least one mechanical cleaning action, followed by measuring the rate of permeate produced by said permeate production facility in function of time. The method further comprises a step of computing a reversible fouling propensity of the feed based on the rate of permeate which is recovered after each cleaning step. The method further comprises a step of computing an irreversible fouling propensity of said feed based on the rate of permeate which remains unrecovered after each cleaning step. The method further comprises a step of calculating the value of at least one first parameter related to a membrane cleaning action based on the reversible fouling propensity.

[0016] The method preferably further comprises a step of calculating the value of at least one second parameter related to a membrane cleaning action based on the irreversible fouling propensity.

[0017] Said first and second parameters form part of a set of parameters related to membrane cleaning actions of the membrane filtration system.

[0018] Preferably, the step of feeding at least a part of the feed to the membrane of the permeate production facility comprises providing a flow of air bubbles along the membrane surface for transporting the feed. Said air bubbles are preferably of slug-type. More preferably, in the iteratively performed measuring steps a larger flow (rate) of air bubbles is used compared to the first measuring step. Said larger flow (rate) of air bubbles is preferably larger by at least 50%, more preferably by at least 75%, even more preferably by at least 100%. [0019] Accordingly, the first cleaning step in said iteration is preferably carried out with differing conditions compared to the subsequent steps in said iteration. The first cleaning step can be carried out during a longer period of time, such as at least 50% longer, preferably at least 75% longer and even more preferably at least 100% longer.

[0020] Preferably, during the measuring steps, the trans membrane pressure (TMP) is held constant. The rate of permeate is hence produced at constant TMP for each measuring step. [0021] Preferably, the iterative steps of cleaning and measuring are performed at least 10 times. [0022] Preferably, the first measuring step is performed until the rate of permeate has decreased to below a predetermined threshold. The rate of permeate can be normalised by pressure and membrane area to yield a permeability.

[0023] Preferably, in each iteration, the measuring step is performed for a predetermined period of time.

[0024] Preferably, the cleaning steps comprise providing a flow of air bubbles along the membrane surface

(aeration) .

[0025] Preferably, the cleaning steps comprise a relaxation step.

[0026] Preferably, said first parameter is related to a mechanical membrane cleaning action. Mechanical membrane cleaning actions can refer to physical cleaning.

More preferably, said mechanical membrane cleaning action is a back-wash. Said mechanical membrane cleaning action can be a back-pulse. Said mechanical membrane cleaning action is preferably aeration. Said mechanical membrane cleaning action can be relaxation.

[0027] Mechanical membrane cleaning actions can refer to the dosing of a floe modifying agent. They can refer to the dosing of a coagulant. The floe modifying agent and the coagulant can be dosed as a preventive measure .

[0028] Preferably, said second parameter is related to a chemical membrane cleaning action. More preferably, said chemical membrane cleaning action is maintenance chemical cleaning.

[0029] Chemical membrane cleaning actions can refer to the dosing of a floe modifying agent. They can refer to the dosing of a coagulant. The floe modifying agent and the coagulant can be dosed as a preventive measure.

[0030] Preferably, methods of the invention comprise the step of using fuzzy set logic for calculating the value of said at least one first parameter. Fuzzy set logic can be used for calculating the value of said at least one second parameter as well.

[0031] Preferably, the reversible and the irreversible fouling propensity are multi-value indexes.

They are preferably expressed in function of permeate

volume. The multi-value indexes are obtained with the data collected in the first and iteratively performed measuring steps. Preferably, the first and iteratively performed measuring steps are repeated in order to monitor the evolution of the multi-value indexes with time.

[0032] According to a second aspect of the invention, there is provided a membrane filtration system, comprising at least one membrane and a lower-level control system implementing a set of parameters arranged for actuating the cleaning actions of said at least one membrane. According to the invention, the membrane filtration system further comprises a supervisory control system and a device for measuring a membrane fouling propensity of a feed applied to said at least one membrane. The supervisory control system is preferably arranged for adjusting said set of parameters. Each parameter of the set of parameters may control or govern a specific cleaning action . [0033] The device for measuring is preferably coupled to the supervisory control system. The device for measuring is preferably separate from the membranes of the membrane filtration system. According to the invention, the supervisory control system and the device for measuring comprise means for carrying out the steps according to methods of the invention.

[0034] Preferably, the supervisory control system comprises a first output arranged for setting one or more first parameters of said set of parameters. The one or more first parameters are preferably related to mechanical (physical) cleaning. The supervisory control system is arranged for calculating the value of said one or more first parameters based on the reversible fouling propensity.

[0035] The supervisory control system can comprise a second output arranged for setting one or more second parameters of said set of parameters. The one or more second parameters are related to chemical cleaning. The value of said one or more second parameters is calculated based on the irreversible fouling propensity. [0036] The device for measuring the membrane fouling propensity of a feed preferably comprises a first output arranged for providing information on the reversible fouling propensity of said feed and a second output arranged for providing information on the irreversible fouling propensity of said feed. The first output of the device is preferably linked to the first input of the supervisory control system. The second output of the device can be linked to the second input of the supervisory control system.

[0037] Preferably, said membrane filtration system is pressure-driven. More preferably, said membrane filtration system is a membrane bioreactor. [0038] Preferably, in the membrane filtration system, the device for measuring the membrane fouling propensity of a feed comprises: a permeate production facility comprising a membrane, means for controlling and setting the trans-membrane-pressure across said membrane and means for measuring the permeate production rate through the membrane .

Brief Description of the Drawings

[0039] Figure 1 represents a scheme of the supervisory control of a membrane filtration plant.

[0040] Figure 2 represents an example of fuzzification of the temperature variable. Three fuzzy sets (A, B, C) are defined and the membership functions of the temperature value to the fuzzy sets are plotted.

[0041] Figure 3 represents the scheme of figure 1 in which the supervisory control system is implemented based on fuzzy set logic.

[0042] Figure 4 represents schematically a device for measuring the reversible and irreversible fouling characteristics of a mixed liquor.

[0043] Figure 5 represents a graph of the measurements carried out with a device of figure 4. The mass transfer coefficient (MTC) of the permeate, also referred to as permeability, is determined over a time period and for each measurement step. Between each step a relaxation and aeration is carried out for removing the reversible fouling layer on the measurement membrane.

[0044] Figure 6 plots the ratio of permeate volume to membrane area in function of the calculated ratio of hydraulic resistance due to reversible fouling to the hydraulic resistance of the membrane, based on the data of figure 5.

[0045] Figure 7 plots the ratio of permeate volume to membrane area in function of the calculated ratio of hydraulic resistance due to irreversible fouling to the hydraulic resistance of the membrane, based on the data of figure 5.

[0046] Figure 8 illustrates the evolution of the reversible fouling propensity with time for the mixed liquor of a membrane bioreactor being fed with municipal waste water, measured according to the invention.

Measurements were performed on the days as indicated.

[0047] Figure 9 illustrates the fuzzification of a reversible fouling fingerprint graph based on four normalised permeate volume on membrane area levels and five fuzzy sets.

[0048] Figure 10 illustrates the membership functions for the fuzzy sets applicable to the 50% level of figure 9.

Detailed Description of the Invention

[0049] Embodiments of the present invention will now be described in detail with reference to the attached figures, the invention is not limited thereto but only by the claims. The drawings described are only schematic and are non-limiting. In the drawings, the size of some of the elements may be exaggerated and not drawn on scale for illustrative purposes. The dimensions and the relative dimensions do not necessarily correspond to actual reductions to practice of the invention. Those skilled in the art can recognize numerous variations and modifications of this invention that are encompassed by its scope. Accordingly, the description of preferred embodiments should not be deemed to limit the scope of the present invention . [0050] Furthermore, the terms first, second and the like in the description and in the claims are used for distinguishing between similar elements and not necessarily for describing a sequential or chronological order. It is to be understood that the terms so used are interchangeable under appropriate circumstances and that the embodiments of the invention described herein are capable of operation in other sequences than described or illustrated herein. [0051] Moreover, the terms top, bottom, left, right, over, under and the like in the description and the claims are used for descriptive purposes and not necessarily for describing relative positions. The terms so used are interchangeable under appropriate circumstances and the embodiments of the invention described herein can operate in other orientations than described or illustrated herein.

For example, "left" and "right" of an element indicates being located at opposite sides of this element. [0052] It is to be noticed that the term "comprising", used in the claims, should not be interpreted as being restricted to the means listed thereafter; it does not exclude other elements or steps. Thus, the scope of the expression "a device comprising means A and B" should not be limited to devices consisting only of components A and B. It means that with respect to the present invention, A and B are relevant components of the device.

[0053] A very important aspect of membrane fouling knowledge is the distinction between reversible and irreversible fouling. The reversible fouling component is linked to the "loose" deposition of fouling constituents, possibly within the membrane pores but mostly on top of the membrane surface. The deposited "layer" comprises a specific porosity which allows for the supernatant to be transported through that fouling layer and further through the membrane. Evidently, this deposited layer encompasses a hydraulic resistance of its own and causes an additional transport energy. Obviously, this energy waste should be minimized. The reversible fouling layer is mechanically weak and can be disrupted through mechanical actions such as aeration (scouring of the membrane surface by air bubbles in order to remove fouling constituents deposited on the membrane surface) , back-washing, back-pulsing and relaxation. Although the term "irreversible" fouling in the strict filtration sense would mean "irrecoverable" it is however used frequently in the literature and by researchers in the field to indicate the type of fouling which can be recovered by chemical cleaning. Hence, the reversible fouling can be removed by mechanical actions. On the other hand, the sticky type of fouling, also referred to as the irreversible fouling, can not. Irreversible

fouling can only be removed by the action of cleaning chemicals .

[0054] Therefore, as the mechanical and chemical cleaning actions are related to different types of fouling, an automated control of those cleaning actions can only be optimized when separate knowledge (information) about the reversible and irreversible fouling of the membranes is gathered. The inventors have developed a measurement method which allows to gain information on the reversible and on the irreversible fouling propensity of a feed.

[0055] The dead-end measurement device of Brauns et al . described above (see the prior art section), can not distinguish between reversible and irreversible fouling propensity, because the membrane element used for the measurement is dead-end and hence can not be cleaned mechanically. The increase of hydraulic resistance measured by such a dead-end device is hence related to the total fouling (sum of reversible and irreversible fouling) . [0056] The present invention therefore provides a device that is able to monitor (to measure) both the reversible and the irreversible fouling propensity of a feed. Such a measurement device opens the way for automating and optimizing the cleaning actions of a membrane filtration system (MFS) . Figure 4 shows schematically an embodiment of a measurement device of the present invention for measuring the fouling propensity of a feed (e.g. a feed to a MFS) . The measurement device is based on a measurement membrane which is made to operate in cross-flow . [0057] Figure 4 shows a scheme of the measurement device 40 with regard to a tubular membrane layout, but layouts for other types of membrane configurations are equally possible, such as for e.g. flat or capillary membranes .

[0058] Device 40 comprises a permeate production facility 42. Permeate production facility 42 advantageously comprises an open-end tubular membrane 421 for carrying out an inside-to-outside filtration of the feed liquid 41. In that case, the feed liquid is fed through the inner lumen 424 of the tubular membrane 421. Permeate is collected at the outside of the tubular membrane, where the permeate production facility provides a sealed, confined space 422. [0059] The feed, as used in disclosing the present invention, refers to a feed as applied at the surface of the membranes of a membrane filtration system. Hence, for a membrane bioreactor for example, the feed refers to the mixed liquor, or activated sludge, being a mixture of waste

(sewage) water with colloidal particles, bacteria, etc. resulting from a biological treatment.

[0060] By way of example, feed liquid 41 is considered to be a sludge. In the present embodiment, permeate production facility 42 is suspended in the liquid 41. Such suspension can be obtained by providing a float 44 to which facility 42 is attached by fastening means 45 (such as e.g. strings or straps) .

[0061] Device 40 further comprises means 43 for producing the permeate at a specific TMP and to measure the corresponding permeate production rate. It is equally possible to maintain a specific permeate production rate and measure the corresponding TMP. However, in order to ensure that there is no influence of the compaction of the fouling layer due to a non-constant TMP, it is advantageous to keep the TMP constant during permeate production. The thus measured (calculated) decrease of membrane permeability can in the latter case be regarded as solely due to fouling and as not being influenced by compaction effects .

[0062] The TMP is preferably selected to ensure a high permeate flux. A high flux advantageously increases fouling rate and allows accurate measurements over even short time periods. The TMP is preferably selected to be less than or equal to 1 bar, more preferably less than or equal to 0.5 bar and most preferably less than or equal to 0.2 bar.

[0063] In the particular embodiment of figure 4, means 43 comprise a vessel 431 for draining away the permeate from facility 42 and a permeate flow meter 432. The permeate is drained from facility 42 by maintaining a subpressure (e.g. 1 bar or less) in vessel 431 by vacuum pump 433. Pressure transducer 434 measures the trans- membrane-pressure (TMP) across membrane 421. The TMP is hence caused by the subpressure in vessel 431, which causes a subpressure in a confined space 422 around the outside of the membrane 421. The confined space 422 is delimited by the membrane 421 and facility 42 and is sealed such that permeate may enter only through the membrane 421. [0064] According to a preferred embodiment, device 40 comprises an aerator 426 at the bottom of membrane 421. Therefore, a compressor 435 produces pressurized air, which is fed to the aerator 426, at the bottom of membrane 421. Large air bubbles 423 are formed, which rise upwards in the tube 424 of the membrane 421. The air bubbles, while rising upwards, transport the sludge 41 upwards along the inner cavity (lumen) 424 of the tubular membrane. The rising air bubbles are of a slug shape and are very efficient in the transporting the sludge along the inner cavity 424. Due to the TMP, part of the sludge permeates through the membrane and generates a cross-flow through the membrane surface. The rising air bubbles 423 may cause turbulences in front of and in the wake behind a bubble.

[0065] A mass flow controller 436 can further be provided in order to monitor and control the flow rate of air 423 fed to facility 42. Device 40 preferably comprises a temperature sensor 425 for measuring the temperature of the sludge for calculating viscosity, which is needed for calculating a mass transfer coefficient (MTC) . [0066] The device of the invention preferably further comprises dedicated software for the data acquisition and data handling/calculation. Such a software may comprise proprietary MeFiAS® software by VITO, Belgium.

[0067] The measurement device 40 measures the fouling propensity of a feed (sludge) in cross-flow conditions and over longer periods of time (several hours or longer) . The measurement device 40 is arranged to determine from the measured data both the reversible and the irreversible part of the fouling.

[0068] The method for measuring the total fouling propensity of the sludge comprises the steps of exerting a specified TMP across the membrane 421 and measuring the permeate production rate in function of time, over a chosen time period. According to the formula described in E. Brauns et al . cited in the prior art section,

wherein:

V = permeate volume (m 3 ) t = time (s)

AP = transmembrane pressure drop (Pa) η = absolute viscosity (kg/m.s) R n , = membrane resistance (m "1 )

Rf = all additional resistance from fouling (m "1 )

A = membrane surface area (m 2 ) .

With all other variables considered constant, a change in permeate flow rate (dV/dt) is caused by a progressive membrane fouling and hence a progressive increase in the hydraulic resistance Rf due to fouling. Hence, information on the fouling propensity of the sludge can be gained by measuring dV/dt.

[0069] In order to measure the reversible and the irreversible fouling propensity of the sludge, it is known that the hydraulic resistance R f due to fouling can be regarded as a series resistance of a resistance Rf, re v due to reversible fouling and a resistance Rf rirr due to irreversible fouling: R f = Rf, re v + Rf,irr- The fouling measurement is performed in a stepwise/cyclic manner as follows . [0070] In a first step, a low air flow rate inside the tubular membrane 421 is used. This results in a low turbulence and causes a high amount of reversible fouling, resulting from a thicker floe layer deposition on the membrane inner wall surface. While maintaining the low air flow rate, the permeate production rate is measured over a given time period.

[0071] Alternatively, in the fist step, the permeate production rate can be measured until the permeability of the membrane, or the permeate rate has decreased to below a predetermined threshold. The permeability of the membrane is proportional to the inverse of the total hydraulic resistance (R m + Rf) of the membrane (and the fouling layer attached thereto) . The permeate production rate can be measured until the permeability of the membrane has decreased to a value falling in the range between 10% and 50% of the initial value (i.e. of a clean membrane, with no fouling) , preferably to a value falling in the range between 20% and 50% of the initial value, more preferably

to a value falling in the range between 20% and 40% of the initial value.

[0072] In this first step, the simultaneous build up of reversible and irreversible fouling is of course not avoidable and during the measurement of the first step the total additional resistance data R f (sum of reversible and irreversible fouling) is thus obtained.

[0073] The first step can be preceded by a conditioning step. In the conditioning step, the feed through the permeate production facility is stabilized, so as to obtain a homogeneous filling of the inner lumen of the tubular membrane 421 with feed liquid (sludge) . This can be achieved by a relatively high aeration (large flow of air bubbles) . [0074] In the second and subsequent steps, a cleaning step is carried out first, in which a higher air flow rate is used to remove the reversible layer while providing a relaxation at the same time (thus no permeate production) . Such a relaxation comprises removing the TMP (no trans-membrane pressure gradient) . This removes the reversible fouling layer. Optionally, other mechanical cleaning actions (e.g. a back-wash) may be performed to remove the reversible fouling layer. Different mechanical cleaning actions may also be combined. [0075] Thereafter, a measurement step according to the measurement procedure of the first step is repeated, either under same conditions or not. The starting measurement data of the second and subsequent steps then corresponds to a permeability recovery. The permeability that is not recovered is due to the remaining irreversible fouling part. The TMP in the second and subsequent steps may differ from the TMP exerted in the first measurement step in cases wherein it is desired to keep the permeate

flux constant. For reasons stated above, it is advantageous to keep the TMP constant.

[0076] The air (bubble) flow during the second and subsequent measurement steps can be different from the air bubble flow of the first step. Preferably, the air bubble flow rate in the second and subsequent measurement steps is higher than the air flow rate in the first step, with values of at least 50% higher (compared to the air flow rate of the first step) being preferred, values of at least 75% higher being more preferred and values of at least 100% higher being particularly preferred.

[0077] The air bubble flow rate of the cleaning steps and the measurement steps is preferably not the same. The cleaning steps are preferably carried out with a higher aeration rate (air flow rate) . The air flow rate during the cleaning step is preferably at least 10% higher and more preferably at least 20% higher than the air flow rate of the (second and subsequent) measurement steps. [0078] Figure 5 shows a graph of consecutive measurements performed with a fouling measurement device of the invention. The graph plots different measurement data sets of permeate mass transfer coefficient (MTC) in function of time. The mass transfer coefficient refers to a permeate rate normalised by the membrane area and TMP (units l/h.m 2 .bar) .

[0079] Set 51 refers to the first measurement step. Set 52 refers to the second measurement step, after removal of the reversible fouling layer on the membrane 421. The difference between the first measurement point 511 of the first step and the first measurement point 521 of the second step, namely an MTC difference of about 1600 l/h.m 2 .bar, is related to an irreversible fouling of the membrane 421. That irreversible fouling is subtracted from set 51 in order to calculate and extract the reversible

fouling information. Indeed, it is assumed that the irreversible fouling between the last measurement of the first measurement step (5In) and the first measurement of the second measurement step (521) is the same. [0080] The difference between the last measurement point 51n of the first step and the first measurement point 521 of the second step, namely an MTC difference of about 1500 l/h.m 2 .bar has been recovered by the mechanical cleaning actions (aeration and relaxation) between the first and the second step. From the combination of curve 51 and point 521 from curve 52 first values for the reversible and irreversible fouling can be calculated.

[0081] As a result of the increasing irreversible fouling, the recovery is smaller with each subsequent step. This effect of a decreasing permeability in each subsequent step is also illustrated in figure 5. The measurement method allows in this way to extract both the reversible and irreversible parts of the fouling. [0082] Further calculations on the measured data may be performed. Figures 6 and 7 plot some results. Figure 6 shows a reversible fouling curve and figure 7 shows an irreversible fouling curve. The plots of figures 6 and 7 represent the ratio of permeate volume V to membrane area A versus the ratio of the hydraulic resistance due to respectively the reversible and the irreversible fouling to the membrane resistance.

[0083] The plots of figures 6 and 7 can be obtained as follows. At each measurement point on the graph of figure 5, R fr rev (the resistance due to the reversible fouling layer) and Rf,i rr (the resistance due to the irreversible fouling layer) is calculated by considering that the total resistance R = R m + R f = R m + Rf r rev + Rf,irr, can be extracted from the permeate rate dV/dt according to the formula indicated hereinabove. The (clean) membrane

resistance R n , can be extracted by regression from the first measurement point 511 (of the first measurement step) of fig. 5. Once R n , is known, R f can be calculated at each sampling (measurement) point. As it is assumed that the cleaning steps remove all reversible fouling, the R f at each first sampling point (521) of a measurement step equals Rf rirr at that sampling point. The assumption also implies that Rf,irr at each last sampling point (5In) of a measurement step equals Rf rirr at each first sampling point (521) of the subsequent measurement step. Based thereon, Rf r r e v at each last sampling point can be calculated. The values of Rf, re v and Rf, 1T r at each sampling point of a measurement step can be calculated by assuming a relationship between Rf, ir r and the permeate rate (or permeate volume) , such as a direct proportionality. As a result, Rf,rev and Rf /irr can be calculated for each sampling point, and the graphs as represented in Figs. 6 and 7 can be obtained. [0084] The fingerprint of the reversible and the irreversible fouling is hence obtained in the form of multi-index values. The multi-index values are advantageously a function of the permeate volume

(normalised) . The entire multi-index values (i.e. the graphs of figs. 6 and 7) can change with time, e.g. due to a varying feed. Hence, in order to monitor the reversible and irreversible fouling propensity of a feed to a membrane filtration system, the measurement procedure as identified above is repeated in time to yield separate graphs (multi- index values) as in figs. 6 and 7 for each complete measurement procedure.

[0085] This is illustrated in figure 8 for the reversible fouling propensity only. The evolution of the reversible fouling propensity graphs for different days is plotted. The data of fig. 8 was obtained by a measurement

device 40, according to the invention, comprising a tubular membrane made of poly (vinylidene fluoride) (PVDF) having an inner diameter of 5.2 mm, a length of 680 mm and a nominal pore size of 0.03 μm. The membranes used for the measurements all had a (normalised) clean water permeability (MTC) higher than 1000 1 m "2 h "1 bar "1 at a constant TMP of 0.1 bar. The feed consisted of municipal wastewater, forming a mixed liquor, pre-screened at 0.75 mm. [0086] The measurement protocol started with a conditioning step during which 100 ml min "1 of air is supplied at the bottom of the membrane tube, to ensure that the mixed liquor is homogenously distributed inside the membrane lumen when filtration starts. A first measurement cycle was carried out, consisting of a filtration step (step IA) . The filtration step was performed at a constant TMP of 0.10 bar and at an air flow rate of 40 ml min "1 , corresponding to a superficial gas velocity of 0.03 m s "1 . Step IA continued until permeability had decreased to 30% of the initial permeability. This ensures that the fouling process has progressed sufficiently to provide useful information about the reversible fouling propensity of the mixed liquor. The following cycles, each consisted of a physical cleaning step (relaxation step) , followed by a filtration (measurement) step. The first cleaning step consisted of a 10 min relaxation at an air flow rate of 100 ml min "1 (specific aeration demand of 0.6 Nm 3 m ~2 h "1 ) . Operational conditions for the subsequent filtration steps were a constant TMP of 0.10 bar, an air flow rate of 80 ml min "1 (superficial gas velocity of 0.07 m s "1 ) and a duration of 5 min. Removal of the remaining reversible fouling during the second and subsequent cleaning steps was achieved through a 3 min relaxation step at an air flow of 100 ml min "1 . At least 10 cycles were applied to ensure an

adequate measurement of the irreversible fouling propensity.

[0087] It is clear from figures 6 and 7 that the reversible and irreversible fouling propensity measured (and calculated) by the measurement device of the invention may be represented as curves and thus need not be a single point value. The measurement approach according to the invention proves in that respect to show a much more detailed fingerprint about the fouling behaviour when compared to classic fouling measurement methods which typically consist of only a one-index value, which then is assumed to completely model the very complex fouling behaviour as an equation y=a with a the fouling index number . [0088] The fouling measurement device of the invention thus allows to measure both the reversible and irreversible fouling propensity of the sludge to induce membrane fouling. At the end of an entire measurement cycle, e.g. the cycle corresponding to the whole of the graph of figure 5, the membrane 421 of facility 42 is not useful anymore for further measurements. It is fouled too much. For a new measurement cycle, the membrane 421 should be replaced by a new or clean membrane. The removed membrane may be disposed of or may be cleaned in order to remove the irreversible fouling layer.

[0089] These actions may be carried out manually, e.g. by an operator. Preferably, they are performed in an automated way. Therefore, facility 42 of measurement device 40 may be provided with a membrane loader (not shown in figure 4) . The membrane loader loads a clean membrane at the beginning of a new cycle of measurements and removes the fouled one. Preferably, the membrane loader is arranged to keep a stock of several clean membranes. The loader then may put the fouled membranes in a stack such that at

defined time intervals they may be collected for cleaning or disposal. The loader then may be loaded with a new batch of clean membranes. More preferably, the loader is provided with a facility to chemically clean the membranes. [0090] By so doing, the measurement device 40 is able to continuously monitor the fouling propensity of the sludge. The measurement device produces an output evaluating the reversible and irreversible fouling characteristics of the sludge. [0091] Having disclosed how the reversible and irreversible fouling characteristics of a feed can be measured in a continuous and automated fashion, the present invention now discloses a system for optimizing and automating the control of the cleaning actions of a membrane filtration system (MFS) . Such a system calculates the parameters relating to the cleaning actions on the membranes (e.g. time interval for cleaning, duration, etc.) in function of the quality of the feed (reversible and irreversible fouling characteristics) . This leads to a more efficient operation of the membrane filtration system.

[0092] In order to achieve this goal, the invention provides a supervisory control system (SCS) and a method of supervisory control, which supersedes the lower-level control system of an MFS, the latter control system being known in the art. Figure 1 shows this schematically. A membrane filtration system 10 is controlled by a lower- level control system 11, which may be a PLC-based control or a MeFiAS® control system (which is proprietary control software/hardware of VITO, Belgium) for controlling e.g. the actuators for the cleaning actions (e.g. the mechanical and/or chemical cleaning) .

[0093] The lower-level control system 11 is implemented with set-point parameters 12, relating to the cleaning actions. The parameters 12 may refer to time

intervals, dosing of chemicals, etc. and will be discussed more into detail later on. In the prior art, the parameter values are fine-tuned when the MFS is installed, and may be adjusted occasionally later on. In the present invention, a supervisory control system 13 is arranged to continuously

(i.e. at regular time intervals) adjust the set-point parameters 12 upon the calculation of new values for them.

[0094] In order for the supervisory control system

13 to calculate appropriate values for the set-point parameters 12, it needs input data. A primary input data is information on the reversible and irreversible fouling of the membranes and/or information on the reversible and irreversible fouling propensity of the feed. Therefore, the supervisory control system comprises a first input 141 for gaining information on the reversible fouling (of the feed or of the membranes themselves) and a second input 142 for gaining information on the irreversible fouling (of the feed or of the membranes themselves) . The reversible and irreversible fouling information is measured by one or more measurement devices 15. It is evident that the optimization of the control of the membrane cleaning actions is fully linked to the fouling propensity of the feed (or the fouling of the membranes) and therefore measurements for evaluating the reversible and irreversible fouling propensity are crucial as inputs 141 and 142 to a supervisory control system 13.

[0095] The measurements of a fouling sensor are thus always primary inputs 141 and 142 of the SCS. The primary reversible and irreversible fouling information may be complemented by other measurement data, such as the EPS concentration (extracellular polymeric substances) , feed temperature, dissolved oxygen, pH, biomass concentration (MLSS), etc. This data is provided by measurement devices 16 and is made available to the supervisory control system

at one or more inputs 17. It is important to note that for specific additional data adequate sensors and measuring devices (methods) are needed.

[0096] The measurement device 15 may perform its measurements on a part of the MFS 10 itself, or on a sample 19 withdrawn from the MFS (e.g. part of the feed withdrawn before it enters the MFS) and fed to the fouling sensor 15. [0097] Preferably, the measurement device 15 is a device which is separate and distinct from the MFS 10. The measurement device 15 is arranged to receive a feed, which is at least a sample of the feed applied to the membranes of the MFS 10.

[0098] The data arriving at inputs 141, 142 and 17 of the supervisory control system 13, is collected, interpreted and based thereon the values of the set-point parameters 12 are calculated. These newly calculated values are made available at an output 18 of the supervisory control system. They are written into the set 12 and used by the lower-level control system 11. [0099] As the inputs 141 and 142 provide distinct information related to respectively reversible and irreversible fouling, a distinct adjustment/calculation of each of the parameters 12 may be performed, based on either the reversible or the irreversible fouling information. Hence, a first set of parameters 12 related to mechanical cleaning actions may be controlled (adjusted, calculated, etc.) based on information available at input 141

(reversible fouling information) . A second set of parameters 12 related to chemical cleaning actions may be controlled based on information available at input 142

(irreversible fouling information) . Such a supervisory control system enables to optimise the operation of membrane filtration plants (optimized membrane surface,

less aeration, less mechanical and chemical constraints on the membranes, etc.) .

[0100] The cleaning actions in a membrane filtration system can be grouped into two types, viz. the mechanical and chemical cleaning actions, which counter two specific types of fouling, respectively the reversible fouling and the irreversible fouling. The supervisory control system is provided with inputs 141 and 142, each input relating to only one of the two cleaning types. This allows to achieve an improved optimization of the cleaning actions and a more efficient operation of the MFS. The input 141 gives information on the reversible fouling component of the total fouling and hence allows for the control of the mechanical membrane cleaning oriented actions. The input 142 gives information on the irreversible fouling component and allows for the control of the chemical membrane cleaning actions. The information supplied at inputs 141 and 142 may be provided by a measurement device 40. [0101] Two main control groups can be considered within the control actions of a supervisory control system of a membrane filtration plant. A first group is related to the reversible fouling and therefore to the related mechanical cleaning controls or to a reversible fouling reduction strategy. A second group is related to the irreversible membrane fouling effects, which can not be removed by the mechanical actions as used for reversible fouling but only by chemically cleaning the membrane. [0102] Having regard to the first group, the reversible membrane fouling control can be tackled in different ways. The reversible fouling component is linked to the "loose" deposition (e.g. weak Van der Waals forces) of fouling constituents, possibly within the membrane pores but mostly on top of the membrane surface. The deposited "layer" comprises a specific porosity which allows for the

supernatant to be transported through that fouling layer and further through the membrane. Evidently, this deposited layer encompasses a hydraulic resistance of its own and therefore such layer causes an additional transport energy (pumping energy) . Obviously, this energy waste should be minimized in an MFS and therefore also the hydraulic resistance according to this layer. The reversible fouling resistance thus should be minimized. [0103] For the particular case of membrane bioreactors, the reversible fouling layer behaviour may be quite complex as a result of the complicated composition of the bioreactor sludge which consists of floes (bacteria) and suspended solids. Since the reversible fouling layer is mechanically weak it is possible to exercise a specific "mechanical" action on the layer in order to disrupt it or to make it as thin as possible. On the other hand, from the weak mechanical characteristics of the fouling layer it is also clear that the trans-membrane-pressure (TMP) exerted during filtration provokes a mechanical pressure gradient across the fouling layer. A sustained pressure gradient can cause a compression of the reversible fouling layer which increases the hydraulic resistance of the fouling layer. A temporary removal of the pressure gradient (called relaxation) thus can be beneficial, in a way that the unwanted compression can be countered. As a result, the fouling layer is restored to a more loosely bounded system which then can be removed more easily by a parallel/succeeding mechanical action or preventive action

(dosing of coagulant) . [0104] Different types of mechanical actions can be applied: back-wash, back-pulse, aeration and relaxation. These are outlined in what follows.

[0105] A back-wash is often used in pressure driven membrane filtration in order to remove regularly the

reversible fouling layer on top of the membrane surface. A back-wash action comprises sending back a specific amount of permeate through the membrane to flush away the non- sticking fouling layer. This action should be fine-tuned in order to spill only a minimum of permeate and to minimize the interruption time of permeate production. Possible set- point parameters related to the back-wash action are the:

— back-wash flux: a high flux can be beneficial for extensive and rapid membrane cleaning, - back-wash duration: an efficient backwash needs a minimum of back-washing time and the

— back-wash interval: a shorter interval time between two back-wash actions is needed with a higher reversible fouling potential of the sludge. [0106] A back-pulse is in fact a high amplitude, very short duration type of back-wash. For more information on this type of mechanical cleaning action, the reader is referred to document EP 1043053. The shock-wave-resembling back-pulse can be very efficient with respect to an improved all-membrane-surface-area cleaning. Since the membrane reversible fouling layer is not necessarily homogeneous over the complete membrane surface area, it is possible that during a common back-wash action the permeate will flow through the lower hydraulic resistance zones but not through the somewhat higher hydraulic resistance zones. Therefore a common back-wash could be of a lower efficiency when compared with the shock-wave type of back-pulse action. Possible set-point parameters related to the back- pulse action are the: - back-pulse amplitude: a higher amplitude means a higher exerted cleaning energy,

— back-pulse duration: a longer pulse increases the removed quantity of the reversible fouling layer and the

— back-pulse interval: a shorter time interval between two back-pulse actions is needed with a higher reversible fouling potential of the sludge.

[0107] Having regard to aeration cleaning, in membrane filtration practice different submerged membrane modules may be used. Such modules may consist of flat, capillary or tube-like membranes. Independent of the membrane type, mechanical cleaning during filtration is performed by aeration. During aeration, air bubbles are produced which travel through the membrane module and thereby scour the membrane surface. As a result of this membrane scouring, the reversible fouling layer is continuously disrupted by the action of the air bubbles such that the loosely bounded fouling layer is removed to a specific extent. In fact, the air bubbles create turbulence at the membrane surface and therefore the aeration effect can be considered as a cross-flow like effect. It is known that this aeration is an important energy-consumer within the MFS. Hence the importance of an adequate control of the aeration. Possible set-point parameters related to the aeration action are the:

— aeration air bubble flow: with a higher flow the turbulence at the membrane surface is higher and therefore also the mechanical cleaning effect with respect to the removal of the reversible fouling layer,

- aeration air bubbling duration: a longer duration of the scouring increases the removal effect of the reversible fouling layer and the

- aeration air bubbling interval: in the case of a continuous aeration, the aeration interval is not interrupted, but in the case of an intermittent aeration, a shorter interval is needed with a higher reversible fouling potential of the sludge; a burst type

of aeration (temporarily burst) could have an efficient mechanical cleaning effect in some cases.

[0108] Concerning relaxation cleaning, since the temporary removal of the TMP pressure gradient across the reversible fouling layer loosens the reversible fouling layer, a modulated filtration mode may be used in membrane filtration, and membrane bioreactor filtration in particular. Permeate is produced during a specific filtration interval, being followed by a relaxation interval, in which permeate is not produced. During the relaxation interval the TMP gradient becomes zero. This allows for the fouling layer to get less compacted during the relaxation interval and therefore allows to be removed more easily by mechanical action (aeration, backwash, etc.). Possible set-point parameters related to the aeration action are the: - relaxation duration: an increased relaxation period simplifies the removal of the reversible fouling layer and the - relaxation interval: a shorter relaxation interval is needed with a higher reversible fouling potential of the sludge .

[0109] In addition to the above types of mechanical cleaning, in some cases a suitable sludge floc-modifying agent is found to be effective with respect to the reduction of the reversible fouling tendency. The floc- modifying agent causes the creation of a floe population with a higher permeability than in normal circumstances on the membrane surface, or possibly with a lower mechanical compaction sensitivity. In these cases, the dosing of such an agent could be considered. As the addition of floc- modifying agent works primarily on the reversible fouling layer, it may be considered as an additional type of

mechanical cleaning action, or at least as an action in support of the mechanical cleaning. Possible set-point parameters related to the dosing of floc-modifying agent could be defined as: — floc-modifying agent dose and the

- floc-modifying agent dosing frequency.

[0110] Besides the reversible membrane fouling effects, a second important fouling factor is the irreversible fouling. This "sticking" type of fouling can not be removed by the mechanical actions for removing the reversible fouling layer described above. In practice, the irreversible type of membrane fouling can only be countered by chemically cleaning the membrane surface. The chemical cleaning can restore part of the original permeability of the membrane but not the irrecoverable part. A second control group may be comprised in the supervisory control system, which is related to the control of the chemical cleaning actions. [0111] Mostly, two types of chemical cleaning are used: a regular maintenance chemical cleaning (MC) and an intensive chemical cleaning (IC) . An IC is only done a few times (or even once) a year and therefore may or may not need to be incorporated in the supervisory control actions, since it is done in a pragmatic way, based on experience. The MC however is determined by the irreversible fouling propensity of the sludge. Possible set-point parameters related to the MC are the:

- chemical cleaning frequency: a higher irreversible fouling rate evidently requires a more frequently executed MC and the

- chemical cleaning dose: an increased irreversible fouling propensity of the sludge requires a higher chemical dose.

Typical chemical cleaning agents that are used for membrane surface cleaning (while trying to restore the membrane resistance) at a specific dosage are sodium hypochlorite, alkaline solutions (e.g. NaOH), acid solutions (e.g. HNO3) , surfactants and enzyme cleaners. Such chemicals can be used solely or combined (in series) and eventually at a higher temperature (e.g. 30 0 C or 40 0 C) in order to raise the efficiency of the cleaning agents. [0112] A preventive action may be an alternative option in order to avoid - completely or partly - the occurrence of irreversible fouling. This can be done by using specific coagulants which envisage to neutralize the irreversible fouling constituents in the sludge

(supernatant) . Set-point parameters according to the corresponding coagulant dosing actions are the:

- coagulant dose and the

- coagulant dosing frequency.

[0113] As stated above, membrane fouling consists of very complex phenomena which in most cases can not be described adequately by mathematical models. There are numerous fouling constituents which act in parallel. These fouling constituents interact with the membrane surface and membrane pores in very complex ways. Therefore, a universally applicable mathematical fouling model which allows to accurately predict the fouling behaviour of a membrane is very difficult to implement.

[0114] The solution to the modelling problem of the fouling adopted in the present invention therefore is to implement in the supervisory control system a set of control rules (mathematical, fuzzy, etc.) which mimic to a certain extent a human operator who uses heuristic knowledge in order to control the cleaning actions of a membrane filtration plant in an acceptable way. This

heuristic knowledge is largely based on membrane fouling knowledge and in particular on reversible and irreversible membrane fouling knowledge.

[0115] The reversible and irreversible fouling knowledge is obtained, according to the present invention, by way of measurement procedures as identified above. The knowledge is advantageously expressed as a hydraulic resistance (normalised) related to the reversible or the irreversible fouling component of a feed. Those hydraulic resistances can be obtained for various permeate volumes (normalised) as multi-value indexes. The indexes can be monitored at regular intervals in time.

[0116] The supervisory control system can be implemented with control rules based on classical mathematical regression analysis of the above multi-value indexes. In the case of regression analysis, the multi- value indexes (graphs) can be represented by an adequate mathematical equation. Another approach, according to the invention, is to implement the control rules based fuzzy set logic. In the latter, the multi-index values can be converted into fuzzy sets, the values of which trigger actions of the supervisory control system according to the implemented control rules. [0117] An embodiment of a supervisory control system and method implemented with fuzzy set logic is outlined in what follows.

[0118] The calculations and control actions taken within the supervisory control system 13 may be performed based on fuzzy set logic (FSL) . In FSL, the value of a variable - e.g. temperature - is accorded one or more fuzzy sets - e.g. low, medium and high. The strength of FSL lies in the fact that the borders delimiting the fuzzy sets need not be crisp. For example, a temperature value of 20 0 C may belong both to the set "low" and to the set "medium", i.e.

the temperature value of 20 0 C has a membership of both the set "low" and the set "medium". This is shown in figure 2, where the membership functions of temperature values are drawn for three fuzzy sets: A (low in the previous example), B (medium) and C (high) . Referring to figure 2, the temperature value of 20 0 C has a membership of 0.5 for set "low" and 0.5 for set "medium". In a fuzzy development environment, control rules may be implemented, based on the fuzzy sets, which allow for actions to be undertaken. [0119] The implementation of fuzzy set logic in a supervisory control system of the invention may be implemented as shown in figure 3. The SCS 13 comprises a control block 131, 132, 133, etc. respectively for each set-point parameter 121, 122, 123, etc. from the set 12. Each control block comprises a fuzzification module 30, a fuzzy inference module 31, a fuzzy rule module 32 and a defuzzification module 33. The role of each of these modules is outlined in what follows. [0120] The fuzzification module 30 translates the measured values of one or more inputs to the SCS (reversible and irreversible fouling and additional input parameters) into linguistic variables, linked to predefined fuzzy sets. Each input is linked to a linguistic variable. E.g. the linguistic input variables "F rev ", "F irr " and "Temp" could be defined. In this way all possible relevant input parameters in the supervisory control can be defined as linguistic variables. For all these linguistic variables the according fuzzy sets need to be defined. [0121] With regard to the fuzzification of the reversible and irreversible fouling information, it is to be noted that, as this information is available as multi- value indexes (see the graphs of figs. 6 and I) 1 a fuzzification is preferably performed based on the multi- value index (or the graph) as a whole. According to an

embodiment, the latter whole content approach can be obtained by using a fuzzy image recognition method, as is explained further for only the reversible fouling fingerprint index (graph) . The method is completely analogous for the irreversible fouling fingerprint.

[0122] A fuzzy control block embodies a fuzzification module, an interference module, a fuzzy rules module and a defuzzification module as set out above. Such set-up can also be used in an image recognition environment to evaluate a multi-value index as a whole, or when the multi-value index is represented as a graph, the "shape" or position of a graph. Since the reversible and irreversible fouling fingerprint graphs, e.g. those in figs. 6 and 7, can be assumed to lie at all possible locations between the extreme of a vertical line (coinciding with the vertical y- axis) and the extreme of a horizontal line (coinciding with the horizontal x-axis) , it is possible to define such extreme locations as respectively 0% and 100% fouling propensity of the feed to the MFS. [0123] Referring to figure 6, a theoretically ideal vertical fouling graph indeed indicates that the system can produce an indefinite amount (V/A) of permeate without any increase of the total (reversible) hydraulic resistance: thus 0% fouling propensity of the feed. A theoretically horizontal fouling graph indicates the worst possible situation, since the system can produce no permeate at all: the hydraulic resistance increases immediately to a very high value, thus 100% fouling propensity of the mixed liquor. All real-case graphs thus will be located between the two extremes.

[0124] The image recognition evaluation of the position or the shape of a fouling fingerprint graph by a human operator could be done in a visual way by linking a percentage between 0% and 100% fouling propensity to that

global position. Such human statement of a fouling percentage can also be mimicked by a fuzzy approach and can be performed as outlined below.

[0125] In figure 9 one can observe a normalized V/A ordinate axis a minimum level of 0% and a maximum level of 100 % (the level of V/A normalised to 100 % could e.g. be linked to a fixed level of 0.020 m 3 /m 2 ) . When introducing the linguistic variables VL (Very Low) , L (Low) , M (Medium) , H (High) and VH (Very High) to categorize levels of (reversible or irreversible) fouling, it is possible to also introduce four normalized V/A levels. By way of example, these can be selected at respectively 12.5 %, 25 %, 50 % and 100 %. As shown in figure 9, at each level distinct VL, L, M, H and VH fouling zones can be identified.

[0126] Appropriate fuzzy sets then can be defined for all VL, L, M, H and VH fouling zones at each level of 12.5 %, 25 %, 50 % and 100 %. This is illustrated for the 50% level in figure 10. [0127] A reversible fouling fingerprint graph

(curve) typically crosses the four levels 12.5 %, 25 %, 50

% and 100 % in distinct VL, L, M, H and VH fouling zones.

Such cross-points, which can be labelled with the symbols

VA12.5, VA25, VA50 and VAlOO respectively, can be used as input variables to a fuzzy control block. In order to be able to execute a logical inference and thereafter a defuzzification into an output as a normalized value between 0 (no fouling propensity) and 100 (extreme fouling propensity) , fuzzy IF THEN rules can be set up for possible combinations of the VL, L, M, H and VH fouling zones at the four levels 12.5 %, 25 %, 50 % and 100 %. A number of these rules are given in table 3.

[0128] It should be noted that a fuzzy inference and defuzzification does not produce a discontinuous, stepwise

output value but a continuously changing value with a varying position or shape of the considered fouling graph

(curve) . An increasing reversible/irreversible fouling propensity of a feed will thus show gradually declining curves (graphs) according to figure 6 or 7, from a more

"vertical" to a more "horizontal" position, in which the fouling layer has an extremely low permeability. In the latter case, the aeration definitely needs to be fully throttled in order to continuously remove said layer. [0129] The linguistic variables from the fuzzification module 30 are processed in the inference module 31. The inference module 31 uses the FSL conditional rules of the fuzzy rules module 32, which are constructed from expert knowledge of e.g. a human operator of a membrane filtration plant. The fuzzy rules comprise the fuzzified values of one or more of the inputs 141, 142 and 17 and combine those through fuzzy inference into a linguistic fuzzy output variable. Each linguistic fuzzy output variable is linked to a set-point parameter 121, 122, 123, etc. The inference result is thus a linguistic value for the corresponding output control variable. Each SCS output parameter is linked to a linguistic variable. [0130] A set of fuzzy rules within the fuzzy rules module 32 comprises also a multitude of fuzzy IF... THEN rules. Moreover, multiple input variables may be processed by such fuzzy rules as fuzzy operators AND, OR, etc. can be integrated.

[0131] During the fuzzy inference handling of the fuzzy rules two major actions are executed. First the IF parts of the rules are calculated (which is called aggregation) whereafter the THEN parts of the rules are calculated (which is called composition) . The calculations result in a linguistic value of the output variable (parameter) . It should be noted that the classic fuzzy

operators AND, OR, etc. are not related in any way to the classic Boolean logical operators AND, OR, etc. The classic Boolean operators in effect can not handle conditions which are "more-or-less" true. Therefore specific fuzzy operators such as AND and OR have been defined mathematically in fuzzy set handling.

[0132] Each set-point parameter' s linguistic value is subsequently translated into a real, physical (crisp) value by the defuzzification module 33. This value is needed to actually adjust the value of the set-point parameter (e.g. 121) of the lower-level classic control system.

[0133] For each set-point parameter of the lower- level MFS-control, a separate and individual Fuzzy Set Logic control block 131 can be needed. In general, each Fuzzy Set Logic control block 131, 132, 133, etc. shows the same features but each block is tuned to the specific set- point parameter linked to that block. [0134] Different defuzzification methods can be deployed in FSL control but often the Centre-of-Gravity method is used. Such methods result in a crisp output parameter value which is even not discontinuous with changing input parameter values. Such defuzzification clearly relates to the human operator mimicking property of a fuzzy control approach.

[0135] It can be stated that a FSL-based MFS-control action, being the setting of the value of a set-point parameter, is produced as a result of the sensing (= input) of e.g. reversible or irreversible fouling and other parameters (mixed liquor temperature, etc.) while checking for the appropriate control within the knowledge base

(fuzzy rules) linked to those inputs and output. This resembles a human operator' s action without the need of mathematical models.

[0136] An EPS concentration sensor may provide an input 17 to a supervisory control system 13 and is preferably used for the control of dosing coagulants in order to prevent (anticipate) membrane fouling, instead of remedying an already occurred irreversible membrane fouling by chemical cleaning (MC) . Heuristic knowledge about the effect of the EPS concentration value on the dosing of coagulants can be implemented in a specific control block of the type 131. [0137] Table 1 gives an overview of the set-point parameters discussed above and related to the mechanical and chemical cleaning actions. In the particular case of an FSL-based supervisory control system, the supervisory control of each set-point parameter of table 1 would imply a separate control block of the type 131. However, not all set-point parameters need to be supervisory controlled. The SCS 13 of the invention may restrict its supervisory control to only a subset of the set-point parameters listed in table 1. Additional set-point parameters may equally be identified and supervisory controlled.

Description of a Preferred Embodiment of the Invention [0138] As part of a preferred embodiment, the supervisory control of the aeration (set point parameters 7-9 in table 1: mechanical cleaning of the membrane surface by air bubble scouring) is described according to an FSL implementation and only in general since different membrane filtration configurations exist (flat, capillary, tubular) and the aeration configuration therefore can also differ in existing MFS layouts. In this embodiment there are three linguistic output variables AirBubFlow, AirBublnterval and AirBubDuration .

[0139] A fouling sensor, such as the fouling measurement device of the invention is used in this example

to measure the reversible fouling propensity of the mixed liquor and thus produces the main input parameter value of the supervisory control system 13, namely a reversible fouling parameter. Next to the reversible fouling parameter, temperature is also considered in this example as an SCS input parameter. Additional input parameters may be introduced as well but in this example, the number of input parameters is confined to two, for simplification reasons. For the same reasons, in this example, only AirBubFlow as an output variable is discussed. The linguistic name for the reversible fouling input variable is "Frev" and for the mixed liquor temperature the linguistic name is "Temp". It is assumed in the present example that the definition of five appropriate fuzzy sets for Frev and also five fuzzy sets for Temp is adequate. The linguistic names of the fuzzy sets for Frev are VeryLow, Low, Medium, High and VeryHigh. The linguistic names of the fuzzy sets for Temp are VeryCold, Cold, Moderate, Warm and VeryWarm. [0140] With regard to the implementation of the outputs of measurement device 40 as inputs 141 and 142 to a supervisory control system of the invention, a membership of Frev to its fuzzy sets can be implemented by fuzzifying the curve of figure 6 as a whole. This can be done by choosing multiple R to t,rev/R m x-axis values and impose on each of these the specified fuzzy sets.

[0141] It is also assumed in this example that seven fuzzy sets are defined for the linguistic variable AirBubFlow and their fuzzy linguistic names are VeryLow, ExtraLow, Low, Medium, High, ExtraHigh, VeryHigh. It is then in principle possible to produce for each possible combination of both input variables Frev and Temp a fuzzy rule and a corresponding linguistic value for the output linguistic variable AirBubFlow.

[0142] The quantitative definition of the fuzzy sets of Frev, Temp and AirBubFlow is not presented here since those can be different for each MFS configuration (membrane type, mixed liquor characteristics, fouling measurement method or protocol, etc.) . However, in a qualitative way, the theoretical combinations can be written as shown in table 2. In this example the linguistic name of the fuzzy sets of the output variable AirBubFlow in the fuzzy rules (table 2) is also not indicated (see the format "AirBubFlow = ...") since its output linguistic value also depends from the actual MFS configuration.

[0143] An additional input parameter with also five fuzzy sets would induce a total of 125 possible fuzzy rules when based on the use of only the AND fuzzy logical operator. When using other fuzzy logical operators (OR, etc.) the number can be even higher when introducing additional input parameters. The large number of all theoretical combinations (thus possible fuzzy rules) can in practice be reduced to a much lower number, while preserving all the (human operator) heuristic knowledge of the specific control of AirBubFlow to ensure its efficient supervisory control. In this example, the thus selected fuzzy rules from table 2 would then be implemented in the fuzzy rules module of an FSL control block 131, arranged to manage the calculation of the set-point parameter value for AirbubFlow.

[0144] While the foregoing description and drawings represent the preferred embodiments of the present invention, it will be obvious to those skilled in the art that various changes and modifications may be made therein without departing from the spirit and scope of the present invention .

Table 1: basic overview of output variables that may be controlled by a SCS (set-point parameters)

Table 2: Fuzzy rules for AirbubFlow

# Fuzzy rules for the control of AirbubFlow

IF Frev=VeryLow AND Temp=VeryCold THEN AirBubFlow=

IF Frev=Low AND Temp=VeryCold THEN AirBubFlow= ...

IF Frev=Medium AND Temp=VeryCold THEN AirBubFlow=

IF Frev=High AND Temp=VeryCold THEN AirBubFlow= ...

IF Frev=VeryHigh AND Temp=VeryCold THEN AirBubFlow=

IF Frev=VeryLow AND Temp=Cold THEN AirBubFlow= ...

IF Frev=Low AND Temp=Cold THEN AirBubFlow= ...

IF Frev=Medium AND Temp=Cold THEN AirBubFlow=

IF Frev=High AND Temp=Cold THEN AirBubFlow= ...

10 IF Frev=VeryHigh AND Temp=Cold THEN AirBubFlow=

11 IF Frev=VeryLow AND Temp=Moderate THEN AirBubFlow=

12 IF Frev=Low AND Temp=Moderate THEN AirBubFlow=

13 IF Frev=Medium AND Temp=Moderate THEN AirBubFlow=

14 IF Frev=High AND Temp=Moderate THEN AirBubFlow=

15 IF Frev=VeryHigh AND Temp=Moderate THEN AirBubFlow=

16 IF Frev=VeryLow AND Temp=Warm THEN AirBubFlow=

17 IF Frev=Low AND Temp=Warm THEN AirBubFlow=

18 IF Frev=Medium AND Temp=Warm THEN AirBubFlow=

1 9 IF Frev=High AND Temp=Warm THEN AirBubFlow=

20 IF Frev=VeryHigh AND Temp=Warm THEN AirBubFlow=

21 IF Frev=VeryLow AND Temp=VeryWarm THEN AirBubFlow=

22 IF Frev=Low AND Temp=VeryWarm THEN AirBubFlow=

23 IF Frev=Medium AND Temp=VeryWarm THEN AirBubFlow=

24 IF Frev=High AND Temp=VeryWarm THEN AirBubFlow= ...

25 IF Frev=VeryHigh AND Temp=VeryWarm THEN AirBubFlow=

Table 3 : Fuz zy inference rules for the revers ible foul ing propens ity f ingerprint .