Filtration

Refinery Process Stream Purification Refinery Process Catalysts Troubleshooting Refinery Process Catalyst Start-Up / Shutdown Activation Reduction In-situ Ex-situ Sulfiding Specializing in Refinery Process Catalyst Performance Evaluation Heat & Mass Balance Analysis Catalyst Remaining Life Determination Catalyst Deactivation Assessment Catalyst Performance Characterization Refining & Gas Processing & Petrochemical Industries Catalysts / Process Technology - Hydrogen Catalysts / Process Technology – Ammonia Catalyst Process Technology - Methanol Catalysts / process Technology – Petrochemicals Specializing in the Development & Commercialization of New Technology in the Refining & Petrochemical Industries

Web Site: www.GBHEnterprises.com

GBH Enterprises, Ltd.

Process Engineering Guide: GBHE-PEG-SPG-300

FILTRATION Information contained in this publication or as otherwise supplied to Users is believed to be accurate and correct at time of going to press, and is given in good faith, but it is for the User to satisfy itself of the suitability of the information for its own particular purpose. GBHE gives no warranty as to the fitness of this information for any particular purpose and any implied warranty or condition (statutory or otherwise) is excluded except to the extent that exclusion is prevented by law. GBHE accepts no liability for loss or damage resulting from reliance on this information. Freedom under Patent, Copyright and Designs cannot be assumed.



Process Engineering Guide: FILTRATION CONTENTS 0 INTRODUCTION 1 THE THEORY UNDERLYING FILTRATION PROCESSES 1.1 The Mechanism of Simple Filtration Systems

1.1.2 Cake Filtration 1.1.3 Complete Blocking

1.1.4 Standard Blocking 1.1.5 Intermediate Blocking

1.2 Cake Filtration – Models and Mechanisms

1.2.1 Classical Theory for the Permeability of Porous Cakes and Beds 1.2.2 The Rate of Filtration through a Compressible Cake – The

Standard Filtration Equation 1.2.3 The Compression or Consolidation of Filter Cakes – Ultimate

degree of dewatering 1.2.4 The Rate of Consolidation 1.2.5 Useful Semi-Empirical Relations for Constant Pressure and

Constant Rate Cake Filtration 1.2.6 Constant Pressure Filtration 1.2.7 Constant Rate Filtration 1.2.8 Multiphase Theory of Filtration

1.3 Crossflow Filtration 2 THE RANGE AND SELECTION OF FILTRATION EQUIPMENT

TECHNOLOGY

2.1 Scale 2.2 Solids Recovery, Liquids Clarification or Feed stream

Concentration 2.3 Rate of Sedimentation 2.4 Rate of Cake Formation and Drainage 2.5 Batch vs Continuous Operation 2.6 Solids Loading



2.7 Further Processing 2.8 Aseptic or “Hygienic” Operation 2.9 Miscellaneous 2.10 Shear versus Compressional Deformation 2.11 Pressure versus Vacuum

3 SUSPENSION CONDITIONING PRIOR TO FILTRATION

3.1 Simple Filtration Aids 3.2 Mechanical Treatments 4 POST-FILTRATION TREATMENTS AND FURTHER DOWNSTREAM

PROCESSING

4.1 Washing 4.1.1 Air-Blowing 4.1.2 Drying

5 TESTING AND CHARACTERIZATION OF SUSPENSIONS

5.1 Introduction – Suspension 5.2 Properties relevant to Filtration Performance

5.2.1 Pre-Filtration Properties of Suspension 5.2.2 Properties of Filter Cake 5.2.3 Laboratory Scale Filtration Rigs

5.3 Means of Monitoring Flocculant Dosage

5.4 Filter Cake Testing

5.4.1 Strength Testing (See also piston press described earlier) 5.4.2 Cake Permeability or Resistance 5.4.3 Rate of Cake Formation



6 EXAMPLES OF THE APPLICATION OF THE FORGOING PRINCIPLES

6.1 Dewatering of Calcium Carbonate Slurries 6.2 Dewatering of Organic Products – Procion Dyestuffs

6.3 Filtration of Biological Systems – Harvesting a Filamentous Organism

REFERENCES TABLES FIGURES



0 INTRODUCTION

For the purposes of this process engineering guide, filtration will be regarded as the process whereby solids are separated from liquids by the use of a porous medium. This definition then deliberately excludes the filtration of gas streams. Filtration together with gravity separation forms the basis for nearly all unit operations to accomplish the dewatering of suspensions [3-5]. It is therefore worthwhile considering the broad factors that favor filtration over gravity separation methods for a given system [1]. Perhaps the most important of these involves the use of a porous medium to effect the operation. The nature of the former may be tailored and designed to best suit the requirements of both phases of the suspension and the sort of dewatering action required. In contrast, all techniques based on gravity separation are completely dependent upon the density difference, ∆ρ, between solid and liquid phases. Since ∆ρ must be regarded for many systems as an invariant, (It may be slightly perturbed by a change in operating temperature), a small value for the quantity almost Invariably means that gravity separation will prove difficult. This Is often the case for biological particles (see Section 3.8). For such cases filtration is then often to be preferred. The penalty that has to be paid for this versatility of filtration process design is usually greater expense and additional complexity when continuous or automated operation is desired. It must be stressed that the above statements are based on broad, generalized principles. For both filtration and gravity separation, the ingenuity of solid/liquid separation equipment designers has led to means of at least partly circumventing many of the disadvantages associated with each [8].

The above definition of filtration encompasses a large number of possible operations ranging from the clarification (or even sterilization> of a very slightly loaded suspension to the removal of a product in the form of a solid cake. Other variations permit filtration to be used as a thickening operation where a feed stream is concentrated in the suspended phase without the formation of a cake or the deposition of solids. Since this section of the manual falls within the dewatering section, most emphasis will be given to processes where it is the suspended phase that contains the desired product and Is usually to be further processed. Clarification by filtration will be discussed briefly but, for greater detail, the reader is referred to the separate chapter on that subject (GBHE-PEG-SPG-400 - Centrifugation) and to references [1,12]



Finally It is appropriate to set the context for the rest of this guide. In Section 3.5.2 some of the basic, and largely classical, theory of filtration will be presented. The range and selection of filtration equipment Is then briefly discussed in Section 3.5.3. In this and all parts of this section emphasis will be placed much more upon relating material properties to the process design than in providing a comprehensive survey of the available technology. To Illustrate process interactions, In Section 3.5.4 a brief consideration is given to those operations most likely to follow filtration in a complete process. In Section 3.5.5, methods for testing and characterizing the filtration properties of suspensions are described together with the interpretation of the results in terms of the theory previously given. To conclude the section, examples are given of the processing of suspensions containing inorganic, organic and “biological” particles. It is hoped that these will illustrate many of the principles previously developed.

1 THE THEORY UNDERLYING FILTRATION PROCESSES The purpose of this part is to provide the necessary theory on which the rest of the section is based. It is intended that each of the following topics should be self-contained and can therefore be read in isolation. The theory pertaining to washing and dewatering by air-blowing is postponed until Section 3.5.4.

1.1 The Mechanism of Simple Filtration Systems

Various authors have proposed schemes for classifying the diverse mechanisms that may operate during filtration operations. The simplest classification for solids retaining systems discriminates between cases where the solids build up on top of a cake and those where they are retained within the filter medium. A useful classification based on this approach was provided a long time ago by Hermans and Bredee [14] who distinguished four mechanisms, which are sketched schematically in Figure 1:



1.1.2 Cake Filtration (Figure 1(a))

This process involves the removal of solids by the formation of a filtered “cake” on the surface of the medium. It Is the cake itself which effects the subsequent filtration. Depending upon whether the suspended particles are smaller or larger than the pores of the medium, this process Is usually preceded by a bridging or straining process in order for cake formation to ensue. Hermans and Bredee proposed the following equations to describe the time dependence for cake filtration:

where V is the volume of filtrate at time t, Qo. the Initial flow rate

and k an empirical constant. Alternatively in terms of a mean (cumulative) flow rate, q(t):

Equations (1) and (2) clearly represent a gross oversimplification of the cake filtration of real systems. Modified forms of these are presented In subsequent sections. Graphical representations of Equations (l)-(2) are shown In Figure 2(a).

1.1.3 Complete Blocking (Figure 1(b)) This mechanism Involves a straining process either at the medium surface or within Its internal structure. This implies that the solids particles are larger than the. local size of the pores of the medium. This then results in completing blocking of pores as the filtration proceeds with a linear decrease in the flow rate with volume:



In surveying a large number of suspensions and filter media, Hermans and Bredee found this type of filtration to be rather rare. It also falls to yield a useful and physically plausible volume-time relationship when integrated. The concept of "complete blocking" as depicted in Figure l(b) still retains some merit for classification purposes, however.

1.1.4 Standard Blocking (Figure l(c))

This is the mechanism most pertinent to depth filtration where particles may pass through the pores of the medium but are retained by eventual adhesion to it. Hermans and Bredee's model ascribed a "fouling" process where the internal volume of the pores decreased linearly with V.

Thus they obtained equation (4).

Since the subject of deep bed filtrations falls outside the intended scope of this guide, no further discussion of this mechanism or topic will be presented. More Information may be found in the guide on clarification or In the books by Svarovsky (Chapter 11 of [1] and Purchas (Chapters 3 and 6 of [12]. In addition, the role of the particle zero-potential has recently been considered by Raistrick amongst others (J H Raistrick in [94]).

The equations (l)-(4) were developed to allow volume, flow rate and time correlations to be tested for each mechanism under conditions of constant pressure filtration. A simpler diagnostic means of distinguishing and interpreting them was provided by the dependence of the rate of change of total filtration resistance, r, (i.e. medium plus accumulated solids) with filtration volume, dr/dV, with r. The three mechanisms described so far were attributed the following dependence:



In order to encompass the sometimes observed relationship, dr/dV a r, Hermans and Bredee's classification allowed for a fourth mechanism:

1.1.5 Intermediate Blocking

Where,

This mechanism may be physically viewed as being intermediate between Figures l(b) and l(c).

1.2 Cake Filtration – Models and Mechanisms

In this section of the dewatering chapter, greater prominence will be given to cake filtration than to the other mechanisms just described. In part this reflects the frequency with which the formation and properties of a cake dominates a separation operation. However, the other reason for this emphasis derives from the necessity to understand and control the influence of the suspension itself as opposed to the hardware; for a properly conceived cake filtration the suspension properties are dominant and the filtration medium of secondary importance.

Before presenting some of the basic theory for cake filtration it is important to delineate the factors which relate to the fundamentals of dewatering presented In Section 3.2. In general for a given cake filtration system one might want to ask the following two questions:

(i) What degree of dewatering Is attainable for a given suspension In a

cake filtration rig? How is this ultimate dewaterabillty influenced by changes in the suspension, filtration conditions and driving pressure?

(ii) What are the kinetics of the filtration process, ie do they allow the

process to operate near to or at the ultimate limit as in (1) above. In general theories describing the permeability of a cake or its rate of deposition are very much aimed at addressing the second sort of question.



However, when one is trying to understand, for example, the expression of liquid in a filter press, it is the ultimate dewaterability that is usually critical though kinetics may again be limiting. For a comprehensive understanding of filtration processes It is therefore necessary to have a quantitative picture of both the kinetics of the process and the maximum degree of dewatering that can be obtained. Both these Issues will be tackled In the following pieces of theory. Additional aspects of the theory of filter cakes and sediments have been discussed In more detail by Tiller and other workers [11]. Other relevant references may be found in the sedimentation section of this chapter (3.3).

1.2.1 Classical Theory for the Permeability of Porous Cakes and

Beds

From an observation of the rate of flow of liquids through beds of sand, Darcy [15] suggested an empirical correlation between the fluid velocity, u, the pressure drop across the bed, ∆P, and the bed thickness, L. The result, Darcy's Law, may be expressed in the simple form:

Some insight into the nature of the constant, K1, is gained by assuming a result from the Poiseuille [16] Equation for the flow of a liquid through a capillary tube of radius, r, and length, L:

This result enables equation (6) to be modified thus:

where the Influence of the fluid viscosity, Q, has been explicitly Included. K2 has dimensions of (length) 2.



A combination of the equations of Darcy and Polseuille together with the Incorporation of two properties of the bed itself (SA and Ɛ, led to the equation attributed to Kozeny and Carman [17-18]:

In this equation the bed properties SA and Ɛ represent the specific surface area (m-1) and the fractional voidage, ie 1 - ɸ, where ɸ is the dimensionless volume fraction of particles In the bed or filter cake.

The Kozeny-Carman Equation, although the precursor for the most commonly used filtration equations, suffers from a number of restrictions and oversimplifications. Basically It Is a sound description for the drainage rate of viscous flow of a clear liquid through a porous bed of constant permeability VULCAN VGP systems the permeability at a given point may be a function of pressure drop, time and the height of that point within the bed. Further discussion of these features is provided later. The constraint of viscous laminar flow is also sometimes not strictly applicable In practice. Where turbulence becomes significant a correction to the rearranged equation may be made as follows, after Burke and Plummer [20]:

This equation was derived for beds of uniform spherical particles of diameter, d; hence SA = 6/d. It can be seen that the leading term for the pressure drop per unit length is the simple, viscous Kozeny-Carman contribution. The second term is the kinetics energy loss to the pressure drop through turbulent flow. Whether such a kinetic energy modification is necessary for a given bed and flow rate may be judged by a plot of the Reynolds Number,



against permeability (eg Morgan's work with sintered metal pores [21]. Fortunately it Is often the case that the simpler, purely viscous treatment of permeability Is sufficiently accurate for practical applications.

1.2.2 The Rate of Filtration through a Compressible Cake – The Standard Filtration Equation

From the somewhat idealized equations for the permeability of porous beds, a straightforward modification, to allow for medium resistance, Rm, yields a general expression for cake filtration rates:

Rc, the cake resistance (units of m-l) may In turn be related to, w, the weight of solids per unit volume of filtrate (kg rnm3) and a quantity, r, the specific resistance of the cake (le resistance/weight of solids per unit area or m kg-1):

As pointed out in (i) for many real systems it is necessary to take account of the finite compressibility of the filter cake under an applied pressure. A simple empirical correction to resistance has been very widely used (1,12).



where ro is the specific resistance at zero pressure drop and the exponent “s” is known as the compressibility factor. This factor ranges from 0 for a perfectly incompressible material to unity a highly compressive cake. Thus a general expression for the rate of cake filtration may be written as:

Equation (15) may be Integrated to yield the total filtration volume after a specified time provided that the functional dependence of the flow rate or the pressure drop with time is known. It is, however, once again very important to note the simplifications and approximations that are Inherent In this equation.

Firstly, It is very common to assume that the medium resistance, Rm, is a constant in time. Physically this corresponds to no blinding or trapping of solids within the medium, ie a complete absence of Hermans and Bredee's mechanisms (ii), (iii) and (iv) of Section 3.5.2(a). In reality the medium resistance is quite likely to change during the initial stages of cake formation. Once a cake of any substantial thickness has been formed, the cake Itself will largely prevent any further particulate matter from reaching the filter medium or support. Hence Rm will usually thereafter indeed be a constant unless further solids are leached into the medium from the underside of the cake. Thus it is often reasonable to treat Rm as a constant during the cake filtration but it may prove erroneous to deduce its value from a measurement of the resistance to "clean" suspension medium alone.

(For an alternative perspective to this Issue, see (b)(v).)

The specific cake resistance, r or r0 will not be a simple constant for a given suspended phase and medium. It will depend critically upon the colloidal properties of the suspension and the consequent structure of the filter cake. It may also depend upon the mode and rate of cake lay-down.



Factors such as particle size and distribution, particle shape, particle surface properties, the presence and nature of any flocculating agent, etc, etc, will all strongly influence r. Thus there arises the ability to control the filtration process by conditioning the suspension by the use of pretreatments and filtration aids. These will be discussed in more detail later. One further reservation concerning equations (14) and (15) must be expressed. It should not be presumed that the simple power law dependence of r on the pressure drop ΔP will apply over a very extended range of pressures.

1.2.3 The Compression or Consolidation of Filter Cakes – Ultimate degree of dewatering

Many of the physical principles pertaining to the consolidation process have already been derived in previous sections of the manual (3.2 - 3.4). The subject is sufficiently central to many filtration situations, however, that an outline of the theory will be reproduced here. Consolidation of a structured filter cake will occur during its laydown and the filtration process proper. Additionally it may be exploited following cessation of the actual filtration by the application of pressure to express liquid from the pores of the cake. This latter may involve cake collapse, that is consolidation, or displacement by gas. Pneumatic dewatering is briefly considered in a later section (3.5.5(b)).

The ultimate attainable degree of dewatering of a filter cake through consolidation is calculated by considering the two opposing forces on the cake. On the one hand, above a certain solids content the cake will possess a structural resistance to densification which may be quantified In terms of Its uniaxial, compressional yield point, Py(ɸ). Methods for measuring Py(ɸ) and a description of its application are given later. This Internal resistance to densification operates against the externally applied pressure differential across the filter cake, ΔP.

Consider a filter cake of solids finit undergoing constant pressure consolidation. Initially, provided that Py(ɸ init) is less than the applied consolidating pressure, ΔP, the cake will be compressed and liquid expelled from it. As this process continues, the concentration of solids in the cake, ɸ, and hence the function, Py(ɸ), will increase.



The compressional yield point, Py(ɸ), is in fact a very strong function of solids content (see Examples 3.5.7) and at some later time it will be of sufficient magnitude to match the applied pressure and consolidation will cease. This defines that the ultimate degree of dewatering will occur when:

when this condition Is reached the internal stresses of the cohesive cake are large enough in magnitude to fully resist the applied pressure ΔP. Thus predicting the ultimte dewaterability simply requires a knowledge of the function Py($) and this may be measured by a simple laboratory scale determination. Examples of this procedure are given later and in Section 3.2.6 of the manual.

1.2.4 The Rate of Consolidation

Equation (16) enables the easy estimation of an equilibrium degree of dewatering for a given filter cake and consolidating pressure. The question of how fast that ultimate solids content is attained is a more complicated one involving additional physical factors such as the drag forces exhibited by the consolidating network on the liquid being expressed from the cake. A full analytical description of the kinetics of such processes are not presently available.

The Consolidation Model of Buscall and White, based on the Yield Stress (Py) concept applied to sedimentation, has already been discussed in some detail in Sections 3.2 and 3.3. This model automatically incorporates the ultimate dewatering limit of Equation (16) for consolidation. This follows from the choice of constitutive equation relating the time evolution of the concentration of solids in the cake (the substantive derivative) in terms of the yield stress parameter:



A number of attempts have been made to couple equation (17) with continuity equations in order to derive scaling relationships for the physically distinct problem of filtration (cf the results for sedimentation, Sections 3.2 and 3.3 of this chapter). Possible strategies are outlined in references [25-273]. Although such an approach has not yet proved entirely successful a number of comments may be made:

(1) As for sedimentation the equation (17) encompasses the notion that the driving force for solid/liquid separation is attenuated by the elastic stress in the cake as described by Py(ɸ).

(2) Thus at low driving pressures (ΔP), Py(ɸ)) the rate of cake consolidation may be enhanced by manipulation of those factors (Sections 3.2.6 and 3.3) that reduce Py(ɸ). Strategies for suspension conditioning may utilize this type of reasoning and are considered later in Sections 3.5.3 and 3.5.4.

(3) In contrast at relatively higher driving pressures (ΔP >> Py(ɸ)), the main factors controlling consolidation will involve properties of the primary particles such as drag coefficients, together with dynamic drag coefficients for the network (λ(ɸ) in Equation (17)). The pore structure and cake permeability will therefore be relevant and hence In this case the controlling factors are similar to those affecting the specific cake resistance as discussed earlier (3.5.2(b)).

Until such a time as a full analysis of filtration in terms of the yield stress concept has proved possible, the kinetics of consolidation of filter cakes will remain a largely experimental science with understanding being at best qualitative. Possible experimental approaches to the problem are given In Section 3.5.6 and examples discussed in Section 3.5.7.



1.2.5 Useful Semi-Empirical Relations for Constant Pressure and Constant Rate Cake Filtration

Constant Pressure Filtration [1,10,12] As a starting point, the general cake filtration equation, (15), is rearranged in the following form:

Integration of this simple form of the equation allows the relationships between filtrate volume, time and Instantaneous filtration rate to be deduced. The results are:



Both the specific resistance and the medium resistance can be evaluated from the general equation, (18), via a plot of reciprocal rate, dt/dV, as a function of cumulative filtrate volume, V. Such a plot has a slope equal to the first bracketted term In the equation from which r o may be evaluated (at a given pressure); the intercept yields the medium resistance, Rm. It Is quite common with cake filtrations for the extrapolated data to pass through the origin such that the medium resistance Is negligible compared with that of the cake. These simple principles are Illustrated schematically in Figure 3. Likewise, for compressible cakes a series of reciprocal rate versus cumulative volume at various pressures yields the relationship between ΔP and specific resistance from which the exponent, S b of equation (14) can be evaluated (see Figure 3).

Once the specific and medium resistances are known, from laboratory (or plant) measurements of V-l versus V, the relations (19)-(22), and those that follow for constant rate filtrations, may be applied In a predictive fashion. It is, however, Important to recall the predictions and restrictions that apply to equation (18) and were discussed in Section 3.5.2(b) (II). The two most important caveats in this context involve the scaling up of the quantities r o and Rm. The medium resistance Rm may be an important parameter and may not hold the same value at plant-scale as measured in the laboratory. Likewise, r o depends critically upon the mode of cake formation and so also may vary with scale, Initial filtration rate.

1.2.7 Constant Rate Filtration

For a constant rate filtration, the pressure drop will increase as the cake builds up. In practical situations there will be a limit to the magnitude of ΔP that can be applied or tolerated. Hence it is necessary to know the cumulative volume, V*, and time, t*, associated with a given limiting pressure drop, ΔP*. The volume- time relationship is trivial: cumulative volume is given by the product of the time and constant rate.

The other relations are again derived simply from equation (15):



The equations (19)-(23) have been derived for strict conditions of constant pressure or rate. In reality many dewatering configurations operate in regimes which are intermediate between these two whereupon a more involved numerical integration of equation (15) will be required in order to derive volume-time relationships. A common configuration involving both extreme cases utilizes constant rate filtration until the pressure drop has reached Its maximum attainable (or tolerable) value whereupon the filtration continues at constant pressure until the rate falls to an unacceptable level.

1.2.8 Multiphase Theory of Filtration

The theory and equations that have been presented in outline here, have long been accepted as a reliable though somewhat empirical description of the cake filtration process. However, more recently (- 1975 onwards), various workers have re-examined this so-called "two resistance* approach (ie r and ), and contrasted its basis with an alternative description, the "multiphase filtration theory" [22,23].

The latter Involves a lengthy derivation of equations based on a continuum mechanics approach. This detail will not be presented here but may be found In the references. Rather an outline of some of the results will be discussed and compared with the "two resistance" formulation.

It is first worthwhile recapitulating on the Interpretation of data analyzed via the *two resistance" approach. In essence this method attributes the resistance to filtration to two additive contributions: that of the cake which may be a function of time and pressure etc and that of the septum or medium which is rigorously regarded as constant (and often negligible). For this case a linear relationship between the inverse filtration rate V-l and cumulative filtrate volume, V, Is taken to imply the following:

(1) The local porosity and cake resistance are uniform.

(2) The average porosity and cake resistance are constant.



Where non-linear reciprocal rate data is encountered, the following conclusions are inferred from the apparent compressibility of the filter cake:

(1) lion-uniformity of local porosity and cake resistance.

(2) Time and other dependence for average porosity and cake

resistance.

(3) Dependence of cake resistance on slurry concentration, pressure and septum.

In essence the “multiphase” description is based upon local continuity and motion equations for both suspending and particulate phases In the cake and septum. Application of dimensional arguments together with estimates of the magnitude of the relevant groupings indicate which terms dominate V-l. Briefly it is the pressure driven and drag forces that control the process whilst inertial and viscous forces may be largely neglected. At the end of the analysis the following multiphase cake filtration equation is gained:

where G Is a function of slurry concentration, solid and liquid densities, the filter area (A), and the average porosity. For the case where the cake height is a linear function of filtrate volume, V, G can be shown to be Independent of V. Then the reciprocal rate equation depends upon the three quantities:

K o (V) :- Septum permeability J o (V) :- Septum pressure gradient Po (V) :- Cake pressure drop

and now deviation from a linear relationship between V-l and V for constant Po, are attributed to changes in the permeability and pressure gradient developed In the septum and its Interface with the filter cake. Willis et al have demonstrated that this theory is plausible by showing that a given cake (formed by the filtration of a Lucite slurry> can be made to undergo a transition from apparent “incompressible” to “compressible” filtration merely by changing the nature of the septum.



In fact, Willis and co-workers cited only one experimental verification of their theory and so It Is unwise to speculate on Its generality. Perhaps the most important feature of the analysis Is that the attribution of non- linear reciprocal rate data to “compressible” cake formation without further corroborating evidence may prove fallacious if septum fouling, blinding or compression Is occurring. Means of obtaining this corroborative evidence and of quantifying cake compressibility without a septum present are considered later in Section 3.5.6.

1.3 Crossflow Filtration [8 -44]

Although a detailed discussion of the various applications and some variants of crossflow filtration will be provided in Section 3.9 of this manual (“Pressure Driven Membrane Separation Processes”), a brief outline of the principles will be given here, The basic objective of a crossflow configuration is to achieve the limited dewatering of a suspension (or solution> using a permeable membrane but without the retention or immobilization of the solid phase. This is achieved by maintaining a tangential flow of the suspension across the surface of the membrane. Thus, Ideally, cake formation is totally avoided and the particulate phase remains evenly distributed throughout the concentrating suspension as a result of the convective effects of the flow [28,35,391.

At a simple level the first requirement for an understanding of crossflow dewatering would be a model relating the permeate flux, ie the rate of concentration to the primary process variables: temperature, driving pressure, starting concentration and transmembrane velocity. For a perfectly ideal case, the flux relationship could be simply derived from treating the septum as a non-blocking bed of constant permeability. The flux would then be directly proportional to the driving pressure gradient as predicted by the Darcy or Kozeny-Carman relationships, equations (6)- (9), In all real cases the behavior of the flux is by no means that simple. An alternative approach that has been applied, also largely unsuccessfully, Is to use some sort of modified cake filtration model. In reality the factors that control the flux of permeate are various, subtle and almost Inevitably time dependent and hence It Is no surprise that simple models are Inappropriate. some useful generalizations may, however, be made.



In many real examples of crossflow filtration, particularly In ultrafiltration [31-37], the flux rate Is limited by a mechanism called concentration polarization. This arises because the layers of suspension closest to the membrane are those which suffer depletion from permeate. There Is therefore a local increase In concentration (but not necessarily a “cake”) which inevitably leads to a fall in the permeate flux.

Acting In opposition to this mechanism are the effects of diffusion and laminar or turbulent convective flows. Clearly the diffusion process Is very dependent on the size and nature of the particulate (or dissolved) phase. Some useful relationships have been given to correlate the effects of the various factors that may operate when this mechanism is dominant. The following have been used for dewatering of biological suspensions by ultrafiltration [32-33]:

It must be stressed that equations such as (25) and (26) are by no means applicable to all crossflow situations.

In addition to the above problems Involving particle concentration gradient, another common source of permeate flux decline is the phenomenon referred to as "fouling" [35-36].

Fouling encompasses a whole series of processes whereby permeate flux falls as a function of time as a result of changes in the membrane Itself. Commonly these changes might Involve deposition of material on the surface or interior of the membrane often leading to a time-dependent decrease in Its porosity. Beyond these generalizations lies an enormous number of experimental studies and observations but unfortunately there Is as yet only relatively poor understanding of the phenomenon and how



to avoid It. Current active research Is being carried out in various centers in the world with notable British contributions being made at Bath 1351 and at Varren Spring Laboratory - the latter under the auspices of a BIOSEP project [39-41]. This project has utilized various electron microscopy based investigations to probe the whereabouts of protein foulants, identified by staining, during biological separations.

In the future, this sort of experiments should at least assist in the elucidation of the mechanisms of fouling In various cases thereby enabling the application of collold and surface science to avoid such problems. Another engineering based strategy for reducing fouling, the deliberate production of transmembrane, turbulent vortices, has recently been investigated by Hltchell I951.

Again with respect to potential future developments, it is interesting to note that developments are being made in new filtration-based dewatering strategies involving the use of electric fields [38,40]. Examples of these process operations Include electrophoresls, electrodecantation, electroflltratlon, electro-osmosis and others.

Some of these will be discussed in more detail in Section 3.9. However, a particularly worthwhile technology target In the present context involves the concept of harnessing a dielectrophoretic effect to prevent fouling or concentration polarization during crossflow filtration processes [40].



2 The Range and Selection of Filtration Equipment Technology [1,12]

It is not the purpose of this short section to attempt to provide a comprehensive equipment selection guide for filtration-based solids/liquid separation operations. There are already established sources for such Information; see, for example, Chapter 9 of reference [12] and Chapter 20 of reference [1] . Rather it is intended to Indicate how an understanding of both the properties of the material and the rest of the envisaged process train will facilitate a choice from the available filtration-based options.

The main factors that Influence the choice of technology are:

2.1 Scale

The scale of the operation is not normally too stringent a constraint since most devices are available in a range of sizes to handle a variety of capacities. In ‘general, however, very small scale separations will not usually command the most expensive filtration plant If thermal drying can follow the mechanical dewatering stage. For high value feedstreams (e.g. pharmaceuticals etc) other factors may override this option, however.



2.2 Solids Recovery, Liquids Clarification or Feedstream Concentration

As a generalization most solids recovery dewatering operations will Involve the formation of a filter cake whilst clarification (Chapter 4) procedures will often avoid cake formation in order to maintain a high flux of liquid. Where feedstream concentration is required two options arise. Either a cake may be formed which Is then reslurried to a higher solids content, or a continuous thickening process may be employed. Very often a crossflow filtration arrangement will be appropriate for such a continuous thickening arrangement.

2.3 Rate of Sedimentation

The rate of sedimentation of a suspension can have various effects on the choice of filtration plant. For example a bottom fed rotary drum filter may not be suitable for slurries containing a fraction of very large or very dense particles since these may settle out to form a "heel" well before they can be transported to the bottom of the drum. The sedimentation behavior Is also often critical In determining the structure of a filter cake closest to the septum. Thus If the initial filtration rate Is properly controlled, the bottom of the cake consists of the largest, fastest settling solids which may help to trap the finer end of the particle size distribution and thus reduce blocking and blinding. A third area in which the suspension settling properties are of paramount importance Is where a filtering centrifuge Is being considered as dewatering device. For these machines, the mechanism of operation entails a rapid settling of the solids phase In the centrifuge bowl followed by flow of the supernatant through the, hopefully, porous bed. The way In which this bed is formed and the properties that result will thus depend on the settling characteristics of the suspension.



2.4 Rate of Cake Formation and Drainage

The rate at which the height of a filter cake rises can easily be assessed using simple laboratory filtration tests (see next section). It will depend on both the solids loading and the porosity and structure of the cake itself. This property has obvious repercussions on the geometry and necessary dimensions of suitable filtration equipment.

2.5 Batch vs Continuous Operation

This is clearly a critical question which must be addressed by looking at the solids loading and rate of cake build-up, etc.

2.6 Solids Loading

As already explained, this factor will affect (Ill), (iv) and (v) above. In addition it will strongly Influence the flow properties and hence the rate at which the suspension can be presented to the filter If this proves to be limiting.

2.7 Further Processing

It is necessary to consider the Influence of additional operations which may either accompany the filtration or follow it in further downstream processing. Possibilities include washing, air blowing and thermal drying. The physical nature of the final product may also be relevant here (e.g. in re-dispersible systems).

2.8 Aseptic or “Hygienic” Operation

When handling biological materials for pharmaceutical, food or other products, the necessity to be able to clean and sterilize a filter my impose particularly stringent demands. A detailed discussion of the relevant Issues and the suitability of various filters (and other plant) to aseptic operation Is given in the BIOSEP Report SAR 1 1401.



2.9 Miscellaneous

Various other factors are likely to Influence decisions about choice of dewatering filter device. Of these the economics of the whole process Is probably the most important. Such considerations will not be considered here but are discussed In reference [45]. It is, however, worth pointing out that process decisions cannot be taken on the basis of economic factors in isolation. Very often physical constraints (e.g. those discussed in Section 3.5.2(b)) render an otherwise economically attractive strategy impossible.

In order to illustrate the influence of the factors described in (i) to (ix) above, Table 1 presents an impression of the range of suitability for commonly available filtration devices.

Having briefly considered the main factors influencing a choice of filtration technology, a short discussion of two related topics is appropriate here. These are the relative merits of dewatering by shear versus compression and by vacuum versus positive applied pressure filtration.

2.10 Shear versus Compressional Deformation

During the latter stages of cake filtration, further dewatering is often achieved by the application of direct mechanical pressure to the cake itself - this Is the consolidation or expression process described In Section 3.5.2(b). Such a densification of the cake, In order to expel further occluded liquid, may be promoted by either an applied shear or uniaxial compressional deformation. For either case no change In the structure will result until a critical stress, the yield stress, has been exceeded. Figure 4 compares the yield stress for both shearing cay) and uniaxial compressional (Py) deformations for samples of BaC12-coagulated, polystyrene latex suspensions. The latter provide a convenient model which mimics a typical flocculated cohesive filter cake [46-47]. It can be seen that shearing forces are effective (In the sense of exceeding the relevant yield stress) at much smaller stresses (by some 1-2 orders of magnitude); these shearing motions will often enable densification In their own right via structural rearrangement and the concomitant collapse of the cake structure. The advantages of dewatering by shear or a combination of shear and compression are already exploited in many filtration rigs, e.g. counter-moving belt filters [49], but there are almost certainly further gains still to be made in this area.



2.11 Pressure versus Vacuum

There are a number of hardware-based factors which favor a choice of pressure over vacuum filtration or vice versa and these are fairly simple to assess. Thus, in general, positive pressure filtration, being capable of yielding larger trans-septum driving forces, can yield greater filtration rates and hence reduce the size of dewatering plant.

On the other hand vacuum filters have the advantage of simple construction and ease of continuous discharge in operation. They are, however, normally limited to total driving pressure drops of - 0.8 bar and, In the normal way, unsuitable for the filtration of suspensions containing volatile solvents.

The above factors relate to the actual filters. In addition, there are more subtle factors, some of them less well understood, that pertain to suspension properties. Of these the most important is the cake compressibility. For a perfectly incompressible cake (s = 0) and a constant pressure filtration, equation (20) indicates that the filtration time for a given slurry volume is inversely proportional to the driving pressure. Thus potentially large gains in rate may be expected by the use of positive pressure drops greater than a bar compared with the vacuum configurations. For compressible cakes (s > 0) the same equation predicts that the advantage to be gained may be considerably attenuated by the pressure dependence of the cake resistance. An assessment of cake compressibility, for example by using the methods described later, Is therefore highly desirable if the efficiency of increasing the trans-septum pressure drop is to be predicted.

Finally, to illustrate the subtlety of some of these effects, attention Is drawn to recent membrane (but not crossflow) filtration studies of Leaver and Bewdick 1421. Studying the filtration of protein (USA) solutions these workers have observed twice the permeate flux for vacuum compared with positive pressure filtration even though the trans-membrane pressure drops were apparently identical. The reason for this behavior is unclear, but presumably involves some sort of membrane fouling.



3 Suspension Conditioning Prior to Filtration

Suspension conditioning may involve a simple mechanical treatment of the suspension, the addition of a so-called filtration aid, or a combination of both. The range of possible treatments may be conveniently divided into these two categories:

3.1 Simple Filtration Aids

Using the term "filtration aid" in its broadest sense there are three general classes of aid. The first class contains those pretreatment chemicals which are added to modify the state of flocculation or coagulation of the suspension prior to filtration [50,51].

Commonly these additives may be inorganic, e.g. Al or Fe salts or polymeric, e.g. starches, gums, polyelectrolyte’s etc. The conventional purpose of such aids is normally to enhance filtration via one of the following:

(i) Production of open aggregates so as to yield a porous filter cake thereby

achieving fast filtration rates [50-52]. (ii) To yield strong aggregates so as to prevent wash-off and attrition; blinding

and septum fouling is therefore reduced [50-52]. (iii) To improve the suspension rheology (Chapter 7). (iv) To modify the wetting behavior of the medium on the suspended phase.

It should, however, be borne in mind that if further, mechanical dewatering of the filter cake by compression is ultimately to be sought, then factors (i) and (ii) will later prove deleterious. A compromise must then be struck to enable a structured cake that may be compressed under modest driving pressures yet retain sufficient porosity during the actual filtration for reasonable flow rates to ensue.

Since the selection and action of flocculants is discussed in detail in a separate section of this chapter (3.7), no further mention of these will be made here.



Additionally the reader may wish to refer to Chapter 2 (Sections 2.4 and 2.5) for details of flocculation mechanisms and the resulting floe structures.

The other two classes of filter aids are the so-called *pre-coat" and "body aid" additives (1, 12). The purpose of the former is obvious and serves to provide an enhanced filter medium surface on which a cake may be laid down. It is usually formed by re-circulating a pre-coat slurry through the filter (typically a rotary vacuum device or similar) prior to the application of the suspension of Interest. A Body Feed on the other hand is completely mixed with the suspension requiring filtration before it reaches the filter device. It serves to Increase the porosity of the developing filter cake (i.e. Factor (i) above > and hence to lengthen the filter cycle time. An indication of the efficiency of either pre-coat or body-feed filtration aids may be gained by incorporating these additives in a small scale laboratory filtration trial such as those described in Section 3.5.6, In the main the function of the former may be assessed by its effect on the measured septum resistance, The body-feed aid on the other hand should have the effect of reducing the specific resistance of the filter cake.

The properties of some commonly encountered pre-coat and body-feed filter aids are presented in Table 2. Further detailed discussion of the use of these is provided in references [1, 12, and 52]. Finally it is worth noting that surfactants are often employed in order to reduce the ultimate moisture contents of filter cakes [53].

More information on this aspect of chemical pre-treatments may be found in a later part of the chapter, Section 3.7.4



3.2 Mechanical Treatments

A large number of options are available for suspension pre-treatments that do not necessarily involve inert or chemically active additives. The following list, plus a few pertinent references, covers some of the most commonly used methods:

(i) Shear Treatment - often employed to reduce the apparent viscosity of the

suspension [46-48]. (II) Degassing - more frequently employed prior to gravity separations. It may

be necessary before the filtration of certain biological products, however (see Section 3.8).

(iii) Suspension Ageing - like (i), (iv), (v), this technique is aimed at improving

filtration performance via a modification of the flocculated structure of the suspension, e.g. in the manufacture of catalyst supports.

(iv) Heat Treatment/Freeze Thaw I543. (v) Acoustic Methods - generally used for biological systems (see Section 3.8)

[55].

It is important to note the immense potential value of suspension conditioning to filtration operations. The field of biotechnology covers many examples where such conditioning has either a profound influence on the process economics, or is absolutely essential to the Integrity of the product. For example, to avoid protein denaturation, degassing may be an imperative conditioning step. A full and valuable review of many aspects of conditioning relevant to bio-separations is provided in BIOSEP SAR Report "Primary Solid/Liquid Separation" [40].

Finally in terms of mechanical treatments it Is appropriate here to mention for completeness a technology development program being carried out by Batelle into "Combined Fields Separation Processes". The objective of this sort of approach Is to identify combinations of separation means such as electric and acoustic fields, such that synergistic advantages In dewatering may be achieved.



In the cases of electro-acoustic and ultrasonic-assisted dewatering, Batelle claim highly significant Improvements In the rate and degree of filtration for suspensions containing particles such as coal, biological’s, paper pulp and food materials. Unfortunately technical details are not yet available though a number of patents have been filed. Although some of these treatments do not strictly involve suspension conditioning, It Is clear that there is considerable potential for the exploitation of filtration-based processes combined with other separation fields in this way.

Post-Filtration Treatments and Further Downstream Processing [56]

An outline of the influence and theory of three typical post-filtration operations, the washing of filter cakes. air blowing and thermal drying, serves to illustrate process interaction with the filtration operation.

4.1 Washing [56, 59]

Filter cake washing is usually employed to effect a purification of the cake by removing entrained soluble’s, or less frequently to recover the mother liquor where the latter is of high value. The two main parameters of interest are the quantity of wash liquor required to achieve the required level of solute removal and the period of time taken for this degree of washing to be attained.

Probably the simplest approach to calculating the required quantity of wash liquor has been provided by Vakeman. He distinguishes between filter cakes still holding filtrate in the voids, i.e., “saturated” cakes, and those that have been blown dry, the unsaturated cakes. For both cases Vakeman has analyzed the various mechanisms influencing the washing process and produced charts of the fraction of recovered solute as a function of the wash ratio, (i.e. the volume of wash liquor X the cake voidage) and one other dimensionless parameter. These then permit a very simple way of calculating the volume of wash liquor from small scale laboratory tests. Further details are not relevant here but good accounts of the use of the tests and theory, together with the charts, are provided in the references [1,12,56-59].



Once the volume of liquor has been calculated, the washing time is very straightforwardly estimated from the final filtration rate that was observed following cake build-up. Inasmuch b both the wash time and volume depend upon cake porosity and tortuosity, it will be appreciated that the factors that influence the mode of cake lay-down (including the various possible pre-treatments) will be very relevant to the washing performance.

4.1.1 “Air-Blowing”

The use of "air-blowing" as a method of dewatering filter cakes is strictly not restricted to air alone; other gases or vapors, for example, nitrogen or even steam may be used. For biological or food suspensions the latter may provide an additional role for purposes of sterilization (see Section 3.5, 7(c)). The gas is propelled through the cake in a fashion appropriate to the filtration mode, hence for vacuum driven systems atmospheric air is commonly sucked through the cake (deliberately or otherwise) following "breakthrough". With filter presses, pressure nutches etc, compressed air is forced through the pores of the cake in order to displace as much moisture as possible. By using heated air or nitrogen some additional drying action is available; such techniques are, however, normally restricted to small scale or high value products usually having special problems of toxicity etc such that normal drying techniques are difficult to apply*

The fundamental guiding principle in "air-blowing" is that the applied gas pressure must be sufficient to overcome the capillary forces tending to hold liquor within the pores of the cake (see Sections 3.2.9 and 3.5.7(a)). Probably the best current model for this process has been provided by Vakeman. Unfortunately, in terms of real operating experience, the predictions that it provides are of limited accuracy even for near-ideal systems containing hard particles of quasi-spherical geometry. Worse than that, for suspensions of high aspect ratio particles (e.g. needles or plates), or for compressible cakes or those prone to cracking, Wakeman's method is of little practical value.

In terms of more empirical approaches, experimental work on a laboratory or semi-technical scale can be used to make predictions of dewatering time, final product moisture content and the air flow required. A number of cautionary points should, however, be noted.



Since the final moisture content can be very sensitive to changes in the particle size distribution and the way that the cake Is formed, it is very Important to use identical material for the laboratory tests and to ensure that factors such as air flow rate and cake thickness are reproduced as closely as possible. Even so, as a "rule of thumb", it should be noted that small scale characterization tests tend to yield an optimistic figure for final moisture content since effects such as cake compression and cracking tend to be more prevalent on large scale.

Further discussion of most of the above features as well as some more practical examples are provided in the references [60-65].

4.1.2 Drying [67 -76]

It is not our intention to treat the subject of drying in any detail here. However, a short discussion is included for completeness to highlight the importance of considering the interaction between the filtration operation and further downstream processes. It is hoped that a future release of the Suspension Processing Manual will contain a more detailed chapter (Chapter 10) based on those aspects of drying that will be alluded to In the present context.

A general guiding principle that is invoked for most large scale dewatering trains is to remove as much water as possible by mechanical means (i.e. the filtration process here). This then minimizes the expenses of the energy-intensive downstream drying operation. However, it is normally the case that physical constraints imposed by the mechanical dewatering step will intervene before the hypothetical economic optimum is reached (see Section 3.10 - “Process Synthesis”).

There is a large literature, both Internal and external to the Company, based on drying. A recent report by the FCMO drying team I661 described three typical regimes of “paste preparation prior to a drying operation”:

a. Where there is no requirement for pipe flow. An example of this

situation is where a filter cake is discharged at high solids content and is transported, perhaps by conveyor, to say an agitated vacuum oven for final drying. It will be typical here to obtain the maximum, physically-possible dewatering during the filtration.

b. Where a filter cake is re-slurried in order to deliver it by pipe flow to

typically a spray-drier. Clearly it is pointless in this case to



mechanically dewater to the ultimate physical limit. The filtration step should be tailored towards facilitating the re-slurrying o f the filter cake to a manageable suspension.

c. Where the paste is formed into a chosen, stable, physical shape to

accelerate the subsequent thermal drying. For such cases the requirement of the filtration stage is to provide a paste with rheological properties that allow this shaping process, e.g. by extrusion. This situation is relevant to the formation of catalyst supports and ceramic materials in general.

The sorts of interactions between mechanical and thermal dewatering indicated in (a - c) above are variously discussed in the drying literature [67-76]. It would appear, however, that relatively little Is known of how the morphology of the filter cake influences the rate of thermal drying. For example the relationship between, say, filter cake porosity and the necessary residence time in an oven drier would be a useful one to establish. Thus such Interactions would usefully be the subject of future research. Finally the subject of drying as part of a solids Isolation process is very critical when a redispersible solid is desired. This latter topic is treated in detail in Chapter 13 of the manual.

5 Testing and Characterization of Suspensions

5.1 Introduction – Suspension 5.2 Properties relevant to Filtration Performance

In order to best utilize the principles and theory that have thus far been presented, it is necessary to know as much as possible about the "colloidal" properties of the suspension requiring filtration. Both the properties of the pre-filtration suspension and those of any filter cake that is formed are of importance. All or any of the following are likely to be relevant:



5.2.1 Pre-Filtration Properties of Suspension

(i) Suspension viscosity including any tendencies towards shear degradation, thixotropic or any other structural modification following shear flow 146-481.

(ii) Suspension medium viscosity and wetting characteristics on the

solid (Chapter 2).

(iii) Settling properties of the suspension, particularly the rate of sedimentation. Relative density of solid phase. Floe size and structure. (Chapter 2, References [50,5])

(iv) The size distribution of particles and/or aggregates that are

present (Chapter 2).

(v) The ease with which flocculated structure, and in particular the above size distribution, may be modified by mechanical treatments or inert/chemical additives. Such modifications will, of course, also influence the other suspension properties above.

5.2.2 Properties of Filter Cake

(i) The mechanical strength of the cake and hence its resistance towards consolidation and the variation of this property with degree of consolidation (Section 3.7, References [50-52]).

(ii) The porosity of the cake as a function of voidage, that Is the

tortuosity of the path that supernatant must follow through the cake. This property then is correlated with the cake resistances [51].

(iii) The influence of mechanical treatments and additives to the

suspension and the actual filtration conditions, e.g. rate of cake lay-down, on the cake strength and resistance.

Once a representative number of the above suspension properties has been determined so as to enable a good understanding of its "colloidal" behavior, the knowledge may be applied to the following targets:

(i) Identification of the most appropriate plant and scale for the

filtration unit operation or suggestion of a better, alternative dewatering means other than filtration (see Section 3.10).



(ii) Optimization of the way that the operation is carried out. For

filtration this will include the choice of filtration conditions and details of cycle time as well as their Impact on further downstream processing.

(iii) Prediction of optimal plant operation. Thus, for example, it is

essential to know what level of performance in terms of rate and degree of dewatering can be expected under a given set of conditions. This Is of paramount Importance for scaleup calculations (see Section 3.2).

(iv) Removal of process "bottlenecks" and correction of plant operating

problems. This again relies heavily on (iii) and the identification of "benchmarks" for optimal performance.

(v) To suggest where conditioning techniques 'and/or filtration aids

may be desirable or appropriate. Whereas the means and optimal extent of pre-treatment should ideally be estimated from small-scale experimentation.

5.2.3 Laboratory Scale Filtration Rigs [77-80]

A number of small-scale rigs exist and these may be applied to the measurement of filtration rates, filter cake properties and the Influence of suspension properties on them. These rigs are commonly used for the Initial derivation of data for scale-up purposes. If there is any doubt, they may also be applied to the question of identifying the filtration mechanisms of Section 3.5.2, although they are predominantly applied to cake filtration tests.

Apparatus for measuring filtration rates on a small scale have been described by various workers [77-80]. The rigs of Allen & Stone [77], Gregory [78] and Bridger [80] are representative and of straightforward construction. The Allen & Stone apparatus, is well automated and their paper describes its mode of operation in detail. A reproduction from their paper is given In Figure 5 from which the basic operating principles are easily deduced. The original objective of the rig was to obtain data for scale-up purposes. In contrast to this, the equipment of Gregory was initially developed in order to assess the value of polymer flocculants as additives to filtration slurries and to derive optimum polymer dosages by experiment. The report of Brldger and Tadros uses a test rig to investigate



some fundamental aspects of the Influence of suspension properties on the mechanistic details of cake filtration.

The "equilibrium" and kinetic aspects of the consolidation process (Sections 3.2 and 3.5.2) may also be studied on a laboratory scale. Tests may be carried out on a small-scale variable volume filter such as the Piston press supplied by Triton Electronics, for example. With these devices it Is possible to measure the solids content of a consolidating filter cake as a function of pressure and also the rate at which this degree of consolidation is approached. In the same vein a gas-pressure driven pressure filter for laboratory scale tests from 0- "10 bar is now available from Schenk.

5.3 Means of Monitoring Flocculant Dosage

The means of selecting appropriate flocculants and assessing optimal dosages is dealt with more fully in Section 3.7 of this manual. However, a recent addition to the range of portable, small-scale testing methods is well worth a mention in the present context. The new test method is an on-line monitor for flocculation control [81-82]. Its operating principles are based on the measurement of turbidity fluctuations In the flowing suspension of interest. Gregory [82] has shown that the root mean square fluctuation intensity can be related to the suspended particle size distribution via a semiempirical relationship. This conclusion enables the RMS signal to be used as a fast and sensitive Indicator of floe formation.

The device, marketed by Rank Brothers of Bottlsham, is relatively cheap (ca $7.5K at the time of writing) and constructed in a way that makes it ideal for portable use and for continuous monitoring. Gregory has described applications where the device has been tested both in clarification and in achieving flocculation of more concentrated suspensions such as those requiring filtration. The method may also be comfortably applied to suspensions that tend to foul the sample cell simply by monitoring the ratio of both the RMS fluctuations and the average light transmission. This ratio has been shown to be relatively invariant to the deposition of modest quantities of material on the surfaces of the sample cell. A recent ad hoc, trial of the Rank Brothers monitor has been made. The device proved a sensitive indicator of flocculation In bacterial suspensions to which high molecular weight cationic polyacrylamides had been added. The response time was also fast demonstrating the potential of such instruments as continuous dosage monitors.



5.4 Filter Cake Testing

For the general case of the suspension processing of fine solids the most common application of filtration involves cake formation and treatment. It is therefore appropriate to consider the parameters and means by which the properties of the filter cake may be characterized. The three principal properties that define the behavior of the cake are its strength, its permeability or, conversely, resistance, and the rate at which it is laid down. Methods for determining these will now be given. 5.4.1 Strength Testing (See also piston press described earlier)

This is relevant both to an understanding of the influence of the pressure drop on the ordinary compressible cake filtration rate (as described by equation (14)) and to the subject of compression dewatering following filtration. Although a number of empirical measures of cake "strength" exist, the most suitable and fundamental parameter to use is the uniaxial modulus of compression, K [83] or the compressional yield function Py(Ø) described earlier. The former my be defined In terms of the effect of pressure on a cake volume (V) or concentration (Ø) change:

The modulus, K, is a very strong function, (K ~ Ø3-4 of concentration, Ø, and depends upon the nature, shape and size distribution of the priory particles as well as the structure of the cake and the Interparticle forces. K Is related to the function Py(Ø) and is also very similar numerically to the conventional infinitesimal modulus of shear G(Ø). This fact enables its determination by straightforward laboratory techniques. (For further details see Section 3.2,4.) Arguably the simplest of these to use is the Pulse Shearometer Cell (Figure 8 of Section 3.3). This device enables the rapid determination of G (~ K) for a small sample of slurry or filter cake by measuring the propagation time of a low strain (~ 10-6) shear wave between two discs mounted on piezo-electric crystals in the cell. Calculation of G requires only the propagation speed of the wave, u (from the disc spacing and propagation time), and the density, p, of the cake or slurry:



The shearometry technique is generally restricted to the range 103 < G < 106 dynes crn-2 but this is not usually a problem. Where cakes of higher strength need testing an alternative strategy may be adopted by measuring the compressional yield point, Py(Ø), in a centrifuge. The filter cake must now be formed in situ from the slurry (to be filtered) in a centrifuge tube. A measurement of the height of the equilibrium sediment as a function of gravitational field enables the evaluation of Py(Ø) over a range of concentrations (Figure 9 of Section 3.3). The upper bound of Py(Ø) measurable by this technique is constrained mainly by the gravitational field that the centrifuge is capable of (safely) producing and the density of the solid phase.

Measurements of either G or K may then be used to evaluate the pressure, Pt, which must be applied to the cake in order to concentrate it to concentrations, (Ø)*:

This then assumes a long enough contact time such that kinetics will not prove limiting. That Is It represents the equality Ps = Py(Ø), the ultimate or structural limit. For the centrifuge technique Py(Ø) may, in principle, be calculated from a single experiment. Using the shearometer cell a series of determinations at different slurry concentrations must be made. In both cases equation (29) is solved either by graphical or numerical integration. An example of the calculation is provided in the next section. Finally it may be noted for completeness that K may also be measured directly In a compression cell (84) but, for practical purposes, one of the two methods described above is usually more straightforward and of sufficient accuracy. For further clarification of the definition, interpretation and measurement of G(Ø), K(Ø) and Py(Ø) the reader is referred to Section 3.2.



Finally It should be noted that the rheological parameters which quantify compressional strength, that Is G, K, Py(Ø), may be related to the compressibility exponent, "s", of equation (14) provided that assumptions are made regarding the dependence of cake resistance on voidage. The simplest and best known model for the latter relationship is that due to Kozeny and Carman. Examples of this approach may be found in the next section.

5.4.2 Cake Permeability or Resistance

A number of approaches may be used to quantify the specific or integrated cake resistance to flow of filtrate. The simplest small scale approach utilizes one of the laboratory filtration rigs just described together with the equations for Idealized constant pressure, cake filtration, (18) - (22). A plot of reciprocal filtration rate, dt/dV, as a function of cumulative filtrate volume, V, should yield a straight line from the slope of which the specific resistance, r, may easily be calculated. The variation of the specific resistance with pressure drop may be evaluated from a series of experiments at different driving pressures. (A plot of Log (S. Resistance) versus Log ΔP yields the coefficient of variation "s" from the measured slope.)

Alternatively this variation may be calculated from a knowledge of the compressional modulus, K(Ø) as will be shown in the next section. However the parameters for cake resistance are determined by laboratory measurements, the warnings given in Section 3.5.2(b) (iv) with regard to scale-up must be noted and checked.

In passing It is worth noting that a number of commercial devices exist for rapid, “on-site”, empirical measures of the resistance of a given paste, sludge, or cake to filtration. For example Triton Electronics manufacture such an instrument to measure the empirical quantity, the "capillary suction time" or CST. Although very useful as quick Indicators of qualitative filtration behavior, such Instruments should In the main be reserved for control monitoring or "trouble-shooting" purposes. In addition to the CST test, many filtration equipment manufacturers have similar quick and simple tests for gaining an immediate feel for the behavior of a given suspension. Thus, for example, the dipping of an inverted Buchner funnel and septum (connected to a vacuum line) briefly into a suspension, enables a good Indication of what thickness of cake would be picked up by a bottom fed rotary vacuum filter.



Despite the usefulness of such as guidelines and the CST approach, It is recommended that serious design data for scale-up purposes should be derived from more careful and fundamental measurements of the type described above.

Finally, an alternative approach to the study of cake collapse in consolidation may be gained from an observation of the rate of fall of the sediment zone boundary in a centrifuge. There are problems in interpreting such experiments to predict filtration behavior since the consolidating pressure varies down the sediment and the experiment does not constitute a true "filtration". However, there is otherwise much to commend the approach. In particular the network drag parameters used in the Buscall and White theory of consolidation rates could be estimated by this means, The experiment to achieve this entails a centrifuge within which the sample tubes are transparent and illuminated by light from a stroboscope. The latter is triggered by the centrifuge rotor and thus a "frozen" image results enabling the kinetics of the consolidation process to be followed. Further research along these lines would be desirable In the future.

5.4.3 Rate of Cake Formation

The principal importance of the rate of cake formation in terms of cake height is for sizing purposes during equipment selection and scale-up calculations. Purchas has defined a standard cake formation time, tF, for a 1 cm thick cake which IMY be related to the specific cake resistance ro. Hence these quantities may be inserted in the equations for constant pressure or rate filtration given previously. Further details and examples of such calculations are given in Purchas' book, "Solid/Liquid Separation Technology" [12].

6 Examples of the Application of the Forgoing Principles

In this final part of the filtration section of the dewatering chapter, some exemplification of the foregoing principles and theory is appropriate. To this end, three different examples of processes involving an important filtration operation will be presented. These have been selected to provide an indication of the variety of suspensions that may be encountered and the concomitant considerations and difficulties that apply to each.



6.1 Dewatering of Calcium Carbonate Slurries

The material for this example has been taken from a recent report, IC 01713 C853, by Asher and Stewart - "Prediction of CaC03 Dewatering and Other Processing Characteristics from “Suspension Property Measurements". The main body of the report involves a comparison of theoretically-predicted and experimentally-observed moisture contents in pressure-dewatered CaC03 magmas. The shear and uniaxial compressional moduli, G and K, were measured by Shearometry and centrifugation respectively as described in Sections 3.2 and 3.5.6. Results for three typical samples are reproduced in Figure 6. It is interesting to correlate the different behavior of these samples with their colloidal properties. Curve (a) shows the greatest resistance to densification which is in good accord with the behavior expected of a suspension of very small particles. For curves (b) and (c) the form of the network modulus, K(Ø), is more similar. The slightly larger slope for curve (b) may be rationalized in terms of a stronger flocculated structure resulting from the MPBD (3% maleinized polybutadiene) coating.

From the data of Figure 6, Asher and Stewart obtained the form of the pressure dewatering curve, Figure 7. This was achieved by the simple numerical integration, described in Section 3.2 and I251 using equation (29) which Is reproduced below:

Inspection of the figure reveals two interesting features. Firstly there is a rapid fall off in the degree of dewatering achieved as the pressure is stepped up - note the logarithmic scale. Secondly the pressure-dewatering relationship is affected by the same colloidal particle properties as was the modulus curve. Hence for a given applied pressure, the much finer suspension, curve (a), is dewatered to a considerably smaller extent. The immense value of using the laboratory-scale modulus measurement in order to predict plant-scale expectations of dewatering is apparent.

In order to test the predictions arising from the figure, Asher and Stewart made laboratory measurements of the moisture content remaining in the samples at various pressures, using a laboratory piston press. The results, expressed in terms of moisture content rather than solids concentration, and calculated by the conventional theory (1), are shown in Figure 8. It



can be seen that there is reasonable agreement between theory and experiment. There are, however, significant deviations between the two at the highest consolidating pressures and it was deduced that the origin of this discrepancy lay in the derivation of the modulus, K, from the centrifugation experiments. A careful re-examination of the approximations inherent in that derivation was made. It was deduced that in general the routine method of analysis, based on a calculation of the pressure head in the middle of the sediment (equation for circles in Figure 8), led to an over-estimate (or upper bound) for K at a given solids content. This follows from the nature of the solids concentration down the sediment. Fortunately the same reasoning led to the conclusion that a calculation of the pressure head and hence K at the sediment base, where the solids concentration was assumed to be [2(H0/H.)-l] Ø0p yielded a lower bound for the modulus. (HO and H. are the initial and subsequent equilibrium heights of the sediment at the various gravitational fields.) As a result of this, the centrifuge data was re-analyzed using a calculation of K at both the middle (Equation 2 of the Figure) and the base (Equation 2) of the sediment. The predicted residual moisture was then derived using the mean of the two calculations (Equation 3 of Figure 8). The agreement with experiment was then found to be excellent and this is also illustrated in Figure 8. A subsequent computer simulation study of the problem demonstrated, the validity of the approach of taking the mean value of the modulus from Its upper and lower bounds. This procedure may be compared with the mathematically more rigorous approaches of Buscall and White (Section 3.2).

Two other aspects of the CaCO3 work are relevant to this discussion of suspension filtration. The first involved a calculation of the pressure at which a significant dewatering would occur due to "air blowing". By combining the Bartell I861 equation for capillary pressure in a porous plug of powder with a relationship for the effective pore size, the following relationship was utilized [87] after the paper of White:



Application of equation (30) yielded the prediction that substantial ‘air blowing" would be expected for the CaCO3 filter cakes at pressures of ~ 20 atmospheres (typical Winnofil-sized CaCO3) and ~ 60 atmospheres (for ultrafine CaC03). These trends parallel those of Figure 6.

The final feature of Interest concerns the rate at which the CaCO3 slurries were observed to filter. Asher and Stewart adopted an approach to filtration rate predictions akin to equations (18) and (20) of Section 3.5.2(b). In particular they looked at the variation in rate with applied pressure drop, ΔP. Increasing ΔP has two main effects on the rate. Firstly it tends to increase it for obvious rheological reasons, le because of the Poiseuille-like flow rate, This increase is, however, attenuated by an increase in cake resistance due to its finite compressibility as per equation (14). The latter was assessed for CaCO3 cakes from the compressional modulus measurements; Figure 9 reproduces the predicted and observed filtration rate as a function of pressure drop. Although the agreement is far from perfect it can be seen that the approach yields a very satisfactory estimate of the dependence. Further examples of work in this area are given In the third set of examples (Example (c)).



6.2 Dewatering of Organic Products – Procion Dyestuffs

An investigation of the dyestuff Procion Blue H-ERD by Ramsay and Sadler [88] revealed that three different physical forms for the dyestuff may be produced following a "salting out" procedure. In the dewatering chain for these dyestuffs the precipitated solids are filtered under pressure in a press (at about 40 psi). The filter cake at about 40% solids is then mechanically conveyed to a Hy-disperser where it is re-slurried with water prior to being pumped to a spray drier. This fairly typical example serves to illustrate the influence of interactions between a filtration operation and other operations up and downstream (cf Section 3.5.5(c)).

The main upstream influence arises at the precipitation step where a change in the salting out conditions can cause a transition from the production of the favored physical form, essentially a precipitate of aggregates, to one of two unfavorable forms. The authors of the report described these unfavorable forms as an amorphous and a quasi-crystalline phase. The morphology of all three forms was deduced from a combination of electron microscopy (TEM, SEW) and X-Ray diffraction (for the quasi-crystalline phase).

In a simple laboratory filtration test using the apparatus described previously, the three physical forms yielded the following results under identical experimental conditions:

Thus a control of the precipitation stage can greatly enhance both the filtration time and also the degree of dewatering obtained. Part of the influence of the physical form on the filtration rate may be attributed to the greatly improved rheological characteristics of slurries of the aggregated material relative to the other possibilities. Figure 10 is a reproduction of a plot of viscosity as a function of shear rate for all three cases. It can be seen that the apparent viscosity of the aggregate form is an order of magnitude lower than the crystalline form.



Finally, this example provides an indication of the sort of consideration that is needed with a view to further downstream operations such as drying. As mentioned previously, the filter cake requires re-slurrying so as to provide a pumpable suspension for spray-drying. The present process utilizes a Hy-disperser to achieve this end but this almost inevitably leads to some breakdown of the aggregates and formation of fines. As a result of this the rheology of the fluidized paste becomes unfavorable and extra water must be back added to enable the slurry to be pumped uphill to the spray drier. Clearly a gentler fluidization method would be desirable and this would be achieved If the filter cake were formed in such a way as to permit facile re-dispersion.

Two other studies of the filtration properties of slurries containing organic crystalline particles by G Taylor and co-workers Illustrate other aspects of the optimization of the suspension processing. In an investigation of plant filtration problems with the product Pyridone CE [89-90], a number of analytical techniques, such as XRD, TGA, DSG and microscopy, were used to characterize the nature and stability of various crystalline modifications. In this particular case the major complication arose from the tendency for Pyridone CE to undergo changes of crystal habit during mu filtration. Hence although the feedstock to the plant filters performed adequately in laboratory tests, when scaled-up to plant process times it led to unsatisfactory filtration rates. By characterizing the stability and means of producing the various crystal habits for the product, Taylor indicated a strategy (by an elevation of process temperature upstream of the filtration) for producing a favorable crystal habit which did not undergo polymorphic changes throughout the duration of the plant filtration times and conditions.

Taylor and co-workers similarly studied the filtration properties of 3,5-Dinitro-2,4,6-Trimethyl Benzene Sulphonic Acid [90]. They indicated three key properties that led to a desirable crystal habit form:

(1) A crystal shape that allows rapid filtration yet yields a cake that Is

sufficiently well packed so as not to occupy too much volume for a given amount of feed. Clearly these requirements are to some extent opposing and a compromise in porosity and cake structure must be met.

(II) The cake structure should enable facile washing of the product (see

Section 3.5.5(a)).



(III) The crystal should have a low solubility to enable thorough washing

without extensive loss of product.

Once again a favorable crystal form, rhombic shaped crystals (~ 100 x 250 µm), was identified as that which yielded the most satisfactory filtrations, though the latter performance was sometimes masked by the presence of “fines”. The cake formed from the rhombic crystals was found to be significantly compressible. Hence it was concluded that a Vacuum filter would be a more appropriate choice of dewatering strategy than the 40 psi Filter Press then currently employed.



6.3 Filtration of Biological Systems – Harvesting a Filamentous Organism

This example is included in order to indicate the very different constraints and requirements of dewatering operations for large-scale bio-technological products. The harvesting of the filamentous organism, Fusarium Graminearum, for human consumption as a foodstuff Is a complex and highly interactive process. The organism leaves the fermenter in which it is grown as a dilute suspension containing about 1.5% of suspended solids by dry weight (Øw) + In order to conserve the growth medium and nutrients the ex-fermenter broth will ideally be pre-thickened with a re-cycle of the spent broth. The concentrated suspension, now at 2-3% dry weight of solids, is then heat treated and finally dewatered to yield a moist but resilient cake for further processing [91].

Figure 11 shows the uniaxial compressional modulus, K for a typical sample of this material before and after heat treatment. Concentrating initially on the pre-reduction material (RNA+), it can be seen that a substantial modulus exists at very low solids contents. The physical implication of this observation is that the material is forming a structural network of considerable compressive strength at very low dry weight fractions, (Øw). As a result of biological constraints and the fragile nature of the RNA+ hyphae (a hypha Is a single organism), the pre-thickening step must avoid both excessive residence times (preferably < 15-20 minutes) and high shear fields, The Information from the modulus measurements then indicates that a filtration means of pre-thickening is likely to satisfy these two constraints most satisfactorily since the strong, open network of hyphae should filter (or "drain") fast whereas unit gravity sedimentation / flotation will be much too slow. Centrifugation is likewise an unfavorable choice at first sight due to the high shearing forces that are encountered during acceleration and discharge [92].

In contrast the slope and position of the modulus curve for the heat treated material (RNA-) is quite different. Since most of the hyphae have undergone only very small changes in their physical dimensions, It is deduced that the change In colloidal properties is due to a reduction of cell-internal pressure or turgor. The network of hyphae is now in effect much more compressible and this factor Is of key Importance in relation to the use of a second filtration step in the process to generate the dewatered cake [93].



Some preliminary investigations of the pressure drop profile as a function of time for RNA- material have been made by A Harrison of Agricultural Division. These measurements, carried out at various constant feed rates, were made on a porous, sintered steel candle filter. Using the expression for the variation of cake resistance with pressure drop given earlier, equation (14),

the experimental data was fitted to an expression analogous to equation (24) with the assumption of negligible medium resistance. Comparison of the measured curves and those predicted from a fit to the filtration equation shows very satisfactory agreement, The fit yielded a value of 0.93 for s, the compressibility factor thus confirming the high degree of cake compressibility. Thus for a constant rate filtration, once the pressure drop has started to rise, its increase with time is very rapid and this is borne out by Figure 12. Thus factors that control the cake compressibility, eg the mean hyphal length, suspension pH, water hardness, etc, will also dictate the length of a filtration cycle at constant rate. Conversely for a constant pressure drop configuration, any advantage in filtration rate to be gained by stepping up the driving pressure will be limited by these same factors. Thus a fundamental study of the cake and suspension properties and their variation with these parameters enables the understanding of the large-scale filtration behavior. Further details may be found in a recent CCSG report [93].



NB The symbols represent different samples of culture removed from the fermenter over a period of 4 months. Some of the scatter therefore reflects changes in the morphology of the organism under varying environments.

Filtration

Technology

Transcript of Filtration