Filtration of Suspensions

- Problems with the Empirical Filter Design and Benefits by Using the Theory Priv. Doz. Dr. -Ing. Ioannis Nicolaou, FOS Filtration Software & Technology, Sophienstr. 10, 76133 Karlsruhe, Germany, Tel./Fax: +49-721-3849727, e-mail: nicolaou@t-online.de, www.fos-filtration.com INTRODUCTION The great diversity of suspensions and filter media, the many filter types and testing devices, the most different operating conditions but foremost the complex physical processes with the cake forming filtration (formation, squeezing, washing and deliquoring of the filter cake) are cause of the fact, that until now with the solution of filtration problems the pure empirical procedure dominates. Under empirical approach is understood that practical and physically based equations for planning, realization and analysis of tests as well as for filter calculation are not referred to. The in practice active researcher is for this reason forced to exert a too high, often unsystematic experimental effort that is expensive and time consuming. In spite of this effort, one often is not in a position to quantify the filtration behavior of the given product accurately, to document the justification of the respective apparatus as a better alternative to other apparatus, and to point out optimisation potentials of his apparatus, be it by change of the adjustment parameters and / or by prevention of equipment related negative effects. Incorrectly applied or /and not optimal operated apparatus are therefore no rarity. To this problem adds also the fact that decisions by not objective factors are affected (not sufficient specialized knowledge, sales interests, refusal by the operator to optimise his installation etc.). The aim should be in case of new filtration applications with a given suspension the suitable apparatus under consideration of operational boundary conditions reliably to design and optimally to operate with minimal cost and time. With already operating filter devices shall the filter performance be checked and be improved by adjustment of important influencing parameters. By still further existing unacceptable filter performance another apparatus should be taken in consideration. An important factor which contributes to the incorrectly employed or operated filter devices is the lacking "common language" among the filtration experts. By that is above all understood on one hand the absence of standardized equipment (hardware) and on the other hand the lacking standardization in respect of the analysis of filtration results and, related to this, in respect of the characterization of the filtration behaviour of suspensions (formation, squeezing, washing and deliquoring of the filter cake). By the most diverse test devices result automatically not uniform test conditions with the consequence that measuring results in spite of tests with the same product depending on the applied equipment deviate from each other in respect of quantity and quality of the measuring data (human errors are here not meant). A laboratory nutsche e. g. with a cake squeezing capability as well as with an automatic filtrate recording can compared to a nutsche without cake expression device and without measuring data recording provide additionally data to the 1

squeezing behaviour of the filter cake and with less experimental effort considerably more information to the kinetics (filtrate mass as a function of time) during the different filtration phases. Depending on the design of the filtrate drain base occur with otherwise identical test conditions different quantities of accumulated filtrate, which with the later analysis results in from each other deviating filtration parameters (cake permeability and filter medium resistance) but also in different cake residual moisture contents. Is e. g. a device supplied with a transparent cover then one can observe the individual phases also visually. Thus can additionally the moment be registered at which the cake formation phase is finished. Looking at the deliquoring phase e. g. occurring cake cracking can be observed. Depending on the size the filter area are falsifications by border effects different. Because of that it is obvious that the absence of a standardized device would make it necessary that each time filtration results are shown the test equipment and other test conditions should be recorded accurately. Not until then is an evaluation of the test results and the hereon basing filter design possible by third party and possibly a comparison with own analyses. The necessarity for standardized filtration equipment does not mean, however, that with certain filtration tasks from the norm deviating solutions should not be employed. A first step to the creation of a common language regarding test equipment represents the working out of the VDI guide - line 2762 /1/. In Fig. 1 is a scheme of this laboratory device to be seen. In the following is only referred to this laboratory equipment, with which by the way the author has been working for over 13 years.
lid with inspection Glass

pressure gauge computer pressure filter cell

gas flow meter gas pressure control valve

double-walled filter cell

button balance

cake formation ring nutsche bottom

pressurized gas

Fig. 1: Scheme of the according to the VDI guide lines recommended laboratory equipment for examination of the filtrability of suspensions /1/, /2/. More important than a standardized filter equipment is the standardization in respect of the characterization of the filtration behaviour of suspensions based on practical and physically based models, that is the inclusion of a generally accepted theory as inseparable component with the 2

planning, carrying out, analysis of tests as well as for the design of filter devices. The standardized modern test equipment alone (only the hardware considered) is equivalent to using a computer with unused possibilities because it lacks the adequate software. The lacking or unsatisfactory analysis of obtained filtration data is the most important cause for the lacking "common language" because no product-specific parameters for the quantification of the filtration results are used. CAKE FORMATION Is e g. the cake formation phase considered then one would with the pure empirical procedure that is with no application of models the cake formation thus characterize that the measured time is stated which was required until a given suspension mass with certain given pressure difference was filtrated. Alternatively or complementing is the filtrate amount stated which occurred at a certain time, the so - called mean filtrate performance. As a rule tests are carried out for different pressure differences and cake heights or suspension quantities, respectively, where the corresponding filtration time or/and filtrate performance are measured. These times or filtrate performances are then the basis for the empirical filter design. Why, however, is the statement of the time or rather of the filtrate performance problematic for the characterization of the filtration behaviour? Because these measured values relate to a certain pressure difference, a certain suspension mass that is a certain cake height and additionally to a certain solids concentration and a certain filter medium and each variation of one of these values results in other times or rather filtrate performances. Because no equations are used one had e g. for each pressure difference and cake height to perform a new test. This is as a rule time consuming and expensive, particularly in conjunction with pilot experiments. Even the automatic data acquisition of the filtrate occurrence (see Fig. 2) does not help, if the test results are not evaluated with the aid of a physical mathematical model. Such an in practice well-tried model represents equation. (1) : p t v =

2 Pc ( p )

v + Rm ( p )

(1)

p (Pa) = Pressure difference, t (s) = filtration time, v (m3/(m2 s)) = specific filtrate volume, (Pa s)= Filtrate viscosity, (-) = solids concentration parameter, Pc (m2) = cake permeability, Rm (m-1) =

filter medium resistance Are however the results of Fig. 2 theory based plotted that is in consideration of equation. (1), as is shown in Fig. 3 then it is possible to obtain a lot of information which otherwise with the nontheory based graphs of Fig. 2 is not available. From Fig. 2 we can e g. make no prediction in respect of the filter medium resistance and its relation to the cake resistance. Also the question concerning the compressibility of the filter cake cannot be answered. Moreover can equipment related problems not be recognised which during the filtrate process act adulterating, such as accumulated filtrate at the start of the cake formation. Such problems are, however, visible with the theory-based plots by the deviation from linear progression. A comprehensive treatment of the in Fig. 1 described filtration equipment and the associated equipment related mistakes regarding cake formation and deliquoring is in /2/.

4 bar
Filtrate Volume / ml

0.8 bar

Filtration time / s
Fig. 2: Time dependence of the filtrate volume for two different pressure differences (automatic data recording) The straight lines of Fig. 3, which result for each pressure difference, confirm the validity of equation (1) and allow the determination of the to the respective pressure difference belonging cake permeability Pc and of the filter medium resistance Rm. The first value results from the gradient of the straight line and the latter from its intersection with the ordinate. From Fig. 3 one recognises e g. a too high filter medium resistance. Such a conclusion can, however, with the data plotting according to Fig. 2 not be drawn. The by the author introduced representation p t/v =f(v) differing from the usual even in the VDI-guide lines recommended t/v=f(v) - representation has the great advantage that by this also the cake compressibility is recognised. According to equation (1) one expects with the plotting according to Fig. 3 that with homogeneous and incompressible filter cakes that is Pc f (p) and constant filter medium resistance Rm F(p) all measured data lie on the same straight line (device related effects not considered). Is this not the case, as is shown in Fig. 3, then one has to do with a not negligible cake compressibility. Such an information can by the t/v=f(v) representation of the results not be obtained. Moreover can in contrast to the p t/v - method by the t/v plotting the dependence of the pressure difference on the filter medium resistance not be recognised. From the analysis of many filtration tests has the following mathematical model for the description of the pressure dependence of the cake permeability and of the filter medium resistance proven well (whereby at this point it is referred to the problems of the exact determination of the latter parameter especially with short filtration times):

 p Pc = Pc 0 p0

p Rm = Rm 0 p0

(2)

Pc, Pc0 (m2) = cake permeability to the pressure difference p or p0, (-) = compressibility factor of the filter cake. Rm, Rm0 (m-1) = filter medium resistance to the pressure difference p or p0, (-) =

compressibility factor of the filter medium Are the tests of Fig. 2 by application of equation (1) evaluated by linear regression one obtains for each pressure difference the corresponding cake permeability and the filter medium resistance. These pairs of values are afterwards used to obtain on the basis of equation (2) by regression analysis the compressibilty factors for cake and filter medium und . By this, cake permeability and filter medium function are determined. This task is accomplished by the userfriendly computer program FILOS written by the author, that besides the cake formation phase evaluates the results for the cake squeezing, washing and deliquoring according to practical equations /3/.

Fig. 3: Plotting of the filtration data according to Fig. 2 and equation (1). Analysis with the programme module FILOS-Analysis of Filtration Data /3/. The mentioned functions for the cake permeability and filter medium resistance represent real product-specific functions (the applied filter medium is in this context considered to belong to the product) and are much more suitable in comparison with the statement of the filtration time or the 5

filtrate performance, as this is the case with the purely empirical procedure, for the characterization of the cake formation behaviour. The application of the theory offers in addition to the already mentioned advantages in respect of the evaluation of the filtration results a drastic reduction of the needed tests because only tests for the determination of the parameters of equation (2) are necessary. As a rule three tests with different pressure differences are sufficient. Once one has determined the parameters of the functions of equation (2), the cake formation for any pressure difference and filtration time can be calculated. Moreover can the influence of the filtrate viscosity and the solids concentration (presupposed these values have a negligible influence on the functions of equation (2) ) be quantitatively obtained. CAKE DELIQUORING During cake deliquoring the cake residual moisture and the gas volume flow rate (also known as gas throughput) are the two measured values. The latter value is in practice not always measured, although the knowledge of this parameter is very important for the compressor design. If one takes for example a belt or a cell drum filter, then it is with the empirical filter design important to know, with given cake height and pressure difference (e g. 25 mm and 0.6 bar), which cake residual moisture in realizable deliquoring times are accomplishable. With this question would theoretically one deliquoring test be sufficient. In this case a filter cake of the given cake height with the given pressure difference is formed and deliquored. With filtration equipment with automatic fitrate recording (see Fig. 1) can then by measurement of the cake residual moisture at the end of the deliquoring test and the registered filtrate mass, which during the cake deliquoring was removed from the cake, the time dependence of the cake moisture content be determined. From the given process then can for a certain deliquoring time, which in the real filter operation shall be realised, the to be expected residual moisture be read off (presupposed no scale upproblems). Thereby must be considered that the time-course of the residual moisture can be faulty due to device related problems. For the standardized device /1/ can the final residual moisture value be measured above real due to a cake remoistening. This source of error can, however, be eliminated by use of an appropriate bottom section, where shortly before turning down the pressure difference the bottom side of the cloth is soaked up. A further still more important source of error is the during the cake formation phase in the bottom section of the device accumulated filtrate mass, which with the gas breakthrough gets on the balance. With the determination of the cake moisture content course with time by filtrate mass balancing is then in the region of the gas breakthrough time incorrectly a quick deliquoring presented. In order to make such equipment related problems visible and quantifiable, so that false residual moisture values are avoided as the basis for the filter design, should in addition to the cake residual moisture (target value with the purely empirical procedure) also the cake saturation be used, defined as liquid volume in the cake to cake voids volume. The cake saturation is obtained from the cake residual moisture by the following formula, whereby for its calculation the cake porosity and the ratio of solids density to liquid density is needed. The determination equation for the cake porosity, defined as cake voids volume to cake total volume, is also shown. For its determination is the knowledge of the solids density and the to a certain cake volume relating solid mass necessary: 6

S=

1 s

RF l 1 RF

=1

ms s A hc

(3)

S (-) = cake saturation, (-) = cake porosity, s (kg/m3) = solids density, l (kg/m3) = liquid density, RF (-) = cake residual moisture, ms (kg) = solid mass, A (m2) = filter area, hc (m) = cake height

Backwards-method

Cake Saturation S (-)

Forwards-method Separate measurement Correct course

Glass beads p = 1 bar hc = 15 mm

Filtration Time t (s)

Fig. 4: To point out equipment caused problems with the acquisition of the cake deliquoring kinetics by balancing of the during the deliquoring registered filtrate mass

In Fig. 4 the course of cake saturation with time is shown with tests with a model suspension of glass beads. The aim of these tests was the demonstration and the quantification of the equipment related errors when measuring the cake deliquoring kinetics with the VDI recommended laboratory equipment /1/ (see Fig. 1). The use of glass beads, which resulted in a homogeneous and incompressible cake, served the elimination of other disturbance factors. It turns out, that both the already described so called backwards calculation method (starting from the residual moisture or rather the saturation of the cake after the end of the test "re-calculate" the remaining values by balancing based upon the registered filtrate mass) as well as the so called forwards method (starting from the state before the start of the deliquoring (S=1) "pre-calculate" the remaining values via the balancing of the obtained filtrate mass) clearly deviate from the correct progression. The correct course is shown by the graph which goes through the crossing points. The 4 crosses (X) stand for the results of four tests with different deliquoring end times, whereby each cross stands for the cake end saturation of the respective test which was obtained from the gravimetrically determined cake residual moisture by applying equation (3). According to the results of the diagram of Fig. 4 is the following conclusion to be drawn: If one wants to have reliable residual moisture values for certain deliquoring times, then the deliquoring test at that point has to be stopped and the sought value has to be determined gravimetrically.
7

Before turning down of the pressure difference there should be soaked up, if possible. With relatively high deliquoring times (far from the gas breakthrough time) is indeed the above described backwards method for the determination of the residual moisture course usually rather reliable.

Z in c S u lfid e

Saturation S* ( - )

D e liq u o rin g In d e x K ( - )
Fig. 5: Cake saturation as function of the deliquoring index K for different cake heights and pressure differences.

If one wants with the empirical filter design to determine the cake residual moisture and the gas throughput not just for one operation point, but for several, as they e g. by variation of the pressure difference and/or the filter speed result, then one is as a rule forced to exert a high experimental effort. The inclusion of a practical theory (simple, physically based equations with adjustment parameters) for the calculation of these values would drastically reduce the experimental effort. As was in a similar way argued in case of the cake formation, only so many tests are necessary, which are required for the determination of the adjustment parameters of the equations. After the equation parameters are known the cake deliquoring can be determined for any operational parameters without having to turn to the experiment for each setting. For the calculation of the cake saturation or the cake residual moisture, respectively, (for the conversion see equation (3) ) is by the author the following equation recommended /2/.: S Sr B = (1+ A K ) 1 S r K= Pc (pd pke ) t l hc (hc + hce ) (4)

S(-)=cake saturation, Sr(-)=remanent cake saturation, A(-), B(-)=adjustment parameters, Pc(m2)=cake permeability, pd (Pa)= deliquoring pressure difference, pke (Pa) = , capillary 8

entrance pressure, t(s)=deliquoring time, (-)=cake porosity, l hc(m)=cake height, hce (m)=substitute cake height

(Pas) =liquid viscosity,

The in equation (4) contained adjustment parameters A and B are determined by evaluation of deliquoring tests by regression analysis (see computer program FILOS-Analysis of Filtration Data /3/). The value K is the so called deliquoring index and contains besides the product parameters all three, during filter operation adjustable parameters pressure difference pd, deliquoring time t and cake height hc. With validity of equation (4) and constant parameters A and B would a plotting of the measured cake saturation values with respect to the deliquoring index K result in a single curve. Such a plot is exemplary shown in Fig. 5. The causes for the in practice also often observed deviations from coincidence of all results on one line are explained in /2/. For practical problems, however, where the deliquoring pressure difference is much larger than the capillary entrance pressure and no crack formation occurs, equation (4) is applicable with good accuracy.

v hc (m 3 cm / (m 2 h) )

hc = 5 mm hc = 7.5 mm hc = 10 mm hc = 12.5 mm h = 12.5 mm hc = 15 mm h = 15 mm

c

Zinc Sulfide t = 240 s

( p / 2 + pu ) p (bar2)
Fig. 6: For the evaluation of a cake deliquoring test regarding crack formation and/or border detachment by a special, theory based plotting of the gas throughput measuring values

In addition to the knowledge of the cake residual moisture is the possibility of the calculation of the gas volume flow rate that during the cake deliquoring occurs of great importance first of all with regard to the compressor design and the determination of the economic efficiency of the process. By comparison of the gas consumption of a filter in operation with the theoretically calculated value the optimisation potential of the filter can be demonstrated. Moreover offers the theory based plotting of the measured data of the gas flow rate the possibility to detect and quantify crack formation and border effects during the deliquoring test as is exemplary shown in Fig. 6 for a mineral suspension. 9

The lower on a straight line lying measured data come from the tests with a comparatively lower cake height. There was no crack formation and no border detachment of the cake (see also Fig. 7). The measured data with the tests with comparatively thick cake lie scattered and clearly above the mentioned adjustment straight line as in this case crack formation and/or border detachment occurred. Such a sorting and evaluation of the gas throughput results would not be possible without the theory (see equation. (5) ). Such an application is therefore so important because it can be used as "diagnosis instrument" also for cases where one cannot visually observe the deliquoring process or not recognize occurring cake micro cracks.

Realative Gas Flowrate v* ( - )

Zinc Sulfide

Calculated

Cake Deliquoring Index K ( - )

Fig. 7: Dimensionsless gas flow rate v* = 1 exp ( K) as a function of the deliquoring index K For the calculation of the gas volume flow rate (norm conditions) which during the cake deliquoring is consumed the following equation developed and verified by the analysis of many tests /2/:
pd + pu ) TN 2 ( + ln (pd / bar ) ) 1 e K v= g ( hc + hce ) p N Tg Pc pd (

(5)

v (m3/(m2 s)) = gas volume flow rate related to the filter area, pd (Pa) = Deliquoring pressure difference, pu (Pa) = 105 with pressure filtration and pu = 1 - pd with vacuum filtration, TN (K) = 273, g (Pa s) = gas viscosity, hc (m) = cake height, hce (m) = substitute cake height, pN (Pa) = 105 , Tg (K) = gas temperature, (-), (-), (-) = adjustment parameters, K (-) = deliquoring index (see definition

equation (4) ) As Fig. 7 shows, it is possible, similar as with the cake saturation (see Fig. 5), to represent all gas throughput measuring values with sufficient accuracy by one curve, whereby the dimensionless 10

gas flow rate v* is applied in dependance of the deliquoring index. The value v* results from division of the measured values v with the first two expressions (fraction and round bracket expression) on the right side of equation (5). The coincidence of all measuring data on one curve is not valid, however, when crack formation or cake border detachment occurs.
CONCLUSIONS

The pure empirical approach for testing the filtration behavior of a suspension often results in an unsystematic, time and cost intensive experimental procedure (especially when combined with pilotscale experiments) whereby the obtained information is limited due to the lack of a theory based result evaluation. The theory based planning and carrying out of filtration experiments in combination with an optimized lab-filtration device including user friendly software for recording and evaluating the measured data is seen as a powerful tool with regard to minimization of needed experiments for characterization of the suspension behavior and proper filter design as well as for a maximal information impact for a given experimental effort. By plotting the experimental results on the basis of practical and reliable equations, the filtration characterizing parameters, as for example, cake permeability, cake compressibility, filter medium resistance, can be calculated thus enabling a quantification and a common language regarding the characterization of the filtration behavior of a suspension. Moreover, depending on the deviations of the experimental results compared to that expected according to theory, product related effects as for example suspension break-through at the beginning of filtration or cake cracking during deliquoring can be detected. Also, device related systematic errors as well as human related bad measurements can be seen and be excluded from the analysis.
REFERENCES

/1/ /2/ /3/

VDI Guideline 2762 Filterability of Suspensions: Determination of filter cake resistance Verein Deutscher Ingenieure, Duesseldorf 1997 I. Nicolaou Progress in Theory and Application of Filter Cake Forming and Dewatering VDI Verlag 1999, 280 pages, ISBN 3-18-358303-8, Language: German I. Nicolaou An Innovative Computer Analysis of Filtration Data Proc. of the AFS - 15th Annual Conference, April 9-12, 2002, Galveston, Texas

11