You are on page 1of 13

Journal of Engineering Studies and Research – Volume 16 (2010) No.



INGLEZAKIS J. VASSILIS Chemical Engineer (PhD, MSc), SC European Focus Consulting, Bacau, Romania, Abstract: Fixed bed is the most frequently used operation for ion exchange and adsorption and although highly useful and utilized, its engineering analysis is quite complex. A predictive model for fixed bed systems requires extensive experimental information, complex mathematical tools and in many cases, for complicated wastewater, the full modeling of the process maybe extremely difficult. The situation is even more difficult considering the maldistribution, as well as the scale up procedures. It is true that the overall effectiveness of a fixed bed operation depended mainly on its hydraulic performance. Even if the physicochemical phenomena are well understood and their application in practice is simple, the operation will probably fail if the hydraulic behavior of the reactor is inadequate. These effects could be critical not only in full scale but also in pilot or even in lab scale reactors. In the present study – composed by two parts – short cut methods is proposed as a first design tool. The methods essentially propose the use of a minimum number of experimental data along with some scale-up and hydraulic-performance considerations, extracted from the related literature. In part I of the study the theoretical models and hydraulics of fixed bed operations are presented. In part II the issues of scaleup and approximate design methods are presented. Key words: ion exchange, adsorption, fixed beds, scale-up

1. INTRODUCTION Ion exchange and adsorption are common operations used for the wastewater treatment. For example, in ion exchange processes, zeolites and resins are used for the removal of heavy metals as Pb2+, Cr3+, Fe3+, Cu2+ and ammonia while in adsorption processes activated carbon, peat, iron oxide, kaolin and fly ash is used for the removal of heavy metals and several organic substances from wastewaters, as benzene, phenolic compounds, carbon tetrachloride, nitrophenol, acid dyes and p-dichlorobenzene [1, 2]. Fixed bed is the most frequently used operation for ion exchange and adsorption as is the predominant way of conducting such sorption separations. In the following figures a typical column operation and its breakthrough curve is presented. In Figure 2 (C/Co) is the dimensionless exit concentration and (Veff) is the effluent volume. Breakpoint concentration is a selected limiting value (according to the needs of the treatment) where the operation should be stopped. The main aim when sizing adsorptive columns is the ability to predict the service time (or total effluent volume) until the column effluent exceeds breakpoint concentration. Then, from a practical point of view, breakpoint volume is the most important information: it is the liquid volume that can be treated by the column until a pre-specified breakpoint. This information could be derived from experiments conducted in laboratory columns and the scale up goal is to “transfer” it to large unit or at least to evaluate the changes during scaling.

3 30 1 Co 0. hydraulic design and scaleup critical factors. in the past. which could be related to low liquid holdup in downflow operation. during time (dt) is [2. In the present study a short-cut design scheme is presented for ion exchange and adsorption fixed bed operations. even in the simple cases of Pb2+. or other design problems. Fig. 2: Breakthrough curve. 7. it is found that equilibrium parameters and/or solid diffusion coefficients seem to be unexpectedly influenced by contact time. 2. The effectiveness of a fixed bed operation is depended mainly on its hydraulic performance. geometric analogies and maldistribution) factors. 3. for complicated wastewater. dead spaces and bypasses that can alter completely the performance of a reactor when compared to the “ideal” presentation [5]. The main factor of failure is liquid maldistribution. complex mathematical tools and in many cases. The situation is even difficult considering the scale up difficulties. industrial plant operation is mainly based on experience [5]. in fixed bed operation.4 0. kinetic (diffusion and convection coefficients) and hydraulic (liquid holdup. THEORETICAL FUNDAMENTALS AND BASIC MODELS The continuity equation for the column process around a segment of solid adsorbent between the height (x) and (x+dx) in the bed. We must be able to recognize the competitive effects of kinetics and fluid dynamics: mixing. peat and activated carbon respectively. capacity).2 0 breakpoint volume C breakpoint concentratio Veff Fig. Although the fixed bed mode is highly useful. Cd2+ and Cu2+ and basic dyes removal from solution using clinoptilolite. making the modeling of the operation a complex task [4]. Frequently. Thus a predictive modeling of fixed bed systems requires extensive experimental information (laboratory and pilot scale experiments). These effects could be critical not only in full scale but also in pilot or even in laboratory scale reactors.6 0. For example. Fixed bed operation is influenced by equilibrium (isotherm.Dd ⋅ ε ⋅ 2 ∂t ∂t ∂x ∂x (1) . Even if the physicochemical phenomena are well understood and their application in practice is simple. it is not possible to achieve full similarity between different scales.8 C/Co 0. its analysis is unexpectedly complex [3]. It is obvious then that the need for simple short-cut design procedures is essential at least for the first approach of the operation design. As a matter of fact. even today.Journal of Engineering Studies and Research – Volume 16 (2010) No. the operation will probably fail if the hydraulic behavior of the reactor is not adequate. 8]: ∂ 2C ∂q ∂C ∂C − ρb ⋅ =ε ⋅ + us ⋅ . considering the basic principles and use of predictive modeling. 1: Fixed bed operation. the full modeling of the process is extremely difficult. decisions on full scale reactor design have not always been sufficiently supported by adequate experimental evidences and. it is well known that scale-up is a complicated procedure and in most cases the engineer proceeds using empirical rules by selecting the essential design variables and taking into consideration the applied safety or over-design factors [6]. On the other hand.

(us) is considered constant throughout the reactor. Then. Dividing the second and third term by the bed height (Z) we have: ∂q ∂C u s ∂C D d ⋅ ε ∂ 2C . 5.Journal of Engineering Studies and Research – Volume 16 (2010) No. (us) is the superficial liquid velocity. 7]: ∂q k f ⋅ au = ⋅ (C − C *) ρb ∂t ∂q Ds = ∂t ρ b The mean solid concentration ( q ) is [9. Hence the expression is in the form of a partial differential equation. relates the stationary and mobile phase concentrations at equilibrium [32]. of the form q=f(C*). 10]: ⎛ ⎛ ∂ 2 q 2 ∂q ⎞ ⎜ ∂ 2q Ds ∂q 2 ⎜ ⎟ ⋅ + ⋅ = ⋅⎜ + ⋅ ⎜ ∂r 2 r ∂r ⎟ ρ ⋅ R 2 ⎜ r 2 r ∂ rR ∂ R b ⎝ ⎠ R ⎝ ( ) ( ) ( ) ⎞ (4. Both (C) and (q) are dependent on time (t) and height (Z). Such effects are usually significant only for gaseous systems [8]. For combined resistances The physical process of adsorption or ion exchange is considered to be so fast relative to diffusion steps that in and near the solid particles a local equilibrium exists.2 ⋅ − ρb ⋅ =ε ⋅ + ⋅ 2 ∂t ∂t Z ∂ x Z ∂x Z ( ) ( Z) (2) and rearranging: − ρb ⋅ Z ∂q Z ∂C ∂C D d ⋅ ε ∂ 2C ⋅ =ε ⋅ ⋅ + ⋅ u s ∂t u s ∂t ∂ x us ⋅ Z ∂ x 2 Z Z ( ) ( ) (3) The term (Z/us) is the liquid residence time (or contact time) while the term ( Dd ⋅ ε ) is the inversed Peclet us ⋅ Z number of the bed. The surface equilibrium relationship between solute in solution and on the solid surface can be described by simple analytical equations as the following [8-10]: . 6) should be used simultaneously. (C) is the solution concentration and ( q ) is the mean concentration of solute in the solid phase. 6) for solid diffusion control. However. The rate equation of the adsorption is [3. (Dd) is the axial dispersion coefficient. the so-called adsorption isotherm. Equation (4) holds for liquid film control and equations (5. For liquid film control equations (4. 5) ⎟ ⎟ ⎟ ⎠ q= 3 R ⋅ q ⋅ r 2 ⋅ dr 3 ∫ 0 R (6) Where: (r) is the distance from particle center and (R) is the particle radius. high concentration levels (concentrated solutions) cause significant variations in (us) and this term should be incorporated into the derivative. q =q. For the common case of dilute (low concentration) liquid systems. 3 31 Where: (ε) and (ρb) is bed porosity and bulk density respectively.

Generally. The continuity. an isotherm is favorable if its shape is convex upward and unfavorable if its shape is concave upward in a X v. 1 0.1 0 0 0. La=1 for linear and La>1 for unfavorable.6 0. In Figure 3 the characteristic isotherm shapes are presented. This system consists of 4 equations and 4 unknowns (C.6 0.5 0.9 0.4 X 0. A review of the correlations suggests that the following one gives the best prediction of the individual liquid-film mass transfer coefficient in water treatment situations [12]: .8 0.2 0. q .s. 3 32 Y= X La + (1 − La ) ⋅ X q qmax (7-10) Y = XW Y= C* X = Co The first equation is the Langmuir isotherm and the second the Freundlich isotherm. which is assumed to be in equilibrium with (q). which can be evaluated from several correlations found in the related literature.2 0. rate and equilibrium equations are solved simultaneously using the appropriate initial and boundary conditions [8].3 0. The kinetic parameters The solid diffusion coefficient (Ds) is theoretically independent of the flow type and rate and it can be calculated using appropriate diffusion models in batch reactors.Journal of Engineering Studies and Research – Volume 16 (2010) No. La<1 for favorable.7 0. Basic isotherm types. It should be noted that for adsorption and ion exchange of equal valence ions each isotherm hold for a specified temperature while in ion exchange of different valence ions hold for the same temperature and normality due to the concentrationvalence effect [11]. In the above equations (C*) is the concentration of the solute in the fluid phase. The parameter (La) is also called separation factor and provides a quantitative description of the equilibrium regions: La=0 for irreversible. q and C*).4 0. The empirical constants (La) and (W) are related to the particular system under investigation and are obtained from laboratory experiments [3]. Y figure. Liquid mass transfer coefficient (or convection coefficient) is a flow dependent parameter.8 1 Favorable Unfavorable Linear Rectangular Sigmoidal Y Fig. 3.

a bed of 50 cm height is sufficient. 66 Sc 0 . especially in low velocities. 3 33 1 ⎞ 1 ⎛ 1 + 1 . for superficial velocities higher than about 0. as zeolites and activated carbon. For packing materials of irregular shape. 3. In most of the cases particle diffusion within the solid matrix is the controlling mechanism.2 cm/s [18]. using granular activated carbon [13]: k f = 2 . 5 ⋅ (1 − ε Sh = ⎜ 2 + 0 . (Sh) is the Sherwood number.4 ⋅ u s ⋅ Re 0 . predictive models could be used in order to model the removal process and determine the controlling step of the whole phenomenon [7]. THE USE OF PREDICTIVE MODELS 3. since in low velocities the liquid film diffusion step could be the controlling mechanism. 85 ⋅ (Re ⋅ Sc ) −2 3 u ⎛1 − ε ⎞ ⋅ s ⋅⎜ ⎟ ε ⎝ ε ⎠ −1 3 (13) Finally. (dp) is the particle diameter and (ν) the liquid kinematic viscosity. In order to use simplified models the following assumptions should be met: 1. Plug flow. 644 ⋅ Re 2 ⋅ Sc 3 ⎟ ⋅ [ ⎜ ⎟ ⎝ ⎠ )] (11) Williamson correlation is proposed and used in adsorption processes for wastewater. Mass transfer controlled systems . 16]: 1 . (Sc) is the Schmidt number. This assumption holds only if the bed Peclet number is greater than 100 [17]. for very low Reynolds number (<0. by using upflow operation this value could be by far lower. Generally by using upflow mode the quality of the flow is much better. as a first step the lower possible linear velocity could be used for the evaluation of the controlling step. and 0. In this case the last term in the equation (3) is neglected.Journal of Engineering Studies and Research – Volume 16 (2010) No. 58 (12) Also Kataoka correlation have been used in adsorption systems from liquid phase [14]: k f = 1 . (Df) is the diffusion coefficient of the solute in the liquid phase. 09 1 1 Sh = ε ⋅ Sc 3 ⋅ Re 3 (14) Where: Sh= Re = Sc= kf ⋅dp Df dp ⋅us ν ν Df (15-17) Where: (Re) is the particle Reynolds number. However. Thus. But controlling step is strongly dependent on liquid linear velocity.01) the Wilson-Geankoplis correlation have been used in adsorption systems from liquid phase [15.“Diffusion” models As a second step. .5 mm particle size.1.

However. (ε) is the bed voidage. 3 34 2. Practically. This simplified design model ignores the intraparticle (solid) mass transfer resistance and the external (liquid film) resistance such that the adsorbate is adsorbed onto the solid surface directly.1) = ln (X ) − La ⋅ ln(1 − X ) +1 1 . . The following dimensionless parameters are defined [7]: ρ ⋅q Λ = b max Co ⎛ Voε ⎞ t −⎜ ⎟ ⎜ Q ⎟ ⎠ T= ⎝ ⎛ Λ ⋅ Vo ⎞ ⎜ ⎜ Q ⎟ ⎟ ⎝ ⎠ Nf = Ns = k f ⋅ au ⋅ Vo Q 15 ⋅ Dd ⋅ Λ ⋅ Vo R2 ⋅Q (18-21) Where: (t) is time. 22]: Ce ≅ Co 1 ⎡Θ ⎤ 1 + exp ⎢ ⋅ (qmax ⋅ M − Co ⋅ Vtot )⎥ ⎣Q ⎦ (25) Where: (Ce) and (Co) are the effluent and inlet solute concentrations.5 (22-24) ΨS = 3. This means that the rate of adsorption is controlled by the surface reaction between adsorbate and the unused capacity of the adsorbent [14]. This condition reduces the continuity equation (3) to the simple relation: C/Co= q /qmax. In the following.0. (Q) is the volumetric flow rate.106 ⋅ La 0.2.La 1 ⎡ La ⋅ ln (X ) − ln (1 − X ) ⎤ ⋅ − 1⎥ N s ⋅ (T . Simplified models under the above assumptions have been proposed and analyzed in the related literature and are in the form of either arithmetic or analytical solutions [3. (M) is the total mass of the adsorbent. which should be evaluated from bed data [4]. Also. (Q) is the volumetric flow rate.1) = ΨS ⎢ 1 . Constant pattern condition. The expression by Thomas for an adsorption column is given as follows [21. (qmax) is the maximum adsorption capacity. (ρb) is the bed density and (qmax) is the operating capacity. simplified models will be presented under the two common applied assumptions.14]. 8]. Subscript (f) refers to the fluid and (s) to the solid phase resistance. The following equations are approximate solutions of fixed bed model under the constant pattern and plug flow assumption for favorable Langmuir isotherm: N f ⋅ (T .894 1 .La ⎣ ⎦ 0. (Vtot) is the throughput volume and (Θ) is the Thomas rate constant (units in: volume/mass*time).Journal of Engineering Studies and Research – Volume 16 (2010) No. (Vo) is the total bed volume. Thomas-BDST Model Thomas model is also referred as bed-depth-service-time (BDST) model [3. this model is essentially a constant pattern model [20]. the constant pattern assumption holds if the equilibrium is favorable and at high residence times [7]. The BDST approach is based on the irreversible isotherm model by Bohart and Adams [20]. constant pattern assumption is “weak” if the system exhibits very slow kinetics [19]. namely the constant pattern and plug flow assumptions.

the values of (C/Co)stoich (corresponding to stoichiometric point) for solid diffusion control are between 0. This approximate method requires only the experimental bed data. and specifically.Journal of Engineering Studies and Research – Volume 16 (2010) No.51 to 0. 3. This point is always lower than about (C/Co)=0. regardless of the controlling mechanism (liquid film.31 and 0. According to Perry [7]. Approximate methods for the determination of the controlling mechanism For the determination of the controlling mechanism. the (C/Co) at which N*(T-1) and thus (T-1) equals to zero is called stoichiometric point. Finally. 10]. (qmax) and (Θ) could be flow rate-depended parameters (see also Section 3) [65].5. at least as a first approximation. 1 0.8 0.70 and the more favorable equilibrium leads to higher value of (C/Co)stoich. involving only (Θ). 3 35 Thomas model is also applicable to the design of ion exchange columns [21]. C/Co curve is constructed. The determination of (La) is requiring the whole experimental equilibrium curve and in case of sigmoidal or other non Langmiur or Freundlich-type isotherms these models are non usable.7. In practice.23]. This equation is simple as it can be used in its linear form: ⎞ Θ ⋅ q max ⋅ M Θ ⋅ Co ⎛ Co − ⋅ Vtot ln⎜ ⎟= ⎜ C − 1⎟ Q Q ⎠ ⎝ e (26) Essentially the above model is empirical as in real systems both solid and liquid film resistances play an important role in the adsorption process. The opposite holds for liquid film control where the corresponding (C/Co)stoich values are between 0. if the experimental data are plotted in N*(T-1) vs (C/Co) graph. (Co) the inlet concentration and (T) is the dimensionless time modulus. Then the (T-1) v. The characteristic (C/Co) vs N(T-1) curves are shown in Figure 4. pore or solid diffusion or the combination between them). where (C) is the exit concentration. the model is used for any equilibrium type.2 0 N(T-1)=0 N(T-1) Fig. BDST models are more easily applied in adsorption operations. By this point of view. as (qmax) is an experimentally derived parameter (see also Section 3).4 0. involving (kf) or (Ds) and (La) while BDST models are oneparameter models.6 C/Co 0. Furthermore. the following method could be used [24]. The above method for the determination of controlling step is of course approximate and qualitative and could be used only if the model could not be applied due to experimental or other problems. for both Langmuir and Freundlich isotherms.s. 9. which are frequently used in adsorption and ion exchange systems [7. the set of exit concentration and the corresponding elapsed time for each concentration. and is independent of volumetric flow rate: that is the point where the amount of solute that has “leaked” past the reference point in the bed equals exactly to the residual unfilled capacity of the solid contained before that point. in the case of mass transfer controlled systems. . 4: Characteristic (C/Co) vs N(T-1) curves for solid diffusion control (dotted line) and liquid film diffusion control. An improved BDST model is found elsewhere [20. The Thomas equation constants (qmax) and (Θ) values can be obtained from the column data and can be used in the design of a full scale adsorption bed. “Diffusion” type models are two-parameter models.3.

while for film diffusion control the kinetics will be affected. above which the controlling mechanism is solid diffusion control. (La) is shown in Figure 5. Miura and Hashimoto defined the mechanical parameter (ζ). while if (ζ) infinite (practically much higher than 1) then the solid diffusion is controlling the process rate. one of them being equilibrium isotherms which a time consuming experimental work. In multicomponent solutions the use of models is extremely difficult as numerous data are required. This equation can be used irrespectively of the constant pattern assumption and only if safe data exist for the solid diffusion and liquid mass transfer coefficients. which are expected in large-scale units.65 0. 0.55 Xstoich 0. For the estimation of the controlling mechanism. .6 0. If modeling or other approximate methods are not applicable then a number of experiments should be conducted in order to examine the effect of superficial velocity on the performance of the bed and more specifically on the breakpoint volume.4 La 0. Keeping the same contact time and particle size one can study the effect of linear velocity only changing accordingly the length of the bed and by this way to examine the controlling step: for solid diffusion control there is no effect of linear velocity on the kinetics in the bed. F: liquid film diffusion control). the mechanical parameter represents the ratio of the diffusion resistances (solid and liquid film).45 0. In fact. as follows [25]: kf ⋅ au q ⎞ K ⋅⎛ ⎜ max C ⎟ o ⎠ ⎝ 15 ⋅ D s ⋅ ρ b K= R2 6 ⋅ (1 − ε ) au = dp ζ= (27-29) According to their analysis.s.2 0. From these experiments a lower limit of linear velocity can be found.Journal of Engineering Studies and Research – Volume 16 (2010) No. if (ζ) is zero (practically much lower than 1) then liquid film diffusion is controlling the process rate. Then. 5: Stoichiometric point curves (S: solid diffusion control.35 0.3 0 0. 6 0. it is guaranteed that the controlling step is unaltered. this is a scale up procedure.4 0. Essentially.8 1 F S Fig. 3 36 Using the approximate “diffusion” models (equations 22-24) the (Xstoich) v. The mathematical complexity and/or the need of knowing multi-parameters from separate experiments in all the “diffusion” models make them rather inconvenient for practical use [26]. for higher linear velocities.5 0.

equation of continuity reduces to [7]: dY dF ( X ) =T ⇔ =T dX dX * (30) Where: Y* = F(X) is the equilibrium relationship between Y and X. using the local equilibrium analysis.6<Re<8. rasching rings. 4. Equilibrium limited system In the case of unfavorable equilibrium the local equilibrium analysis could be applied. Finally.g. In the related literature there are some correlations for the evaluation of particle Peclet number (Ped) and liquid holdup (%h.3 mm. For high Reynolds number the following expression hold [3]: .Journal of Engineering Studies and Research – Volume 16 (2010) No. for spherical and other irregular-shaped particles (intalox saddles. 3 37 3. This situation is ideal and does not correspond to real situations. the isotherm of a system can be found from breakthrough experiments using the following equation: 1 Y * = ∫0 T ⋅ dX (31) In the case of favorable equilibrium. Then. as a percentage of the total bed volume) for materials that are frequently used in adsorption and ion exchange systems. physical mechanisms that are sensitive to size are investigated separately from chemical (kinetic or equilibrium) studies [27]. as a vessel e.2-1. berl saddles) particle Peclet number is found to be between 0.65 for upflow and 0. local equilibrium analysis predicts that at T=1 the concentration (X) will rise instantly from 0 to 1 (ideal step change).28 %h = 21 + 99. mass transfer always is controlling the rate. superficial velocity is in (cm/s). the fixed bed is examined in respect to its flow patterns. (v) is the kinematic viscosity of the liquid phase. The limits of validity of eq.05 for downflow. assuming local equilibrium between the liquid and solid phase. the bed Peclet number can be evaluated using the following equation [17]: PeL = Ped ⋅ Z dp (34) Generally. 33) Where. the mass transport step is neglected or it is considered to have a minimal effect on the system evolution when compared to that of the unfavorable equilibrium.52 for upflow and 0. irrespectively of the specific chemical reaction or physical phenomenon that will take place in it. BASIC HYDRAULIC CONSIDERATIONS After the study of the thermodynamics and chemical kinetics of the system the hydraulic behavior of the fixed bed apparatus should be investigated.72 ⋅ u s (32. In this case. which are common in wastewater treatment applications. This could be done by experiments (tracing techniques) or by use of correlations derived from literature. while liquid holdup equation hold for particle size of 1. This investigation basically involves the evaluation of liquid holdup and maldistribution of the flow.48 for downflow and. Essentially. the following equations can be used [18]: Ped = L ⋅ Re k 0. Peclet correlation hold for 0. as when a system exhibits favorable equilibrium. For this kind of materials. as for example zeolites and similar particles of irregular shape [3]. in %h correlation.30 can be found setting X=0 (for T minimum) and X=1 (for T maximum).5.3 and 0.4. (L) is 0. During this step. Here.8 for Reynolds number between 0.01 and 150 [3]. (k) is – 0.

In order to “transfer” experimental data from the small to the large unit. The constant part in liquid holdup correlation (equal to 21) is the static liquid holdup. the part of liquid holdup.2 and 2 mm...72. Certain geometrical analogies should be kept within the following limits in order to avoid large-scale maldistribution of the flow [3]: Z D Z ≥ 5.. particle diameter is presented (u: upflow. for 0. In Figure 6 the maximum linear velocity v. the particle Peclet number is between 0. The minimum fluidization velocity can be evaluated using the following equation [7]: 150 ⋅ μ ⋅ u fm 1 − ε 1.75 ⋅ ρ ⋅ u fm 2 1 ⋅ + ⋅ = g⋅ ρp −ρ Φs ⋅ d p Φs2 ⋅ d p2 ε 3 ε3 ( ) (40) Where: (dp) is the particles diameter. The bed voidage is considered to be the same for different particle sizes.6 and 0.01<Re<10 and is steadily increased for Re>10. In the typical case of ε=0...8 times the same velocity for downflow operation. which is a function of velocity (dynamic liquid holdup). Furthermore. ≥ 12 − 30.. where (ε) is the bed voidage [24]. This correlation is derived using glass beads. (ρp) and (ρ) the density of the particles and of the liquid respectively.5 static holdup is between 14 and 27%. maldistribution and liquid holdup should be kept in a satisfactory level and similar in both scales. aluminum beads and steel beads.. (ε) is the bed voidage. (g) is the gravity acceleration constant (9.81 m2/cm).. according to a literature review presented in [28].48 ( ) (35) This expression hold for 25<Re<320.95 [31].3.4-0.30]. According to Ruthven [8] a maximum linear velocity should not be exceeded in order to avoid extended friction between the packing material in both down and upflow operations. Finally.2 + 0.. .06 and 0. d: downflow).s.8 times the minimum fluidization velocity for upflow operation and 1. For granular particles sphericity is between 0..11/ε) and (0. (Z) is the bed height and (dp) is the particle diameter. which should be used not only in full scale unit but also in the laboratory scale unit. what is essential here is the geometrical analogies. This value is for activated carbon particles. which is true for low dp/D values (see scale up analysis) . which are of similar shape as zeolites [29. For spherical particles is equal to 1..011 ⋅ Re 0. This part is between (0. showing no particular trend.Journal of Engineering Studies and Research – Volume 16 (2010) No. This velocity is 0. for other particle diameter can be evaluated using the following analogy: %h(d1 ) ⎛ d 2 ⎞ ≅⎜ ⎟ ⎟ % h( d 2 ) ⎜ ⎝ d1 ⎠ m (36) Where: (m) is 0. Furthermore. (ufm) the minimum fluidization velocity and (ΦS) the sphericity of the particle.. 3 38 Ped = 1 ε ⋅ 0. ≥ 50 − 150 D dp dp (37-39) Where: (D) is the bed diameter. (μ) the dynamic viscosity of the liquid..072/ε) for particle size between 0.

as it is desirable. for example 0. which is true for low dp/D (see scale up analysis). The maximum linear velocity for this particle size is approximately 0. again because in downflow the permissible maximum linear velocity is higher. the velocity will be about 1.5 cm/s for upflow and 1. ε=0. In laboratory beds this is not a problem as by this way the need of high residence times is satisfied. under upflow conditions. However. it is not possible to work under upflow conditions and downflow is recommended. upflow is avoided. the linear velocity will be increased and probably it will be higher than the maximum linear velocity. particle diameter. Also consider that the adsorption will operate at high contact time. Also.3 cm/s. keeping the same contact time and particle size in both scales. ρ=1 g/cm3. It is obvious that particle size should be chosen considering in the same time the kinetics and the hydraulics of both beds (laboratory and full scale). μ=0. Then it is safe to work.2 d Fig.FS QFS a ⋅ VFS = ≤ umax. which is essential in order to avoid liquid maldistribution and low liquid holdup. for the same contact time.9 cP.5 and Φ=0. 11].5 us (cm/s) 2 1. For safety the 85% of this height could be used.5 1 0. in large beds. . beds with small particles should be operated in low velocities. On the other hand. big particles should be avoided. can be easily used without exceeding the maximum linear velocity.05 0. 3 39 3.65 (for sandtype granular materials).5 3 u 2. especially in upflow operations. Using this maximum working velocity and the contact time during pretreatment/regeneration (which is generally higher than the treatment one) the maximum bed high could be determined as follows: QFS = a ⋅ VFS ⇒ uFS = umax.Journal of Engineering Studies and Research – Volume 16 (2010) No. Furthermore. leading to a safer operation.2 h (5 BV/h relative flow rate). (umax) is the maximum allowable linear velocity and (A) is the bed cross sectional area.5 cm/s for downflow operation. From Figure 6 some important conclusions can be withdrawn. The following typical parameters where used: ρp=2. Thus. another problem could be the increased pressure drop.s. is well known that increasing particle size the adsorption and ion exchange rates are decreased [8. Then.FS ⇒ Z FS ≤ AFS AFS a (41) Where: (a) is the number of bed volumes per hour (BV/h). The bed voidage is considered to be constant. in these beds upflow operation.08 g/cm3. consider a solid of 1 mm particle size. For example. For a laboratory bed of 100 cm height and for the specified contact time the linear velocity is about 0.13 cm/s. First of all.1 d p (cm ) 0.15 0. However for a large fixed bed of 1000 cm height. Apart from the velocity limitations using small particle sizes. which in turn resulted in flooding of the bed in downflow operations. which should be used. subscript (FS) is for full scale bed. 6: Maximum linear velocity v.5 0 0 0.

[3] Inglezakis. Liquid holdup and dispersion in zeolite packed beds.171-189. A. Environmental Pollution. Viraraghavan. [13] Crittenden.. H.Chem.Eng. . S. Pirbazari. Scale up of chemical reactors.Journal of Engineering Studies and Research – Volume 16 (2010) No. H.. Y. REFERENCES [1] Weber. Vermeulen.. Low-cost adsorbents for heavy mrtals uptake from contraminated water: a review. 36. O. [22] Rao. Mashitah. 219-243. 38.. [5] Donati.. Poulopoulos. 2nd Edition.P. [21] Kapoor. [4] Inglezakis. J.. Chemical Engineering Science. John Wiley and Sons... Ind.D. 1972. Adsorption of toxic and carcinogenic compounds from wastewaters. McGraw-Hill.J.J. Timmerhaus.. 10.H. [10] Fleck.. 112. 483-533. Principles of Adsorption and Adsorption Processes. Adsorption . V... International Editions.Chem. L... Biosorption of phenol from an aqueous solution by Aspergillus niger biomass. Jr. Z. M. [12] Chern.. Heavy metals removal in fixed-bed column by the macro fungus Pycnoporus sanguineus. 1984.J. Lykins. C. In any case the lower height derived from the last two equations (41-42) should be used. 2002. 55. [15] Xiu. 2003. D. H. Carbon. 212-223. M.. 463-470. Hand. Journal of Hazardous Materials.J.Res.W. Perry's Chemical Engineer's Handbook. S.. 74-82. [6] Peters. D. Green...Fund. Research on ion exchange by means of the percolation technique. 32. Jan. J. F. J. New York. V.H. 3 40 In the bottom of the bed the particles will be under the pressure not only of the other layers but also under the hydraulic pressure of the liquid...J..P.R. Wat. [9] Hall.W. 61.. [2] Babel. (Mk) is the particle weight and (Fc) is the maximum force that can be applied to a single particle without breaking it (particle strength)..D. Pore. Kurniawan. K. K. The following equation could be used for the evaluation of maximum bed height [24]: Fc ⋅ d p ⋅ ρ b (42) Z≤ g ⋅ M k ⋅ (ρ b + ε ⋅ ρ) Where: (Z) is the bed height. near 50 m [24]. Ind. 1968-1977. W. 5.. Chemosphere. 5049-5057.Eng. Lemonidou. 2006. Modeling of ion exchange of Pb2+ in fixed-beds of clinoptilolite. 647-655. R.Jr. Catalysis Today. 95-99...Sci.. 1984. T. Chemical Reaction Engineering. 2001. 4.. G.Eng. 54. 2002. M..Eng.4-1. R. Grigoropoulou. A. 1998. 85. McKay. (ρb) is the bed bulk density. Optimized correlations for the fixed bed adsorption of metal ions on bone char. 2000. Bhatia. Design considerations for GAC treatment of organic chemicals. 975-981. T. Wiley Eastern Limited. L... 2004. Acrivos.. J.... G. Water Research.. 2001. Viraraghavan.P.D. [18] Inglezakis.G. Hall.. Microporous and Mesoporous Materials.H. 203-209. Bioresource Technology. Arora. 2000.. Chem. 1968. 1973.S. 34.. Ion Exchange.S. B97. 2nd Ed.. [19] Wevers. Wang. 1997.7 mm particle size is very high. Eagleton. [17] Levenspiel. V.. Dover Publications. Chien..A. 1959. R. [11] Helfferich. [8] Ruthven . Paludetto. S.. 56.M. 5819-5829. T. Plant Design and Economics for Chemical Engineers. Chem.D.. M. G. D. Characterization of metal adsorption kinetic properties in batch and fixed-bed reactors....and solid-diffusion kinetics in fixed bed adsorption under constant-pattern conditions..C.165-171.C.. 12. McGraw-Hill International Edition. 1987. Elsevier. Thus a maximum bed height should be determined also by using particle strength data. 1962. P. and Grigoropoulou. Li.. Kirwan. Porter.W. Sci. Journal AWWA. 273-282. [7] Perry.. 1966.. 1982.M.Ion Exchange and Catalysis: Design of Operations and Environmental Applications.Fundam. [14] Zulfadhly. International Edition. Prediction of breakhrough curves for adsorption of lead(II) on activated carbon fibers in a fixed bed..R. J. [20] Ko. For example the maximum bed high for clinoptilolite (a common zeolite species with particle strength of about 2Nt) and 1. T. Journal AWWA... Adsorption of nitrophenol onto activated carbon: isotherms and breakthrough curves. B. 2003.K. 397-404.F. K. D. Removal of heavy metals fro aqueous solutions using immobilized fungal biomass in continuous mode. [16] Chen. Mixed-resistance diffusion kinetics in fixed bed adsorption under constant pattern conditions..

E.F.. A.. Analytical solutions for the breakthrough curves of fixed-bed adsorbers under constant pattern and linear driving force approximations. Wang. 10. Baldi. V. [24] Inglezakis.. V.Journal of Engineering Studies and Research – Volume 16 (2010) No.1289-1295. J.. Chem. P. 2002.I.. 1221-1228. Porter. 2002. J.K. 515-523. Chem.C.. K. [25] Miura.C... McKay.. 1990. [26] Juang. [31] McCabe. G. 53. PhD Thesis. Journal of Chemical Technology and Biotechnology. [28] Chung. T. A. Unit Operations of Chemical Engineering. 31..S.. Harriott. D. Solid-Liquid contacting effectiveness in trickle bed reactors. 1977. [29] Specchia. Pressure drop and liquid holdup for two-phase concurrent flow in packed beds. National Technical University of Athens.Sci.Eng. Longitudinal dispersion of liquid flowing through fixed and fluidized beds.. 23-31.. V.Sci. Reactor scaleup methodology... 3 41 [23] Ko. Hashimoto... Wen. R. W.Eng..J. 1977.Y. C. 5th Edition. Lin. K. S. Chemosphere. 1968.Y.H. Baldi.. Feb. G. [27] Trambouze.F.Ch. Smith. 2003... G. [30] Colombo. 857-866. P. 1101-1108. 14. Chemical Engineering Progress.C. . Lee. Journal. S.J. 1976. McGraw-Hill. 1983. Design of Ion Exchange Column Using Natural Minerals as Packing Materials. Greece. Improved design models for fixed bed adsorption of acid dye and zinc ions from effluents. 32.K... Athens. Journal of Chemical Engineering of Japan. 490-493. 77. Removal of metal ions from the complexed solutions in fixed bed using a strong-acid ion exchange resin.