Electrodialytic recovery of some fermentation products from model solutions: techno-economic feasibility study

Journal of Membrane Science 164 (2000) 129–140

Electrodialytic recovery of some fermentation products from model solutions: techno-economic feasibility study Mauro Moresi ∗ , Fabiana Sappino Istituto di Tecnologie Agroalimentari, University of Tuscia, Via S. C. de Lellis, I-01100 Viterbo, Italy Received 18 February 1999; received in revised form 8 April 1999; accepted 20 May 1999

Abstract In this work, the technical feasibility of the batch electrodialytic (ED) recovery of trisodium citrate from aqueous solutions was compared to that of sodium itaconate and lactate by assessing the effect of electric current density (j) on the main ED performance parameters (i.e. solute recovery, ρ, and Faraday, η, efficiencies; specific energy consumption, solute and water fluxes). In particular, ρ and η were found to be practically independent of j. However, while ρ was almost unitary for all salts tested, η increased from 0.52 for trisodium citrate to 0.61–0.62 for the other two salts. Solute and water fluxes were proportional to j, whereas specific energy consumption () was a power-law function of j. Then, a study-grade estimation of their specific ED recovery costs was carried out, thus showing that sodium lactate recovery is by far much less expensive than that of trisodium citrate or disodium itaconate. ©2000 Elsevier Science B.V. All rights reserved. Keywords: Electrodialysis; Citric acid; Itaconic acid; Lactic acid; Electric current density; Economic feasibility; Solute recovery efficiency; Faraday efficiency; Specific energy consumption; Solute flux; Water flux; Parameter sensitivity analysis

1. Introduction Electrodialysis (ED) is a unit operation for the separation or concentration of ions in solutions based upon their selective electromigration through semi-permeable membranes [1–2]. Its largest area of application is in the desalination of brackish water for the production of potable water [2–3] and de-ashing of milk whey to obtain a valuable raw material for baby foods [4]. In biotechnology, ED was used to remove continuously lactic [5] and itaconic [6] acids from fermentation broths, in order to enhance microbial acidic pro∗ Corresponding author. Tel.: +390-761-357494; fax: +390-761-357498. E-mail address: [email protected] (M. Moresi).

ductivity and make their final processing easier. This technique was also suggested as an environmentally friendly alternative to the conventional citric acid recovery process, that gives rise to enormous amounts of calcium sulphate (ca. 2 kg of CaSO4 ·2H2 O per kg of monohydrated citric acid) that are to be disposed of. To avoid such solid waste formation and alternatively for citric acid extraction from fermentation broths using tertiary aliphatic amines as actually applied by Pfizer Inc. in Europe and Haarman and Reimer Corp. in the USA, citric acid was recovered by ED using either bipolar [7] or monopolar [8–10] membranes. In previous work, batch electrodialytic recovery of citric acid from model solutions was studied by varying the feed solute concentration (cSi ), pH, temperature (T) and electric current density (j) [11] and then optimised by balancing the investment and mainte-

0376-7388/00/$ – see front matter ©2000 Elsevier Science B.V. All rights reserved. PII: S 0 3 7 6 - 7 3 8 8 ( 9 9 ) 0 0 1 8 6 - 6

130

M. Moresi, F. Sappino / Journal of Membrane Science 164 (2000) 129–140

nance costs against energy costs [12]. More specifically, citric acid was found to be much more conveniently recovered from clarified fermentation broths once converted into the trisodium salt. The main aim of this work was to compare the economic feasibility of the batch electrodialytic recovery of trisodium citrate from aqueous model solutions to that of two other well-known fermentation products, such as the sodium salts of lactic and itaconic acids.

2. Experimental A laboratory-scale electrodialyser (Aqualyzer P1, Corning EIVS, Le Vesinet, F), previously described [10–11], was used. This apparatus was equipped with two graphite electrodes and a sheet-flow stack containing 40 cation- and 40 anion-exchange membranes, separated by spacer gaskets which ensured a flow parallel to the membranes themselves. Their main characteristics are given in Table 1. The direct current (DC) generator could supply voltage (φ) and current (I) in the following ranges: 0–60 V and 0–2.5 A. The dilute (D), concentrated (C) and electrode rinsing solutions were stocked in three PVC tanks and re-circulated through the electrodialysis stack by means of three polypropylene centrifugal pumps with a nominal capacity of 250 dm3 h−1 and a total discharge head of 2.5 m of water. Electrode rinsing was carried out by using a NaCl solution (10 g dm−3 ) at an initial pH value of 6.5 to assure adequate electric conductivity in the electrode rinsing channels, as suggested by the rig’s supplier. All batch recycle runs were carried out under constant re-circulation flow rate (170 dm3 h−1 ), equivalent to a channel superficial velocity of 2.5 cm. s−1 , by varying the initial feed solute concentration (cSi ) and current density (j) in the ranges 50–100 g dm−3 , 110–145 A. m−2 , respectively. Each run was continued for as long as reduction of the solute content in the diluting (D) compartment at less than 0.5 g dm−3 was taking place to allow the diluting stream to be potentially reused in the preparation of the fermentation medium. The feed solutions were prepared by dissolving technical grade citric and itaconic acids in or diluting a 90% w/w lactic acid solution with de-ionised water, so as to obtain the aforementioned cSi values.

By adding analytical grade NaOH, the pH of any acidic solution was increased up to a value greater than the corresponding pKn at 25◦ C, the degree of dissociation n being equal to 1 for lactate (pK1 = 3.86) [13], 2 for itaconate (pK2 = 5.45) [14] and 3 for citrate (pK3 = 6.396) [15]. The specific electric conductivity and refractive index at 20◦ C of all the feed solutions used were determined by using, respectively, a conductivity meter mod. 3420 (Jenway, DunMow, Essex, UK) and an Abbe refractometer mod. 1T equipped with a sodium lamp. Since for any solution tested their refractive indexes were found to be linearly related to solute concentration with correlation coefficients greater than 0.999, the instantaneous concentrations of solute in the diluting (D) and concentrating (C) streams were estimated by refractometry. The ED stack was routinely cleaned at ca. 20◦ C by performing a series of re-circulation cycles of 20 min each with neutral (de-ionised water), acidic (HCl at pH 1.0), and alkaline (3 g/l NaOH) solutions, as described before [10].

3. Results and discussion 3.1. Overall performance of solute recovery by ED To assess whether diffusion-limited conditions were achieved in the diluting compartment as the electrodialytic process was prolonged up to recovering almost any of the solutes initially charged in the compartment D, the limiting current density test was performed as described before [10–11]. In all cases, by using the Cowan and Brown’s method [16] it was impossible to assess the limiting current density, as also observed by Yen and Cheryan [17] and Novalic et al. [7] during the ED of lactic and citric acid solutions, respectively. The typical time course of the electrodialytic recovery of the three electrolytes at 33◦ C and j = 145 A m−2 is shown in Fig. 1. The feed solutions contained 100 g dm−3 of citric and lactic acids, but just 50 g dm−3 of itaconic acid to avoid sodium itaconate precipitation onto the anionic membranes in the concentrating compartments due to the smaller solubility of such a salt with respect to sodium citrate

M. Moresi, F. Sappino / Journal of Membrane Science 164 (2000) 129–140

131

Table 1 Specifications of the electrodialytic stack used (AQUALYZER P1, Corning EIVS, Le Vesinet, France) and manufacturer’s data on membrane properties Membrane type Thickness Temperature range allowance pH range allowance Ion-exchange capacity Electric resistance at 0.5 M NaCl and 25◦ C Perselectivity 1.0/0.5 M KCl Membrane overall size Membrane effective size (am ) Number of cell pairs Overall membrane surface area (Am ) Intermembrane channel

Anionic

Cationic

AMV 0.14

CMV 0.15 0–35 1–9

1.9 2–4.5 92

and lactate. Recovery of lactate, itaconate and citrate was practically brought to an end within about 1.7, 1.0 and 2.6 h. This corresponded to solute accumulation rates (defined as the derivate of solute concentration in compartment C with respect to time, dcC /dt) of about 54, 31 and 29 kg m−3 h−1 , respectively. However, by accounting for the different equivalent masses of the solutes examined, their equivalent accumulation rates exhibited a more restricted range of variation, that is 0.48, 0.36 and 0.34 kequiv. m−3 h−1 , respectively. As the concentration in compartment D was reduced, the electric resistance of the stack increased and, therefore, the direct current generator had to increase automatically the voltage (φ) applied to the electrodes up to the maximum value of 60 V to keep j = const. From this point onward the current density began to decrease (Fig. 2). By installing two graduated Plexiglas tubes (i.d. of 3 cm and height of 2 m) on the top cover of the concentrate (C) and diluate (D) tanks, it was possible to monitor the volume increase in C and the volume decrease in D as a function of time, as reported in Fig. 3. The liquid level in both tanks was not equal at the beginning of the ED process, but it was greater in the diluting compartment (thus involving a higher initial volume VDo ≈ 2.7–3.0 dm3 ) and smaller in the concentrating one (VCo ≈ 2.1–2.3 dm3 ) to allow its easy monitoring throughout any trial. All data collected during each batch recycle run were used to calculate the overall performance indicators listed in Table 2, where any symbol is reported in the Section 5. In particular, despite the volume varia-

2.4 2.9 95 125 × 125 138 40 0.55 0.4

Unit (mm) (◦ C) (–) (meq/g) ( cm2 ) (%) (mm × mm) (cm2 ) (–) (m2 ) (mm)

tion in C (1VC ) was found to be slightly smaller than in D (1VD ) in all cases studied, their difference was found to be statistically insignificant at the 95% confidence level and their absolute values (|1Vi |) were averaged as indicated in Table 1. Fig. 4 shows the effect of current density (j) on the main ED performance parameters for the three solutes considered, while Table 2 lists their corresponding empirical correlations. For any solute mass recovery (ρ) and Faraday (η) efficiencies were found to be independent of j (Fig. 4a). Moreover, the average η values (ca. 61%) for lactate and itaconate were greater than that for citrate (52%), probably because of their greater ionic mobility through the anion-exchange membranes. In accordance with Faraday’s law, any solute flux (JS ) was proportional to j (Fig. 4b). Among the regression equations established here that regarding lactate flux prediction allowed a fairly good estimation (0.45 kg m−2 h−1 ) of the lactate flux (0.40 ± 0.05 kg m−2 h−1 ) measured by Yen and Cheryan [17] in another electrodialyzer (equipped with a tortuous-flow-type stack containing 20 cationand 20 anion-exchange membranes having an effective area of 230 cm2 ) when operating at a current density of ca. 174 A m−2 . Regardless of the solute used, specific energy consumption () was found to be a power-law function of j (Table 2). Since the energy consumed is proportional to the square of the electric current (I) times the overall electric resistance (R) of the ED stack while the amount of solute recovered is proportional to j, 

132

M. Moresi, F. Sappino / Journal of Membrane Science 164 (2000) 129–140

Fig. 1. Time course of the electrodialytic recovery of sodium citrate (a:䊏,䊐), itaconate (b:䊉,䊊) and lactate (c:N,4) from aqueous solutions at 33◦ C and j = 145 A m−2 : Solute concentrations in the concentrating (cC : closed symbols) and diluting (cD : open symbols) streams as a function of time (t).

Fig. 2. Time course of the electrodialytic recovery of sodium citrate (a), itaconate (b) and lactate (c) from aqueous solutions at 33◦ C and j = 145 A m−2 : Voltage (φ:– –) and electric current density (j:——) vs. time (t).

depends on j and the average value of R during the separation process. For the sake of simplicity, such average R value was regarded as a power function of the current density only, thus explaining the empirical model ( ∝ jν ) listed in Table 2 and used to fit the experimental  data against the current density (j), as shown in Fig. 4c. In accordance with Audinos [3], the water flux results from the transport of the hydration water bound

to ions (which is called electro-osmosis) and from the osmotic flux caused by the difference in solute concentrations across the membranes. In previous work [11], the osmosis contribution was measured at 28◦ C by using the procedure described by Yen and Cheryan [17]. By charging the concentrate and diluate tanks with 150 g dm−3 of disodium citrate and de-mineralised water, respectively, and running the system without

M. Moresi, F. Sappino / Journal of Membrane Science 164 (2000) 129–140

133

Similarly, in Yen and Cheryan’s electrodialytic experiments on model lactic acid solutions [17] using the aforementioned tortuous path-type stack, water flux was found to be linearly related to j. However, its regression coefficient (equal to 0.0172 ± 0.0002 dm3 A−1 h−1 ) was about double of that (0.0083 ± 0.0003 dm3 A−1 h−1 ) estimated here. In brief, the whole set of regressions listed in Table 2 represent the ‘best estimates’ of the performance parameters for the electrodialytic recovery of the three solutes under study and will be used to evaluate their economic feasibility, as reported below. 3.2. Estimation of the ED operating costs In order to evaluate the economic feasibility of the electrodialytic recovery of the aforementioned metabolites from clarified fermentation broths, the overall operating costs of the downstream processing plant shown in Fig. 5 were estimated on the assumption that its overall capacity (Qp ) varied from 250 to 6000 Mg (metric tons) of free acid recovered per year. The first unit of such a plant allows the pH of the clarified broth (its free acid concentration being equal to cSi ) to be adjusted to about 5, 6 or 7 by adding approximately 1, 2 or 3 moles of NaOH per mol of lactic, itaconic or citric acid in accordance with the following reaction: Hn R −n + nNaOH → Nan R −n + nH2 O

(1)

Thus, the solute concentration (cSe ) in the pH-controlled solution will be: cse = Fig. 3. Time course of the electrodialytic recovery of sodium citrate (a:䊏,䊐), itaconate (b:䊉,䊊) and lactate (c:N,4) from aqueous solutions at 33◦ C and j = 145 A m−2 : Variation of the volumes in the concentrated (1VC : closed symbols) and diluted (1VD : open symbols) compartments against time (t).

any electric current being applied, the water transport due to osmosis was as low as 0.036 dm3 h−1 [11]. This allowed the osmotic water transfer to be neglected with respect to the electro-osmotic one, which, being proportional to the solute flux, is proportional to the current density (j), as shown in Fig. 4d.

Mfs cSi Mac

(2)

where Mfs and Mac are the molecular masses of the final salt and its reference acid, respectively. If both tanks C and D of the electrodialyzer (ED) are charged with the same volume (Vi ) of clarified broth, the water transport associated with the ionic flux through the ED membranes will increase the final volume (Vo ) of the concentrated stream, thus limiting the final solute concentration (cSo ) in the concentrate. Both these variables can be estimated as reported previously [10]: cSo =

2ρ cSe 1 + [Jw (2ρ − 1)cSe /JS ]

(3)

134

M. Moresi, F. Sappino / Journal of Membrane Science 164 (2000) 129–140

Table 2 Electrodialytic batch recycle recovery of a few sodium salts of lactic, itaconic and citric acids from aqueous solutions at 33◦ C: Definition of the overall performance indicators (solute recovery efficiency, ρ; Faraday efficiency, η; specific energy consumption, ⑀; solute flux, JS , and mean water flux, JW ) used to describe each run and their empirical correlations to current density (j), expressed in A/m2 Indicator

Definition

Solute recovery

final solute mass in C initial solute mass in C and D

efficiency (ρ) Faraday Effi-

Equation ρ=

ciency (η)

effective solute mass recovered theoretical solute mass transported η =

Specific Energy Consumption ()

Energy consumed per unit solute recovered

ε=

Solute Flux (JS )

Solute collected into C (1mS ) per unit cell pair surface area (Am ) and time (1t) Average of the absolute volume variation due to water transfer into the ith comp. (C or D) (|1Vi|) per unit cell pair surface area and time

Js =

Water Flux (JW )

  JW Vo = Vi 1 + (2ρ − 1)cSe JS

Lactate

Itaconate

Citrate

Unit

mCo +1ms mCo +mDo

1.06 ± 0.03

1.02 ± 0.04

0.99 ± 0.06

dimensionless

F (1ms /ME ) R N 01t I (t)dt

0.62 ± 0.07

0.61 ± 0.05

0.52 ± 0.05

dimensionless

0.0306 × j0.48 (r2 = 0.70)

0.00016 × j1.65 0.0039 × j (r2 = 0.72) (r2 = 0.94)

kWh kg−1

0.0026 × j (r2 = 0.83)

0.0020 × j (r2 = 0.85)

0.0016 × j (r2 = 0.69)

kg m−2 h−1

0.0083 × j (r2 = 0.89)

0.0067 × j (r2 = 0.96)

0.0052 × j (r2 = 0.99)

dm3 m−2 h−1

R 1t 0

φ(t)I (t)dt 1ms

1ms Am 1t

Jwi =

|1Vi | Am 1t

(4)

with Vi =

1 Qp 2 ρcSe

(5)

where JW and JS are, respectively, the water and solute fluxes, while ρ is the electrodialytic solute recovery efficiency. In accordance with Table 2, ρ was assumed as unitary for sodium lactate and itaconate and as equal to 0.99 for sodium citrate. The lactate-, itaconate- or citrate-rich concentrates will be further de-watered in a multiple effect evaporator up to about 88 [13], 37 [18] or 64 [18]% w/w, respectively, depending of their respective solubility. While lactate-concentrated solutions may be used as such [13], the other ones are fed to the corresponding crystallisation unit [18]. By referring to a sheet flow-type ED recovery unit, the overall effective cell pair surface area (Am ) required for a given plant capacity (Qp ) can be estimated as follows: Am = ξ

[(106 /24)(Qp /τ )] = N am Js

(6)

where Qp is the amount of free acid annually removed from the given feed solution, τ ( = 330 days/year), the overall working period of the recovery plant, JS the solute flux — that can be estimated via the empirical correlations listed in Table 2, ξ ( = 1.1) a correction factor for safety overdesign of the membrane surface area required, N the overall number of cell pairs and am the effective surface area per each cell pair. Since each cell pair consists of one anion- and one cation-exchange membranes, the overall membrane surface area installed is therefore twice Am . The investment costs (CI ) of the ED unit were supplied by Largeteau (Eurodia Industrie, Wissous, F, 1996) and made up of three items, such as equipment (Ce ), ED plate-and-frame (Cpf ), and membrane (Cm ) costs: C I = Ce + Cpf + Cm

(7)

where Cm accounted for all the membrane sheets installed and their associated spacers, Cpf included the plate-and-frame assembly to tighten the ED stack and two graphite electrodes, and Ce included tanks, centrifugal pumps, piping and valves, DC power generator, instruments, insulation, civil work, electrical,

M. Moresi, F. Sappino / Journal of Membrane Science 164 (2000) 129–140

Fig. 4. Effect of current density (j) on the main performance parameters of the electrodialytic recovery of sodium salts of lactic (N,4 ,——), itaconic (䊊,䊉,– –) and citric (䊐,䊏— . — .) acids from aqueous solutions at 33◦ C: (a) ρ (closed symbols) and η (open symbols) efficiencies vs. j; (b) JS vs. j; (c)  against j; (d) JWi in the C (closed symbols) and D (open symbols) compartments vs. j. The continuous, broken and dash–dot lines were calculated by using the correlations listed in Table 2.

135

Fig. 4 (Continued).

installation, etc. All these costs were found to be linearly correlated to the overall effective cell pair surface area (Am ) [12], as summarised in Table 3. The operating costs (Co ) of the ED unit in Fig. 5 accounted for the investment-related costs (CIo ), such as depreciation (Cd ) and maintenance (Cmain ), utility (CUo ) and labour (CLo ) costs: Co = Cd + Cmain + CUo + CLo

(8)

The first item was made up of depreciation and maintenance. The former was estimated over a 7-year amortisation period (nD ), while the latter included the general plant maintenance (assumed as being equal to about 3% of the investment costs) and annual replacement of a fraction of all electromembranes installed provided that their life span is of the order of nm years.

Fig. 5. Schematic flow-sheet diagram of the downstream processing plant used to recover electrodialytically the sodium salts of the acidic metabolites tested from clarified fermentation broths and feed the evaporation unit.

The utility costs included the electric energy required by the electrodialytic process only, being the energy necessary to pump the solution through the stack negligible. Finally, the labour costs was evaluated on the

136

M. Moresi, F. Sappino / Journal of Membrane Science 164 (2000) 129–140

Table 3 Main correlations and equations used to estimate the investment (CI ) and operating (Co ) costs for the ED unit shown in Fig. 5. The empirical correlations were derived from several cost data supplied by Mr. D. Largeteau (Eurodia Industrie, Wissous, F, pers. commun.) with reference to industrial ED plants with overall effective cell pair surface areas (Am ) ranging from 100 to 1000 m2a COST ITEMS Investment costs (CI ) Equipment costs Membrane costs ED plate-and-frame costs

Operating costs (Co ) Depreciation costs Maintenance costs Utility costs Labour costs a

Equation

Unit

Ce = α + β A m Cm = γ Am Cpf =δ Am α = 158,000 ± 30,000 β = 718 ± 46 γ = 632 ± 5 δ = 279 ± 0

Euro Euro Euro Euro Euro.m−2 Euro.m−2 Euro.m−2

Cd = ‘CI /nD Cmain = 0.03 CI + Cm /nm CUo = Qp cee CL = 46,500

Euro/year Euro/year Euro/year Euro/year

1 Euro = 1936.27 Italian Liras

basis of 1/2 skilled worker per shift whatever the number of stacks of the ED unit on 3 (plus a replacement) shifts per day at a yearly overall rate of Euro23,250 per SW, and were equivalent to Euro46,500/year, as reported in Table 3. Once noted that the specific energy consumption () is a power function of the current density applied (j), while the solute flux (JS ) linearly varies with j (as listed in Table 2):  = ψ jν

(9)

JS = ϕj

(10)

it was possible to combine the regression equations listed in Table 2 into the expressions used to evaluate the single items of the overall operating costs (see Table 3), thus yielding the operating costs per kg of any solute recovered (co ): co = Co /Qp = A/Qp + B /j + ψcee j ν

(11)

with A = CLo + α(1/nD + 0.03) B=ξ

(1/nD + 0.03)(β + γ + δ) + γ /nm τϕ

(12) (13)

The specific recovery costs (co ) are dependent on plant capacity (Qp ), current density (j) and specific electric energy cost (cee ), labour costs being negligible with respect to the other terms of the operating costs. They may be therefore minimised by setting all its partial derivatives with respect to the above independent variables to zero, as follows: B A ∂ co ∂ co = − 2 + ν.ψcee j ν−1 = 0; = − 2 = 0; ∂ Qp ∂ Qp j ∂ co = ψ jν = 0 ∂ cee

(14)

More specifically, the partial derivative of co with respect to Qp is a monotonically decreasing function of Qp itself that nullifies for Qp → ∞, this being in agreement with the general law of economy of scale. The partial derivative of co with respect to cee is a power function of j and, provided that ν is greater than zero as listed in Table 3, leads to a trivial optimum point for j tending to zero, upon which the specific recovery costs are, of course, minimum when a zero-current intensity is circulating through the ED stack. Only by setting the partial derivative of co with respect to j to zero, an optimal value for j can be estimated:  1/ν+1 B (15) jopt = ν ψ cee For the specific electric energy costs currently charged in Italy (cEuro3–10/kWh), the optimal conditions for sodium lactate and citrate were generally connected with equal or greater current densities than the maximum allowable current density (350 A/m2 ) for the ion-exchange membranes used here and this clearly points out the greater contribution of the investment-related costs to the overall operating costs. On the contrary, the optimal current density for sodium itaconate varied from 320 to 200 A/m2 for cee ranging in the above interval. Obviously, the greater the cee the smaller jopt becomes to minimise the contribution of energetic costs to co . Under the assumption of operating at the above optimal values of j (provided that they do not exceed the maximum allowable one) and cee = cEuro5/kWh (which is a typical figure for co-generation plants where high-pressure steam is used to generate the necessary plant power and the exhaust steam for processing and heating), it was possible to evalu-

M. Moresi, F. Sappino / Journal of Membrane Science 164 (2000) 129–140

137

Table 4 Main process variables, investment costs (CI ), overall (Co ) and specific (co ) operating costs for the ED unit shown in Fig. 5 (size capacity, Qp = 2500 Mg/year; working period τ = 330 days/year; cee = cEuro5/kWh) ED Operating variables and parameters Optimal current density (jopt ) Solute conc.n initial (cSi ) Solute conc.n in the conc.ed stream (cSo ) pH Solute recovery efficiency (ρ) Solute flux (JS ) – Water flux (JW ) Specific energy consumption () Overall membrane surface (Am ) Investment costs (CI )

Citrate 350 50 110 7 0.99 0.56 1.82 1.37 620

Itaconate

Lactate

Unit

260 50 109 6 1.00 0.52 1.74 1.54 668

350 50 104 5 1.00 0.91 2.91 0.51 382

A.m−2 g.dm−3 g.dm−3 – – kg m−2 h−1 dm3 m−2 h−1 kWh kg−1 m2

1168

1246

780

kEuro

Depreciation costs (Cd ) Maintenance costs (Cmain ) Utility costs (CUo ) Labour costs (CLo )

167 166 171 46

178 178 194 46

111 104 64 46

kEuro/year kEuro/year kEuro/year kEuro/year

Overall operating costs (Co )

550

596

325

kEuro/year

– – – –

Specific solute recovery costs (co )

0.22

0.24

0.13

Euro/kg

was found to be less expensive than that of the other two sodium salts, since for cee = cEuro5/kWh the ED separation of the sodium salts of lactic, itaconic and citric acids would cost Euro0.13 kg, Euro0.24, or Euro0.22 kg, respectively. Therefore, since the current market prices of anhydrous lactic, itaconic and citric acids are of the order of Euro1.9, 2.0 and 1.1 per kg, respectively, the relative contribution of their ED separation costs varied from about 7% for lactate to 12% for itaconate and to 20% for citrate. Fig. 6. Effect of ED plant size (Qp ) on the overall recovery costs (co ) of sodium citrate (䊐), itaconate (䊊) and lactate (1) under the following assumptions: cee = cEuro5/kWh; j = 260 A/m2 for disodium itaconate and 350 A/m2 for monosodium lactate and trisodium citrate.

ate the effect of plant size (Qp ) on co , as plotted in Fig. 6. Such a diagram shows that the decrease in co drastically slowed down for Qp greater than 1500 Mg/year. As an example, Table 4 lists all the main process variables and cost items of an ED unit with a size of 2500 Mg/year, fed with aqueous solutions containing 50 g dm−3 of sodium citrate, itaconate and lactate. In particular, the ED recovery of sodium lactate

3.3. Cost sensitivity analysis As reported above, the specific ED recovery costs (co ) of sodium salts examined here depend on several independent design and cost parameters, such as plant size (Qp ), overall working period (τ ), optimal current density (j), depreciation period (nD ), electromembrane life span (nm ), specific cell pair area cost (γ ), specific electric energy cost (cee ), and specific labour costs (CL ). However, only a few of them (viz. cee , j, γ and τ ) were found to exert a significant effect on the recovery costs of sodium citrate, as revealed by a parameter sensitivity analysis previously carried out [12].

138

M. Moresi, F. Sappino / Journal of Membrane Science 164 (2000) 129–140

Once the working period (τ ) of the plants under study was set to as high a value as 330 days per year, the effects of cee , j, and γ on co were assessed by performing the following procedure: 1. Obtain a first estimate of the specific recovery costs (co ) by assuming an initial set of independent design and cost parameters, that is: plant size (Qp = 2500 Mg/year), overall working period (τ = 330 days/year), optimal current density (jopt ≤ 350 A m−2 ), depreciation period (nD = 7 year), electromembrane life span (nm = 3 year), specific cell pair area cost (γ = Euro630/m2 ), specific electric energy cost (cee = cEuro5/kWh), and specific labour costs (CL = Euro23 250/year/SW); 2. Vary the values of cee , j, and γ one by one by, e.g., 10 and 20% and register their influence on co . Fig. 7 shows that both γ and cee had a positive effect on co , while j a negative one. More specifically, a 20% variation of γ or cee caused increments of 7–7.5% or 4–6.6% for co , respectively. The effect of j on co tended to amplify as the current density applied was different from the optimal one, as determined via Eq. (15). In fact, the variation of co for a 20% decrease in j was of the order of about +12.4% for sodium lactate, +7.7% for trisodium citrate, and +3.5% for disodium itaconate, being the reference j values used for sodium lactate and sodium itaconate quite less than or coincident with the corresponding jopt ones, respectively. It is, however, worthwhile pointing out that the maximum allowable current density for the monopolar membranes used was 350 A/m2 , thus making any higher j value purely theoretical. For all solutes tested at j = jopt ≤ 350 A m−2 , the specific operating costs were mainly controlled by the investment-related costs, such as depreciation and plant maintenance costs.

Fig. 7. Relative variation of the overall recovery costs (co ) of sodium citrate (a), itaconate (b) and lactate (c) as a function of the relative variation of a few independent design and cost parameters (Xi : j, 䊏 ; γ , 䉬; cee N).

4. Conclusions Although ED recovery for all the solutes studied appeared to be technically feasible, its economics resulted to be less expensive for lactate anions than for citrate or itaconate ions. However, this technique is highly likely to receive further attention in the short–medium term period since it might avoid

the great pollution problems caused by the disposal of the enormous amounts of calcium sulphate produced by the traditional citrate recovery process which is most largely used in the industrial-scale or simplify the complex sequence of operations necessary to obtain white crystals of itaconic acid [18].

M. Moresi, F. Sappino / Journal of Membrane Science 164 (2000) 129–140

In conclusion, any attempt to enhance ion mobility through the ED membranes or to extend their service life would minimise the overall surface of ED membranes to be installed (thus reducing plant investment and maintenance costs) and therefore would foster ED applications in the food biotechnological sector.

5. List of symbols A Am am B Cd Ce CI CIo CL CLo Cm Cmain Co Cpf CUo cee ci co cSe cSi cSo F I Ilim JS JW j jopt

coefficient defined by Eq. (12) (Euro) overall effective cell pair surface area (m2 ) effective surface area per each cell pair (m2 ) coefficient defined by Eq. (13) (Euro.A.m−2 kg−1 ) depreciation costs (Euro/year) equipment costs (Euro) investment costs of the ED unit (Euro) investment-related costs (Euro/year) skilled worker overall annual rate (Euro.SW−1 year−1 ) labour costs (Euro/year) membrane costs (Euro) maintenance costs (Euro/year) overall operating costs (Euro/year) ED plate-and-frame costs (Euro) utility costs (Euro/year) specific electric energy cost (Euro/kWh) solute concentration in the ith compartment (g/dm3 ) operating costs per unit mass of solute recovered (Euro/kg) solute concentration in the pH-controlled solution (g/dm3 ) inlet feed concentration (g/dm3 ) solute concentration in the concentrate (g/dm3 ) Faraday’s constant (96,486 Coulomb/g-equiv.) electric current (A) limiting electric current (A) solute flux (kg.m−2 h−1 ) water flux (dm3 .m−2 h−1 ) current density (A/m2 ) optimal current density as estimated by Eq. (15) (A/m2 )

ME Mi mio N n nD nm pKn Qp T t Vi Vo Xi

139

solute equivalent mass (g/g-equiv.) molecular mass of the ith generic component initial solute mass in the generic ith compartment (kg) number of cell pairs (dimensionless) degree of dissociation (dimensionless) depreciation period (year) average electromembrane life span (year) cologarithm of the nth dissociation constant (dimensionless) plant size (Mg/year) temperature (◦ C) process time (s) initial volume in the generic ith compartment (dm3 ) final volume of the concentrated stream (dm3 ) generic ith design parameter

5.1. Greek symbols α β,δ,γ 1co 1mS 1t 1Vi 1Xi  η ν φ ϕ ␺

ρ τ

empirical coefficient in Table 3 (Euro) specific cell pair area cost (Euro/m2 ) variation of the specific recovery costs (Euro/kg) solute collected in compartment C (kg) overall duration of the electrodialytic process (h) volume variation in the ith compartment due to water transfer (dm3 ) variation of the generic ith design parameter specific energy consumption (kWh/kg) Faraday efficiency (dimensionless) empirical exponent of current density (dimensionless) voltage applied to the ED electrodes (V) solute flux per unit current density as defined by Eq. (10) (kg h−1 A−1 ) specific energy consumption per unit current density as defined by Eq. (9) (kWh m2ν kg−1 A−ν ) solute recovery efficiency (dimensionless) overall working period of the ED unit (days/year)

140

ξ

M. Moresi, F. Sappino / Journal of Membrane Science 164 (2000) 129–140

correction factor for safety overdesign (dimensionless)

5.2. Subscripts ac C D fs

referred referred referred referred

to to to to

the the the the

generic acid concentrating compartment diluting compartment generic final salt

Acknowledgements This research work was supported by a grant from the National Research Council of Italy.

References [1] R.E. Lacey, S. Loeb, Industrial Processing with Membranes. Wiley-Interscience, Wiley, New York, 1972, pp. 21–106. [2] H. Strathmann, Design and cost estimates and applications, in: W.S.W. Ho , K.K. Sirkar (Ed.), Membrane Handbook, Chapman and Hall, New York, 1992, pp. 246–262. [3] R. Audinos, Liquid waste concentration by electrodialysis, in: N.N. Li, J.M. Calo (Eds.), Separation and Purification Technology, Marcel Dekker, New York, 1992, pp. 229–301. [4] B.T. Batchelder, Electrodialysis applications in whey processing, FIL-IDF Bulletin 212 (1987) 84–90. [5] P. Boyaval, C. Corre, S. Terre, Continuous lactic acid fermentation with concentrated product recovery by ultrafiltration and electrodialysis, Biotechnol. Lett. 9 (1987) 207–212.

[6] B. Bauer, H. Chmiel, Th. Menzel, H. Strathmann, Separation of bioreactor constituents by electrodialysis using bipolar membranes, in: M. Reuss, H. Chmiel, E.-D. Gilles, H.-J. Knackmuss (Eds.) Biochemical Engineering-Stuttgart, G. Fischer, Stuttgart, 1991, pp. 38–45. [7] S. Novalic, F. Jagschitz, J. Okwor, K.D. Kulbe, Behaviour of citric acid during electrodialysis, J. Membr. Sci. 108 (1995) 201–205. [8] R. Datta, E.P. Bergemamm, Process for producing of citric acid and monovalent citrate salts. U.S. Patent no 5,532,148; 2 July, 1996. [9] M. Mancini, M. Moresi, F. Sappino, Sodium citrate concentration by electrodialysis, in Abstracts of the 2nd Italian Conference on Chemical and Process Engineering, (ICheaP-2), Florence, Italy, 15–17 May 1995, AIDIC, Milano, 1995, pp. 894–897. [10] F. Sappino, M. Mancini, M. Moresi, Recovery of sodium citrate from aqueous solutions by electrodialysis, Ital. J. Food Sci. 8 (1996) 239–250. [11] M. Moresi, F. Sappino, Effect of some operating variables on citrate recovery from model solutions by electrodialysis Biotechnology and Bioengineering 59(3) (1998) 344–350. [12] M. Moresi, F. Sappino, Economic feasibility study of citrate recovery by electrodialysis, J. Food Eng. 35 (1998) 75–90. [13] T.B. Vickroy, Lactic Acid, in: M. Moo-Young, C.L. Cooney, A.E. Humphrey (Eds.), Comprehensive Biotechnology: Industrial Chemicals, Biochemicals and Fuels, vol. 3. Pergamon Press, Oxford, 1985, pp. 761–776. [14] R.C. Weast, M.J. Astle, CRC Handbook of Chemistry and Physics, 63rd edn, CRC Press, Boca Raton, FL, 1982, D.171. [15] Kirk-Othmer, Citric Acid, in: Encyclopaedia of Chemical Technology, 3rd edn. vol. 6, Wiley, New York, 1979, pp. 150–178. [16] D.A. Cowan, J.H. Brown, Effect of turbulence on limiting current in electrodialysis cells, Indust. Eng. Chem. 51(12) (1959) 1445–1448. [17] Y.-H. Yen, M. Cheryan, Electrodialysis of model lactic acid solutions, J. Food Eng. 20 (1993) 267–282. [18] W.S. Fong, Fermentation Processes, Report no. 95, Stanford Research Institute, Menlo Park, Ca, 1975, pp.71,179.