Reverse osmosis treatment of process water streams

Desalination, 82 (1991) 267-280 ElsevierScience PublishersB.V., Amsterdam

by A. Saavedra*

, G.

BertoniX, D. Fajner,

G.C.

Sarti

Dipartimento di Ingegneria Chimica e di. Process0 Universitj di Bologna Viale Risorgimento 2, I-40136 Bologna, Italy *Permanent address: Chemical Engineering Department Ilniversity of Chi.Le, Beaucheff 850, Santiago, Chile *ENIPlONT ANIC s.r.l.,

Enirnont Group, Mantova,

Italy

AIJSTHACT R.O.

separation has been considered for the treatment of a relevant water stream involved in a phenol production plant of The stream contains significant quantities of two high capacity. peroxides and ot' sodium salts of organic acids which organic demand appreciable processing in order to allow for its reuse within the process or, alternatively, for its discharge. Different R.O. membranes have been tested in a pi-lot plant in order to examine the feasibility of the process and to obtain reliable parameters for the design of a reverse osmosis plant.

INTRODUCTION Reverse osmosis process is a we.Ll established and worldwide spread technology in several fields as the desalination of brackish water and of seawater, of the concentration of fruit juice and the treatment of streams in food and in farmaceutical industries. High capacity R.O. plants suitable for the treatment of several tons per hour have been built so fnr only for desalination purposes. On the other side there i.san increasing demand for the treatment of process water streams in chemical and .Ln petrochemical plants due to both the increasing attention to environmental protection and to the tendency to develop selfcontained processes which reduce to the least minimum possible the amounts of eff.Luents. 267

268 The use of R.O. separations within traditional processes of chemical and petrochemical plants is very attractive in many respects; it involves very large aqueous streams and it is undoubtedly a worldwide very large potential application, to which an increasing attention has been payed in recent years. However, extending R.O. t0 this field is not always straightforward, mainly because it is very difficult to obtain a standardization in view of the major differences both qualitative and quantitative in the composition of the water streams; that forces us to consider every single applications as cases which must be analyzed separately from one another. The aim of this work is to establish wether R.O. is a suitable process to treat an aqueous stream entering one of the major phenol production plants such as the one located in Mantova and contains several operated by ENIMONT ANIC s.r.1. The stream phenol and many different organic compounds such as peroxides, organic acid salts. Our attention is focused on the possibility: of using R.O. with commercially available membranes, in the i) presence of the organic peroxides; a permeate stream quality which makes it ii) of obtaining suitable for a recycle within the process; the rate below iii) of reducing the concentrate flow predetermined target value of 'L'S%of the feed flow rate. On the other side our aim is also to establish the quantitative relationships which are needed in order to proceed to a reliable design of the R.O. plant. To that extent several tests have been performed in a R.O. pilot unit by using the real process mixture as well as model aqueous feeds containing either single solutes or groups of solutes in order to inspect the response to single components of the process stream.

EXPERIMENTAL Materials

and

Analysis

The process stream has a composition changing in time due to production fluctuations, therefore different samples have been used. The major characteristics of these streams are reported in Table 1.

269

mtOPERTIES ANALYSIS TYPE Conductivity COD

TABLE 1 OF THE PE?QCESS WATER

1 (wsiemens)

PH Silica (plxn) Iron (Ppm) Peroxide A (ppm) Peroxide B (pyxn)

2

SAMPLE Nr. 3 4

8312

8072

6.4 0.2 < 0.3 1607 5480

5.4 7.9 0.2 < 0 1 < 0.3 < 0.3 1810 137 6855 4434

6813

6996 24165 6.1 < 0.1 < 0.3 1160 7063

5 8679 25088 5.5 < 0.1 < 0.3 1733 5760

Model feeds were obtained either by considering only water and sodium organic salts or by considering the presence of single water and peroxide A or by adding peroxide A to a water-salts mixture. Tests have also been performed by considering aqueous solutions containing sodium salts of organic acids and peroxide A only or, alternatively, peroxide B only at the concentrations which are present in the process stream. The concentration of all the more significant organic acid salts, of phenol and of peroxides A and B, were initially measured in the permeate as well as in the concentrate streams in order to calculate the modu.Le rejection to single components. It has been later observed that the measures of the electrical conductivity, of COD, and of the total peroxide contents were sufficient to determine the rejection to salts and to the peroxides respectively. Tests have been performed by using different temperatures in the feed between 20"' C and 40' C and by considering different pressures between 15 and 30 bars. Based on the known variation with pH of the phenol rejection, tests have been performed at different pH in order to determine possible changes in the rejection to single components; the PH interval between 3.5 and 12 was inspected, with particular attention to the pH range of 6-10, -suggested also by the properties of the membranes used. Membranes Different membranes have been tested. A preliminary membranes have been made based upon the resistance

choice of the to oxidizing

270 environments. Use has been made of poliamide membranes, of composite membranes as well as, for the sake of reference, of cellulose acetate membranes. Specifically, the commercial modules choosen were the following: SEPAREM 495T (SEP-l), SEPAREM MGCc4040 (SEP-51, SEPAREN MGAC404OR96 (SEP-6), FILNTEC BW304040 (FT-11, and DESAL SG4040F (DES-l). The characteristics of these modules are reported in Table 2.

TABLE 2 CBARACTERISTICS OF THE MEMBRANES USED

ltlNHRANB TYPB

HMIHutl KAXIKUH AXIAL FUlW PRBSSURK TMPBRATuuB RANGE lbarl (l/b) ( ‘(7

pHEANCEHAXIHUH OXIDANT HAXIHUH TUWDITY CONCBNTEATION ww (PP Cl1I

SEP-1

42

40

1200- 4500

2 - 12

SBP-5

42

50

1200- 4500

4 - 11

SEP-I

31

40

1200- 4500

n-1

41

45

DB-1

40

50

Experimental

30

5

1000ppn-h

5

5,5 - 5,5

1.0

5

1200- 3500

2 - 11

( 0.1

5

1500- 4000

4 - 11

1000ppll-b

( 5

Equipment

The reverse osmosis unit used for the experimental runs, is a pilot plant equipped with a high pressure piston pump (15 kW, 80 4.4 ti/h), a multistage centrifugal prem (1.5 kW, 5 bar, bar, 4.0 ma/h), a prefilter system (20 microns cartridge prefilters), one vessel 2.5"x40" and two vessels 4"x40", for spiral-wound membranes, a feed tank (100 liters). The temperature is measured within the feed tank while presure gauges are set both in the feed and in the concentrate streams. The system is protected by automatic shut-down devices sensitive to controllers and temperature and pressure. A conductivity cell was installed on-line for the continuous determination of the salt rejection parameter. The equipnent was operated either in batch mode, by recycling both the concentrate

271 and permeate streams to the feed tank, or in semibatchmode recycling only the concentrate and collecting continuously permeate (concentration runs).

Experimental

by the

Procedure

The following types of experiments were carried out: i)

ii)

Initial state characterization of each reverse osmosis with standard solutions (NaCl 2000 ppn), at 25'C, modules, and at the suggested test pressure for the specific membrane. Measurements of transmembrane fluxes and rejections to the single species, present in the wastewater streams were carried out, as a function of the relevant operating conditions, mainly pressure, temperature, pH and concentration.

iii) Studies of the membrane behaviour, in short and relatively long lasting runs with the actual wastewater. iv) Studies in semibatch mode on subsequent concentration stages of the wastewater. Studies on membrane cleaning routines. v)

RESULTS

AND

DISCUSSION

For all the membranes, the flux behaviour vs the concentration salts for aqueous solutions which contain only sodium salts organic acids is very close to the flux which is obtained using NaCl mixtures at the same osmotic pressure, x, according the following general relationship: JW = Kw

(AP-

An,

of of by to

(1)

The typical behaviours are shown in Fig.1. As it is very general the transmembrane flux increases with increasing the temperature values. A typical bshaviour is shown in Fig.2. it is very cormnon, the transmembrane flux, JW, increases linearly with pressure so long as the rejection proves to be a constant value, while on the contrary a non linear dependance is observed in the low pressure range where a simultaneous decrease in the rejection is also observed. As

272 220-z

B

a***.

DESIT-1

00000

160-

8 *. Oo -

zclso3

SEP-_I

ooooo

2oO-

00

_

?3 140lz!

- **

00

0

:,

loo-

l

0

rw a 120ii! k!

1

0

0

_

0

0

eo-

so-

40

0

-l

:....,....,....‘..‘.,..mr...(,........,...,.,...,...r....~

0.05

0.10

SALT

O.PB

CONCENTRATION

0.20

(as

0.25 Formiote.

mold)

0.3”

Fig.1. Transmembrane flux vs salt concentration as formiate, for different membranes, at 30-C

Fig.2. Transmembrane flux vs temperature different membranes

for

0.

273 Major differences with respect to the usually expected behavior were associated to the presence of the organic peroxides A and B. In particular in all the R.O. membranes tested the presence of peroxide A causes a significant decrease in the transmembrane flux which is not apparently associated to the very minor in the osmotic pressure of the mixture (Fig.3). The increase specific interactions between the membranes and peroxide A, which for the above mentioned reduction, were not are responsible the inspected in great detail so far. Suffice it to say that indeed, after a cleaning effect is only partially reversible; subsequent tests with sodium chloride solutions show a routine, permanent reduction in the membrane permeability. Among all the inspected SW-1 has shown the minimum flux reduction membranes while ET-1 showed the minimum irreversible reduction. Peroxide B, on the other side, does not cause per se a significant change in the transmembrane flux.

-

SEP-1

Fig.3. The sharp decrease in transmembrane flux versus the time lapsed after the addition of peroxide A, at 25'C

274 A rather high rejection to the sodium organic salts has been observed up to almost the same values which were obtained with respect to NaCl in the reference tests. In view of this the overall salt rejection was later used by simply measuring the electrical conductivity of the feed, concentrate and permeate streams. Parallel to this, a very common behaviour has been observed as far as the rejection to phenol is concerned, either in a single solute mixture or in the complex multicomponent solution; namely for pH 10.5 phenol is salified and therefore very high rejections up to the value of 95% have been obtained. The same behaviour has been observed for all membranes qualitatively, although the actual rejections to salified phenol obviously differs -from membrane to membrane. Changes in pH in the range 6-11 did not affect appreciably the rejection of other components but phenol. On the other hand, when pH is decreased progressively down to acid values, also the rejection to specific organic salts becomes very low; that occurs as soon as the concentration of the free organic acid ions becomes very low, due to the dissociation equilibria. In all cases, at pH values below 4 the rejection to all the organic salts becomes irrelevant. No particular rejection has been observed to either peroxide A or peroxide B. In most cases the membrane does not act as a barrier side and the feed solution for both between the permeate peroxides. Thus, the values of the rejection which has been observed do not allow for a significant separation of the two peroxides from the solution, based on reverse osmosis. The actual values of the rejections are reported in Table 3. Tests have mainly been performed for a relatively short period of time in the order of 15-20 hrs, although longer tests have also been considered up to 200 hrs. From the data the parameters entering the expression for the transmembrane flux, Eq.(l), have been obtained in the form of JW = Wexp(-6/T)*[AP

-T*kz*XscJ

(2)

where l3,6 and k2 are membrane characteristic constants been experimentally obtained for any membrane tested.

that

have

275 For the membranes SEP-1, Fl'-1and DES-1 the parameters reported in 'fable4 both in the presence and in the absence peroxide A.

are of

TABLE 3 SOLUTE REJECTION AT 25 C, p1.i = 7 MEMBRANE TYI'E

PRQCESS PRESSURE (bar)

SEP-1 SEP-5 SEP-6 Fr-1 DES-1

SOLWE ORGANIC SALTS

30 23 27 15 15

94.5 97.5 95.5 99.5 918.0

REJECTION (%I PEWXIDE

A

35 20 0 55 10

PEROXIDE B 0 0 0 20 18

TABLE 4 EXPERIMENTAL PARAMEOF R.O. MODULES MODULE TYPE

MW3RANEPARAMETERVALUES 9 (l/h m'l kz (bar/K) 6 (K) with A without A

SEP-1 m-1 DES-l

1.119Et3 8.2413+4 3.55OEt3

1.3413+3 2.1173~5 9.362Et3

2004.9 2988.9 2096.9

2.3233-6 2.3963-6 2.4013-6

For each membrane,,the rejection to salts changes by changing the transmembrane flux; in terms of changes in pressure the rejection varies approximately according the well known expression [3]: l/R = 1 t B/Jw

Plant

(3)

Analysis

A possible R.O. plant suitable to treat the considered process stream and to meet environmental and economical requirements associated with the conduction of the phenol plant, should reduce the content of salts and phenol in the permeate, possibly leaving only a content of peroxides, which make the permeate itself suitable for a complete recycle within the chemical plant.

276 Preliminary economical estimates indicate that the introduction of a R.O. plant is interesting if it gives rise to a concentrate stream which is of the order of no more than 20-E% of the feed. There are different possible configurations for such a R.O. plant. For the entire flowrate, which is of several tons per hour, two possible configurations are indicated in Fig. 4. For each configuration different operating conditions may be followed in terms of both pressure and temperature. When operating at higher pressures the requirement of reducing the concentrate flow rate to the given fraction of the feed is obtained by using a smaller number of modules. A rather thorough analysis of the configurati.ons and of the different operating conditions has been performed. We like to notice here that the may be obtained either in the configuration of requirement Fig!.4/A at P=35 bar and T=30'C, or in the configuration of Fig.4/B at F=45 bar and T=35"C. The simulation calculations have been performed by using which hold true for the membrane FT-1 according parameters the measurements performed in the pilot unit. Noteworthy parameters have been obtained after the operating time of hours.

the to the 'LOO

In Fig.5 we report the values obtained at 35 bar and 30°C in terms of salt concentration in the concentrate and in the permeate streams, out of each module in the configuration of Fig.$/A; each module is characterized by its position in the series sequence. For clarity sake the total cumulative water recovery after each membrane module is also indicated, as well as the osmotic pressure in the concentrate. Analogous data are reported in Fig.6 for the plant configuration of Fig.4/B. PERtKATE

I

1

I

J

J

I

I

I

4

CONCENTRATE

Fig.$/A. Reverse osmosis configuration with eighteen modules for the treatment of process stream

PERMEATE

CONCENTRATE I

F&.4/B.

I

Reverse osmosis conriguration with eleven modules for the treatment. ot' process stream

Fig.5. Process in the module series

txuxmeters out of each membrane module piant configuration of F&.4/A; each is indicated by its position in the sequence; T=30'C, p=35 bar.

. concentration in the concentrate (ppm*10S3 1 Ez; concentration in the permeate (ppn*lO-' 1 EZ: osmotic pressure in the concentrate (bar) cumulative water recovery B3:

278 Long time operation runs are needed before giving a more complete answer to the feasibility problem of the operation, and in order to know what is the cleaning plicy which must be followed for the plant. It is in closure worthwhile to point out explicitly that our calculations have been performed for the low flux data which are obtained in the presence of peroxide A. Of course, should peroxide A been removed by a pretreatment of the R.O. feed all flux values increase significantly (see Tab.4). stream, In such case, in addition, the life problems and the needs for frequent cleanings become less severe.

go/

0

1

2

3

4

5

e

7

ST&X MJMBER

Fig.6. Process naramcters out of each membrane module in the plant configuration of Fig.$/B; each module is indicated by its position in the series sequence; T=35'C, p=45 bar. . concentration in g; concentration in ES3: osmotic pressure water tZ2: cumulative

the concentrate (ppm*lO-3 1 the permeate (ppm*lO-1) in the concentrate (bar) recovery

279 CONCLUSIONS Reverse osmosis was applied to process a water stream of a phenol production plant. The stresm considered contains, in addition to traces of phenol, several sodium organic salts and two organic peroxides, indicated as A and B respectively. The rejection to the organic salts is rather high and close to the values obt,ained for NaCl for all the membranes inspected. On the contra.ry, no appreciable re.jection has been observed for the organic peroxides. The major effect produced by the latter comrx,nenLs is a marked decrease in the transmembrane flux, which is not associated to a suitable increase in the osmotic pressure. Such a flux reduction is not completely recovered after a cleaning routine. The amount of the flux decrease, as well as the percent ot' recovery after cleaning, change from membrane to membrane. The latter effect must be further inspected in order to obtain informations on both membrane life and periodicity of the need of a as well as on the possible cleani.ng routines, pretreatment which reduces the peroxide content. In all cases, even with the low flux values obtained in the presence o!Z'the peroxides, it. is feasible Lo introduce reasonable reverse osmosis plants in order to decrease the organic salt content to a sufficiently low value which makes it possible to recycle the water stream within the process, and simultaneously to reduce the concentrate stream to values below 25% of the feed stream.

Acknowledgements This work has been supported by ENIMONT ANIC s.r.l., Resins and as a par-t of its program PoIyammidc Intermediate Division, devoted to the continuous improvement of the environmental impact Grateful thanks are due to Mr. the ,production plants. of F.Or1andin.i for his help in conducting the experimental tests.

LIST Ju: Ku : Ii2:

AP:

m

SYMBOLS

solvent flux solvent (water) permeation coefficient osmotic pressure coefficient (bar/K) hydraulic pressure applied across the membrane

280

An:

osmotic pressure difference across the membrane T: absolute temperature (K) solute concentration in the concentrate side (ppm) KS,: t-03? factor solvent permeation 13: preexponential the coefficient for the membrane (kg/mzh) energy factor for the solvent permeation coefficient of the 6: membrane (IO

LITERATURE

CITED

D. and Madadi, 1. Bhattacharyya, M.R. Separation of Phenolic Compounds by Low Pressure Composite Membranes: Mathematical Model AICHE Symposium Series, ~01.84, N"261, and Experimental results. (1987) 139-157. 2. Tone, S. Shinohara, K. Igarashi, Y. and Otake, T. Separation of Aromatic Substances from Aqueous Solutions Using a Reverse Osmosis Technique with Thin, Dense Cellulose Acetate Membranes. J. Membr. Sci., ly (19841 195-208,

3. Rautenbach 1989.

R., R. Albrecht,

"Membrane Processes"

J. Wiley Fd.