Water Research Vol. 11, pp. 379 to 385. Pergamon Press 1977. Printed in Great Britain,
REVERSE OSMOSIS ON SECONDARY SEWAGE EFFLUENT: THE EFFECT OF RECOVERY ROBERT WECHSLER*
Australian Atomic Energy Research Establishment, Lucas Heights, New South Wales, Australia
(Received 2 May 1976) Abstract--In the treatment of chemically clarified secondary effluent by reverse osmosis, product quality, flux stability and flux regeneration were examined as a function of the recovery of the operation. As the recovery of the plant increased, the salinity of the product increased while the level of organic contamination in the product remained constant. The effect of recovery on the economics and productivity of reverse osmosis is examined quantitatively using the experimental results. Provided the precipitation of calcium sulphate can be prevented, it is economically desirable to maximise recovery; the upper limit of recovery will be determined "by the salinity of the product which can be tolerated. l. I N T R O D U C T I O N
In recent years, the reuse of waste water has received serious attention as it presents not only a possible source of domestic water supply but also an ideal solution to the problem of disposal of sewage effluents. The extent to which these benefits can be realised increases with the extent of reuse and so operation of the reclamation system at high recovery is desirable. If domestic sewage is to be recycled, dissolved salts must be removed in addition to ammonia, BOD, bacteria and virus etc. For this reason, reverse osmosis is ideally suited for incorporation into domestic waste water reuse systems. There is a fundamental economic argument for high recovery operation of reverse osmosis units in sewage application. The membrane fouling problems associated with such applications dictate the use of extensive treatment prior to reverse osmosis. Since 100% of the influent to the reverse osmosis unit must be pretreated, the total operating cost of the reclaimed water is decreased if the recovery of the reverse osmosis unit is increased. The extent to which the cost savings can be realised by high recovery operation will, of course, depend on the relative operating costs of the pretreatment and reverse osmosis operations. Consider a plant which employs primary treatment, secondary treatment and chemical clarification as pretreatment to the reverse osmosis plant. The relative operating costs of the pretreatment and reverse osmosis will depend to a certain extent on the scale but, for plants of approximately 10 MGD, the operating cost of pretreatment will be about 70% of the unit production cost of reverse osmosis treatment. For larger plants, this figure will tend to decrease while for smaller plants it will tend to increase due to the greater influence of scale on operating
costs of the pretreatment operations compared to reverse osmosis. In any case, the cost of pretreatment is a significant part of the total operating cost of the reclamation process, and hence operating costs can be reduced by high recovery operations. A third argument for high recovery operation of reverse osmosis is found in the problem of brine disposal. The disposal of the brine is always a problem in reverse osmosis installations. The unit cost of brine disposal is generally not dependent on the quality of the brine. High recovery operation of reverse osmosis plants produces less brine and hence lowers the operating costs. These are strong economic arguments for high recovery operation of reverse osmosis plants for sewage applications. There are, however, technical difficulties in operating reverse osmosis at high recovery. High brine concentrations cause both depreciation of product quality and increase the rate of membrane fouling. The increased rate of fouling will cause the operating costs of the reverse osmosis treatment to rise so there will be some optimal recovery for which the operating cost of the whole reclamation process will be a minimum. This paper deals with the effect of recovery on plant performance and product quality in a pilot scale municipal waste water reclamation plant in the city of Leiden (South Holland). 2. EXPERIMENTAL
The reclamation pilot plant was operated on secondary effluent from the municipal sewage installation of the city of Leiden. The Leiden municipal sewage plant serves a population equivalent of 60,000 and is typical of many plants in Western Europe. The complete purification treatment sequence is shown in Fig. 1. Raw sewage is comminuted and * Work carried out at: Testing and Research Institute screened before passing to the grit chambers and the of the Netherlands Waterworks, Sir Winston Churchill- pre-aeration tank where it is mixed with waste actilaan 273, Rijswijk Z-H, Netherlands. vated sludge, aerated for 45 min and then passed into 379
SCREENING r ' m l R I ~ A L I - ] P I A E R A T I O N ~ T ~ I ~ _ N ~ ~Pp-
GR SLUDGE DIGESTERS
RECLAMATION[ PILOT PLANT I
POD m PERMANGANATE OXYGEN 0EMAND
Fig. 1. Flow diagram of complete sewage treatment plant. the primary settling tanks. Primary settling takes place over four hours after which the effluent enters a conventional step aeration activated sludge plant with 3 h aeration period. The wasfe sludge is recycled to the pre-aeration tanks. The secondary effluent enters the canal system of Holland without further treatment. This effluent was used as a feed water to the reclamation pilot plant. The pilot plant consists of chemical clarification and reverse osmosis with optional granular carbon polishing. A detailed flow diagram of the reclamation unit is given in Fig. 2. Clarification was carried out batchwise. A 6001 batch of secondary effluent was dosed with 500ppm alum (as A12 (SO4)a. 18H20), agitated slowly for 15 min and then allowed to settle over several hours. The clarified effluent was pumped through a dual medium filter of crushed anthracite and sand to give a very clear, almost colourless, effluent which was further polished by a 20/an cartridge filter before entering the reverse osmosis plant. The reverse osmosis plant consisted of two macroscopic tubular modules. One of these modules contained a membrane annealed at 80°C whilst the other contained a membrane annealed at 86°C. The membranes were supplied unannealed by Patterson Candy Int. Annealing is the final step in the manufacture of cellulose acetate membranes. It involves placing them in hot water at specific temperature in the range from 60°C to 90°C for about 10 min. This raises the salt rejection of the membrane at the expense of the water flux and therefore allows some degree of tailoring of membranes for specific application. A recirculation pump provided a brine velocity of 120 cm/sec. The pressure in the brine loop was maintained at 40 atm. by means of a 2 cylinder piston pump with a bypass regulator. The temperature of
the brine was maintained automatically at 25°C. The plant recovery was varied by variation of the reject brine flow by means of a time proportional device. The effect of recovery on the product quality and on the flux decline was studied. 3. RESULTS (a) Product quality The brine conductivity was continuously monitored to indicate the establishment of steady state conditions. At each steady state, the flux and brine flow rate were measured and samples taken for analysis, Two parameters were measured as indicators of product quality, the concentration of chloride ion and the chemical oxygen demand (permanganate number) determined by the permanganate oxidation method (Hitlebrand et al., 1955): The permanganate number is directly related to the concentration of organic substances present which can be oxidized by permanganate under prescribed conditions. Aromatic and straight chain aliphatic hydrocarbons are not included in this determination. The effect of variation in recovery on these two indicators of water quality is shown in Fig. 3. The chloride concentration in the product increased rapidly as the recovery of the plant increased. The observed behaviour of the chloride ion concentration is as expected (Kedem & Katchalsky, 1958, Katchalsky & Curran 1967, Spiegler & Kedem, 1966). As the recovery increased, the brine concentration increased and hence the salt flux increased in obeyance of Fick's law of diffusion. On the other hand, permanganate number of products from both membrane types did not vary significantly as the recovery was varied over a wide range. There was practically no difference in quality between
If I . . . . I I ~ PODC~ OOppat P~DOcdSSppnt P00"~4.8 ppR~
POD--PERMANGANATE OXYGEN DEMANO
! CAFIgOH I p R ^ ~ U ~ T I C O L U M N L . - ~ ''~' " lin~ ~ l l t .I
Fig. 2. Flow diagram of sewage reclamation plant:
Reverse osmosis on secondary sewage effluent
ecular dimensions of the membrane impermeable group. The concentration of the membrane impermeable group is increased by increasing the recovery but not that of the membrane permeable species. If such a mechanism is operative, the concentration of organic species present in the product will be equal to the concentrations of membrane permeable organics present in the reverse osmosis plant influent. This amounted to about 5 ppm or about 12~o of the permanganate oxidisable organics in the reverse osmosis plant influent. In this study, only the behaviour of permanganate oxidisable organic species was studied. There is no reason to assume that other organic species not included in such analysis would behave differently.
~ 100 U
e - - * - * . ~ - - e--j..
40 60 RECOVERY ( % )
Fig. 3. Product quality as a function of product recovery. 0------(3 Concentration of chloride in product from 80°C cured membrane. O O Concentration of chloride in product from 86°C cured membrane. O------O Permanganate oxygen demand in product of 80°C cured membrane. • • Permanganate oxygen demand in product of 86°C cured membrane.
z 8S 0 I-
the products of the two types of membrane as determined by the permanganate number. This independence of product permanganate number with recovery (and hence brine permanganate number) is inconsistent with normal behaviour of cellulose acetate membranes. The rejection of chloride ion and permanganate oxidisable organics is plotted against the brine concentration in Figs. 4(a) and (b). The chloride rejection is independent of the brine chloride concentration. This indicates that the Fick's diffusion law is obeyed. However, the rejection of permanganate oxidisable organics increased with increasing concentration of permanganate oxidisable organics in the brine. This indicates that when examined as a group, the flux of permanganate oxidisable organics did not obey the Fick Diffusion equation but may have done so when examined individually. This rather anomalous situation can be explained by considering the membrane as being completely permeable to a fraction of the organic species present in the influent and completely impermeable to the rest. In other words, if one considers the pore model to be an adequate description of the selective mechanism of cellulose acetate membranes, the mean pore size in this case will be greater than the molecular dimensions of the membrane permeable group of organics and smaller than the mol-
W.R. I1 4
I ~) I I 200 4 0 600 IK)O I000 BRqNE SOLUTE CONCENTRATION (rag/t)
Fig. 4. (a) Rejection of chloride and permanganate oxygen demand by 80°C cured membranes as a function of brine concentration. • Chloride. • Permanganate oxygen demand. I00
4OO 6 0 I 0 I000 BRINE SOLUTE CONCEN'rRRTION ( m g / t ~
Fig. 4. (b) Rejection of chloride and permanganate oxygen demand of 86°C cured membranes as a function of brine concentration. • Chloride. • Permanganate oxygen demand.
pattern to be expected when the Fick law of diffusion is obeyed. The product concentration of chloride ion increased rapidly as the recovery was increased. The level of salt in the product will limit recovery on grounds of quality, not the product concentration of organic species.
Therefore it can be concluded, that the level of organic contamination in the product does not increase with the recovery, but remains remarkably constant at a level, in this case of 5 ppm permanganate oxygen demand, which corresponds to a reduction of 88% of the permanganate oxidisabte organics present in chemically clarified and filtered secondary effluent. From the engineering standpoint, the behaviour of the permanganate oxidisable organics indicates that the level of organic contamination in the product is independent of the plant recovery and hence high recovery operation does not depreciate the product quality with respect to dissolved organic pollution. Also high recovery operation does not increase the operating costs of granular activated carbon post treatment, as would be expected if the rejection behaviour of organics was similar to that of, say, chloride ion. The chloride ion rejection behaviour showed the 0,70("
(b) Flux decline as a function of recovery The flux decline of the reverse osmosis unit was also studied as a function of the recovery. The rate of flux decline was determined using a single batch sample for each of the recoveries studied. Although this technique limited the duration of the test, it eliminated the effects of stop/start operation and of fluctuations in influent quality. Variations in influent quality between experimental runs were minimised by sampling at the same time of day. The flux decline behaviour of the reverse osmosis plant is shown in Figure 5(a) and 5(b). Typical decay curves were found for both membrane types at 35. J
io',~o',~o',:,o' •r I M E
Fig. 5. (a) Flux decline behaviour of 86°C cured membrane as a function of recovery.
,,,o' ,~o~ ,so' ~do 22o
Fig. 5. (b) Flux decline behaviour of 80°C cured membrane as a function of recovery.
Reverse osmosis on secondary sewage effluent 50 and 80% recovery. At 93% recovery, the flux decline curve showed an abnormally sudden and very severe decline after about 56 h of operation. This behaviour was observed with both types of membrane and was found to be reproducible. Analysis of brine samples showed them to be supersaturated with CaSO4. Such brine produced a crystalline precipitate on standing for several days. This behaviour at 93% recovery was probably due to precipitation of CaSO4 on the membranes. To examine this possibility, an ion exchange softener was placed in the reverse osmosis influent line and the experiment repeated at 95% recovery. Under these conditions, the flux decline curve followed the typical decay curves obtained at lower recovery operation. The rate of flux decline can be described by Equation (1) (Merten et al., 1967) FT = K . T ~
F r = flux at time T K = constant m = flux decline index. Equation (1) was fitted to the experimental data by a regression analysis to obtain the value of the flux decline index, m, and the constant K. The flux decline indices for both membrane types are plotted as a function of recovery in Fig. 6. There is an increase in the flux decline rate with recovery which is gradual provided that precipitation of brine components does not occur. The flux decline indices of the 80°C membrane are typically higher than those for the 86°C membrane. This is a common observation and is one of the hazards of using high flux membranes. In view of the ease with which the initial fluxes could be regenerated (discussed below), it is safe to conclude that membrane fouling does not constitute a major obstacle to high recovery provided that no species reaches saturation concentration in the brine.
by depressurisation is a simple process which can be easily automated and does not add significantly to operation costs. As a result, more frequent membrane cleaning operations can be used than would be practical if the more laborious methods of foam ball swabbing or chemical cleaning methods were necessary. This further results in increased membrane productivity and lower operating costs. (d) The effect of recovery on membrane productivity The flux decline index may be translated into membrane productivity by integrating the flux decline equation (equation (1)). Q=
[ T ' ' + ' -- T~+I].
Where Q is the amount of product (per unit area of membrane) produced during the interval between To and T (i.e. the period between successive cleaning operations). m is the flux decline index as previously defined. The flux decline equation (equation (1)) describes the flux as a continuous function of time. In practice, however, membrane cleaning operations result in discontinuities so that operation over periods where membranes have been cleaned cannot be adequately described by an equation of the form of equation (1). The flux decline equation as presented is only followed closely during the interval between consecutive membrane cleaning operations and can only be applied over that period. Tables l(a) and (b) show the membrane productivity for both membrane types to be expected from a plant operated at various recoveries and with cleaning intervals of one, three and five days. Also shown in these tables is the relationship of each productivity -I00
(c) The effect of recovery on membrane cleaning
The response of the flux to a series of cleaning tests was observed after each determination of the flux decline rate. Three cleaning methods were examined: (1) sudden depressurisation for 10 rain followed by a sweet water flush; (2) foam ball swabbing; (3) chemical cleaning for 30min with a fresh 2% sodium perborate solution brought to pH 7 by addition of hydrochloric acid. The flux response was measured after each test. Complete regeneration of the original flux was obtained by a 10 min depressurisation, followed by a sweet water flush to remove the dirt desorbed during the depressurisation. No further flux improvement was obtained in any of the flux decline studies in this series by use of the more rigorous methods of foam ball swabbing or perborate cleaning. Cleaning
80% ~,F,,,E . . / . . " _1 o
Fig. 6. Flux decline index as a function of recovery.
Table l(a). Projected membrane performance for 80°C membranes Recovery t%)
1 day cleaning interval Productivity Max. in mam -2 production (~o)
35 50 80 95 (So~ened influent)
0.676 0.658 0.612 0.606
3 day cleaning interval Productivity Max. in m3m -2 production (Vo)
99.4 94.6 89.8 89.5
2.08 1.99 1.80 1.76
98.6 94.6 87.2 84.0
5 day cleaning interval Productivity Max. in m3m - 2 production (°/o) 3.38 3.35 3.12 2.92
93.8 93.8 87.0 82.8
Table l(b). Projected membrane performance for 86~'C membranes Recovery (~,;,)
1 day cleaning interval Production Max. m3m- 2 production (%)
35 50 80 95 (Softened influent)
0.591 0.591 0.553 0.541
3 day cleaning interval Production Max. roam- 2 production (~,)
98.5 98.5 92.1 91.3
to that which would be expected if there was no flux decline. An off stream period of 1 h was allowed for the membrane cleaning operation in these calculations. The data corresponding to 93Vo recovery have not been treated as the precipitation of CaSO4 produced a situation which would not correspond to practical operation. Tables l(a) and (b) show that, as the recovery is increased, there is a consistent decrease in membrane productivity. The extent of this decrease is dependent on the cleaning interval, particularly at high recovery operation. The losses in productivity caused by the higher rate of membrane fouling at high recovery operation can be reduced to about 10~o by daily membrane cleaning. In view of the ease with which the membranes could be cleaned, such a cleaning frequency is clearly a practical possibility. (e) The effect of recovery on reverse osmosis economics
High recovery effects the performance of reverse osmosis in two ways, firstly by increasing the salt concentration in the product and secondly by increasing the rate of flux decline. The latter results in the necessity for a higher frequency of membrane cleaning operations and hence increased down time and greater labour costs which lead to increased operating costs. On the other hand, high recovery operation allows a saving in the pressurising pump energy costs and permits the use of a smaller capacity pump for this purpose. The installed cost of the pressurising pump is a very small part of the investment and any saving from the use of a smaller capacity pump will not, in general, significantly effect the operating costs of the reverse osmosis installation. Energy costs are, however, a significant part of the total operating cost. Lacey (1972) estimated that such costs amount to 13Vo of the total operating costs of
1.81 1.8l 1.61 1.59
97.8 97.8 88.5 86.0
5 day cleaning interval Production Max. m3m- 2 production (%) 3.02 3.02 2.07 2.6l
97.5 97.5 86.5 84.3
a 1 M G D reverse osmosis plant operated at 50% recovery. In high recovery operations (i.e. 95Vo), this energy cost may be decreased by 47~o. This saving corresponds to 6.1~o of the total operating costs and offsets to a major extent the extra costs entailed in the loss in production due to the more severe flux decline which accompanies high recovery operation. The cost of activated carbon polishing is greatly dependent upon the organic loading of the carbon columns. It has been shown that the product concentration of organic contamination is independent of the recovery of the reverse osmosis plant. As a corollary to this the cost of activated carbon polishing is independent of the recovery of the reverse osmosis operation. If the influent of the reverse osmosis has been extensively pretreated, as is the case in the present study, a very substantial reduction in costs can be realised by high recovery operation as the whole of the reverse osmosis influent must be pretreated. The extent of this cost saving will depend on the relative operating costs of the pretreatment processes and the reverse osmosis process. A detailed cost analysis of the complete waste water reclamation is beyond the scope of this paper. 4. CONCLUSIONS This paper describes a field pilot plant experiment designed to ascertain the effect of increasing the plant recovery on the product quality and production rate. It was established that: (1) the concentration of salt in the product increases as the recovery increases; (2) the product concentration of organic components in the product does not increase significantly as the recovery is increased; (3) the flux decline rate is not an obstacle to high recovery operation provided that the crystallisation
Reverse osmosis on secondary sewage effluent of hardness producing substances is prevented. The use of lime clarification is strongly recommended if high recovery is to be realised at reasonable cost. The membrane productivity can be maintained at about 90~o of maximum by daily cleaning by depressurisation and flushing with sweet water; (4) the ease with which the membranes can be cleaned is not affected by the plant recovery; (5) it is expected that the limit of recovery is set by the concentration of salt that can be tolerated in the product and not by the rate of flux. decline or the concentration of organic species in the product; (6) because considerable investment has been made in primary treatment, secondary treatment and chemical clarification prior to the reverse osmosis step, it is economically advantageous to operate the reverse osmosis plant at maximum recovery.
Acknowledgement--The assistance of Mr. F. M. Schotel throughout this work is gratefully acknowledged. REFERENCES
Hillebrand W. F., Lundell G. E. F., Bright H. A. & Hoffman J. I. (1955) Applied Inorganic Analysis. 2nd Ed. p. 185, Wiley, New York. Katchalsky A. & Curran P. F. (1967) Non-Equilibrium Thermodynamics in Biophysics. Harvard University Press, Cambridge, Mass. Kedem O. & Katchalsky A. (1958) Thermodynamic analysis of the permeability of biological membranes to non electrolytes, Biochim. biophys. Acta 27, 229. Lacey R. E. (1972) The costs of reverse osmosis, Industrial Processing with Membranes. Ed. Lacey, R. E. and Loeb, S., Wiley, New York. Merten U., Lonsdale H. K., Riley R. L. & Voss K. O. (1967) Reverse osmosis membrane research U.S. off saline water, Res. Dev. Rep., p. 265.