Stage Cut Permeate Flow Feed Flow

Permeate Flow Rate

If the permeate flow rate is sufficiently high it can generate a back pressure that reduces the differential pressure and rate of permeation.

From: Gas Purification (Fifth Edition) , 1997

Reverse Osmosis

Hisham T. El-Dessouky , Hisham M. Ettouney , in Fundamentals of Salt Water Desalination, 2002

Example on Semi-Empirical Model

Use the statistical mechanical model to obtain the permeate flow rate and membrane area at the following conditions:

–

Membrane constants C₁, C₂, D₁, and D₂ with

${\text{C}}_1=-3.01\times {10}^{-1}$

${\text{C}}_2=1.195164$

${\text{D}}_1=-1.56\times {10}^{-4}$

${\text{D}}_2=1\times {10}^{-2}$

–

Other data includes:

Membrane salt rejection = 99%

Feed salinity = 34,000 ppm

Feed flow rate = 1000 m³/d

Permeate flow rate = 325 m³/d

Feed pressure = 6000 kPa

Pressure of brine reject = 5900 kPa

Permeate pressure =101 kPa

Salt concentration in membrane = 1.76 C_f

–

The salt rejection definition is used to calculate the product salinity, where

$\begin{array}{l}\text{SR}=\text{1}-{\text{X}}_{\text{p}}/{\text{X}}_{\text{f}}\\ \text{0}\text{.99}=\text{1}-\text{Xp}/\text{34000}\end{array}$

which gives X_p = 340 ppm

–

From the above the pressure drop across the membrane is given by

$\begin{array}{cc}\text{Δ}\text{P}\hfill & =\text{0}\text{.5}(\text{Pf}+\text{Pb})-\text{Pp}\hfill \\ \hfill & =\text{0}\text{.5}(\text{6000}+\text{5900})-\text{101}\hfill \\ \hfill & =\text{5849 kPa}\hfill \end{array}$

–

The osmotic pressure is calculated for the feed, brine, and permeate

$\begin{array}{l}{\text{π}}_{\text{f}}=75.84{\text{\hspace{0.17em}X}}_{\text{f}}=\left(75.84\right)\left(34\right)=2578.6\text{\hspace{0.17em}kPa}\\ {\text{π}}_{\text{b}}=75.84{\text{\hspace{0.17em}X}}_{\text{b}}=\left(75.84\right)\left(50.2\right)=3806.3\text{\hspace{0.17em}kPa}\end{array}$

${\text{π}}_{\text{p}}=75.84{\text{\hspace{0.17em}X}}_{\text{p}}=\left(75.84\right)\left(0.34\right)=25.8\text{\hspace{0.17em}kPa}$

–

The resulting average osmotic pressure on the feed side is then calculated

$\overline{\text{π}}=0.5({\text{π}}_{\text{f}}+{\text{π}}_{\text{b}})=0.5(2578.6+3806.3)=3192.4\text{\hspace{0.17em}kPa}$

–

Therefore, the net osmotic pressure across the membrane is given by

$\text{Δ}\text{π}=\overline{\text{π}}-{\text{π}}_{\text{p}}=3192.4-25.8=3166.7\text{\hspace{0.17em}kPa}$

–

The permeator area is calculated from the following equation

$\begin{array}{l}{\text{M}}_{\text{p}}/\text{A}=({\text{D}}_1{\text{\hspace{0.17em}c}}_{\text{w}}+{\text{D}}_2)(\text{Δ}\text{p}-\text{σ}\text{Δ}\text{π})\\ 325/\text{A}=(-0.000156(60.2)+0.01)(5849-3166.6)\\ \text{A}=199{\text{\hspace{0.17em}m}}^2\end{array}$

–

The salt rejection is then calculated from the following relation

$\begin{array}{cc}\text{SR}\hfill & ={({\text{C}}_1\frac{1}{\left({\text{M}}_{\text{p}}/\text{A}\right)}+{\text{C}}_2)}^{-1}\hfill \\ \hfill & ={(\left(-0.301\right)\frac{1}{325/199}+1.19514)}^{-1}\hfill \\ \hfill & =0.99\hfill \end{array}$

Read full chapter

URL:

https://www.sciencedirect.com/science/article/pii/B9780444508102500099

31st European Symposium on Computer Aided Process Engineering

Çağla Odabaşi , ... Ömer Çağlar , in Computer Aided Chemical Engineering, 2021

3 Results

In order to predict RO membrane performance indicators (salt passage, permeate flow rate and pressure difference), different ML methods were implemented and compared. The best performed models that resulted in minimum RMSE on test data prediction are used to detect the most influential parameters on three RO membrane performance indicators. Feed pressure, flow rate, conductivity, TSS, ORP, COD, turbidity and temperature were used as input parameters. The RMSE values of different models for each RO membrane performance indicator were presented in Table 1.

Table 1. RMSE values of validation and test data prediction using different methods

Methods	Salt passage		Permeate flow rate		ΔΡ
	10-fold-cv RMSE	Test RMSE	10-fold-cv RMSE	Test RMSE	10-fold-cv RMSE	Test RMSE
XGBoost	0.91	0.77	3.90	6.64	0.33	0.64
Random forest	0.71	0.65	7.32	10.51	0.35	0.63
ANNs	0.76	0.91	4.36	5.12	0.34	0.58
MLR	0.76	0.85	3.77	4.35	0.39	0.66

For salt passage prediction, a random forest model performed superior to other methods (Table 1). The RMSE of test data prediction was found to be 0.65. The actual and predicted salt passage values of test data are presented in Figure 1a . In random forest method, mean decrease in Gini (IncNode Purity) is used as a measure to analyse the variable importance (Kuhn et al., 2008). This measure represents the performance of each split considering the input parameters using Gini index. If the value of mean decrease in Gini of an input variable is high, this variable has a higher variable importance. Hence, considering the random forest model, it was found that the feed conductivity and temperature had more effect on salt passage than other parameters (Figure 1b).

Figure 1. Analysis of salt passage a) actual versus predicted values of test data (number of trees = 100), b) variable importance analysis

Bartels et al. also showed that when the feed water salinity increases, salt passage also increases (Bartels et al., 2005). Hence, the conductivity of the feed water has a significant effect on salt passage and ion removal efficiency. Secondly, a change in the feed water temperature of RO membrane trains causes the RO membrane pore diameters to change. If the feed water temperature increases, pore diameters expand and more ions can pass through the RO membranes. Hence, salt passage increases (Al-Bastaki and Al-Qahtani, 1994; Jin et al., 2009).

For the permeate flow rate prediction, the accuracy of the MLR model was found to be the highest and the RMSE value of test data prediction was found to be 4.35. The actual and predicted permeate flow rate values are shown in Figure 2a . In order to determine the relative importance of the input variables according to MLR model, the significant input variables of which p-values were found to be smaller than 0.05, were considered. In Figure 2b, the absolute values of the coefficients of significant input variables are given. Feed flow rate was found to be the major factor affecting permeate flow rate. Considering the strong linear correlation between feed and permeate flow rate (0.96), the success of the MLR model and the importance of the feed flow rate were expected. Temperature appeared to be the second influential factor for permeate flow rate. Regarding the expansion of the RO membrane pore diameters due to high temperature operation, the feed flow can pass through the pores more easily and the permeate flow rate increase. (Boulahfa et al., 2019).

Figure 2. Analysis of permeate flow rate, a) actual versus predicted values of test data, b) variable importance analysis

Lastly, ANN model performed better than other methods for predicting the pressure difference across RO membranes with a minimum RMSE of 0.58 on test data prediction (Table 1). The actual and predicted pressure difference values are given in Figure 3a . In order to determine the variable importance according to ANN model, Olden's method (Olden et al., 2004) was employed. Olden's method is based on connection weight approach that uses raw input-hidden and hidden-output connection weights in the neural network. According to the Olden's method, feed flow rate was found to affect the pressure difference significantly and this situation is expected considering the system hydraulics. Secondly, temperature was found to affect pressure difference. An increase in temperature causes a pore expansion of the RO membranes, hence, the pressure difference between the stages of the train is expected to be lower.

Figure 3. Analysis of pressure difference a) actual versus predicted values of test data (neural network with two hidden layer (2-12 neurons)), b) variable importance analysis.

Read full chapter

URL:

https://www.sciencedirect.com/science/article/pii/B9780323885065500383

Water and Membrane Treatment

Rajindar Singh , in Membrane Technology and Engineering for Water Purification (Second Edition), 2015

2.4.3 RO/NF array design

The following basic membrane separation phenomena should be kept in mind when designing a membrane system:

•

The permeate flow rate (convective) is proportional to the net driving pressure (NDP) differential across the membrane.

•

The salt flow rate (diffusive) is proportional to the concentration difference across the membrane, and is independent of applied pressure.

•

Permeate TDS depends on the relative mass transfer rates of water and dissolved solutes through the membranes.

•

The chemical and physical nature of the membrane determines the preferential transport of water over dissolved solutes.

•

The higher the permeate flux, the greater the likelihood of higher concentration polarisation (CP). As CP increases, the osmotic pressure of the solution in the feed-reject channel increases, salt passage increases, and the risk of scaling and/or fouling increases.

Membrane manufacturers provide design guidelines based on the parameters given below. These guidelines are usually modified based on the type of feed water and pre-treatment [33,40].

•

Maximum feed flow rate to any element in the pressure vessel

•

Maximum reject-flow rate from any element

•

Maximum product water recovery for an element

•

Maximum flow rate (or flux) for any element

•

Maximum average flux for a system

•

Maximum applied pressure

The optimal design of the RO system incorporates certain rules of thumb based on the membrane for the particular application:

•

Recovery per element is <   19% for softened water or well water with SDI   <   3

•

Recovery per element is <   16% for unsoftened water or surface water with SDI   =   3–5

•

Net pressure drop across the array is <   7   bar

•

Average flux for each element depending on the type of feed water ¹

•

Percent variation in permeate flow rate is <   10% between the first and last elements in the same pressure vessel

•

Feed water flow rate to the first element of each stage is the same, <   10%

In order to avoid excessive concentration polarisation at the membrane surface, permeate recovery per membrane element should not exceed 18%. In the case of brackish water RO systems, the average recovery per 100   cm (40-in.) long membrane element is usually about 9%. The overall recovery for a staged system with pressure vessels containing six elements is usually as follows [46]:

•

One-stage array (1)   =   52–56%

•

Two-stage array (2:1)   =   75–80%

•

Three-stage array (4:2:1)   =   85–90%

The recovery in each element is controlled by the concentration of rejected species, especially, sparingly soluble salts of calcium and magnesium and silica in the brine stream. When the product recovery is 50%, the salt concentration in the reject stream is doubled, whereas the salt concentration increases fourfold when the recovery is 75% due to the concentration factor. Hence, the RO system is operated below the design recovery point. In general, the product water recovery is maintained well below 15%, and the systems are usually designed for a recovery of 8–10% per element. The scaling and fouling potential is usually the highest in the last elements of the final stage as stated in Table 2.9.

The above conditions are taken into consideration when modelling a RO/NF membrane system. The engineers use computer-generated performance projection software provided by membrane manufacturers to design an optimal membrane array design that maximises the operating conditions and minimises fouling and scaling. A typical RO/NF programme calculates permeate quality (conductivity, pH), feed-pressure requirements, and the final concentrate stream solubility numbers such as LSI and SDSI depending on (a) permeate flow rate, (b) %recovery, (c) feed water composition, (d) feed water temperature, (e) type and number of membrane elements, (f) the rate of flux decline, and (g) the rate of salt passage increase.

The programme algorithm is an iterative calculation in which the computer first estimates a feed pressure to satisfy the desired recovery and then calculates the performance of the first element of the system [46,47]. The concentrate from the first element becomes the feed to the second element, and a second calculation of membrane element performance is made, and so on from element to element through the complete array of the proposed design. The programme then sums the permeate flow from all elements and compares this value to the target value. The programme adjusts the feed pressure based on this comparison, causing the solution to converge to the required feed pressure to achieve the required permeate recovery given the user-defined system parameters and until the programme has converged on a single unique solution. Calculations can be repeated with different design parameters or membrane element array configurations. If the programme does not converge, a warning is issued requesting a revised number entry. The programme also calculates the concentration polarisation coefficient called the β factor:

$\mathit{\beta }=\mathrm{concentration}\mathrm{at}\mathrm{the}\mathrm{membrane}\mathrm{surface}/\mathrm{concentration}\mathrm{in}\mathrm{the}\mathrm{bulk}\mathrm{fluid}$

β is a function of the ratio of permeate flow from an element to the feed-brine average flow for that element. The optimum value is 1.13 for the last element of the last stage of a membrane array. To maintain this value of β when using 100   cm (40   in.) long SW elements, the maximum recovery is usually limited to 15% for one element, 30% for two elements in series, and so on to 50% for six elements in series in a pressure vessel.

Computer-generated design of a single-pass, three-stage RO membrane array (4:3:3) with concentrate recycling is shown in Figure 2.24 and given in Table 2.11. The stream numbers in the figure are 1 is raw water, 2 is blended water, 3 is membrane array feed, 4 is reject, 5 is reject/concentrate recycle, 6 is reject to drain, and 7 is permeate. The RO unit is designed to produce 27 gpm (6.13   m³/h) permeate corresponding to an overall recovery of ~   75% (27 gpm permeate/36 gpm raw water feed). The TFC polyamide RO membranes reduce the TDS content from 421   mg/l in blended feed water to 5   mg/l in product water at an average rejection of 98.8%. A portion of reject flows to the drain at 9 gpm (2   m³/h), and the remaining recycles to the RO pump inlet at 5.5 gpm (1.25   m³/h). The design feed pressure is 145 psig (10   bar   g). The β factor is within range, and the LSI is <   1, which is within range with anti-scalant dosage. The RO membranes are spiral-wound modules, 4   in. diameter   ×   40   in. long.

Figure 2.24. Block-flow diagram of the single-pass RO membrane design given in Table 2.11 with reject recycling. Stream 1 is feed, 2 is blended feed, 3 is membrane array pressurised feed, 4 is reject, 5 is reject/concentrate recycle, 6 is reject-to-drain, and 7 is permeate.

Table 2.11. RO membrane system performance projection (Hydranautics)

Project name	Single-pass RO design	Permeate flow	27.00 gpm
HP pump flow	41.5 gpm	Raw water flow	36.0 gpm
Recommended pump pressure	144.6   psi	Total system recovery	75.0%
Feed pressure	115.6   psi	Permeate recovery ratio	65.1%
Feed water temperature	25.0°C (77   °F)	Concentrate recirculation	5.5 gpm
Feed water pH	7.14 (0.00)	Element age	0.0   years
Acid dosage, ppm (100%)	0.0 H2SO4	Flux decline % per year	7.0
Acidified feed CO2	11.1	Salt passage increase, % per year	10.0
Average flux rate	15.2 gfd	Feed type	Well water

Stage	Perm. flow (gpm)	Flow/vessel
		Feed (gpm)	Conc (gpm)	Flux (gfd)	Beta	Conc and throt. pressure		Element type	Elem. no.	Array
						(psi)	(psi)
1–1	12.8	10.4	7.2	18.1	1.13	100.5	0.0	ESPA2-4040	12	4   ×   3
1–2	7.8	9.6	6.9	14.8	1.11	83.8	0.0	ESPA2-4040	9	3   ×   3
1–2	6.3	6.9	4.8	11.9	1.12	72.2	0.0	ESPA2-4040	9	3   ×   3

	Raw water		Feed water		Permeate		Concentrate
Ion	mg/l	CaCO3	mg/l	CaCO3	mg/l	CaCO3	mg/l	CaCO3
Ca	32.6	81.3	45.5	113.4	0.16	0.4	129.9	323.8
Mg	7.4	30.5	10.3	42.5	0.04	0.1	29.5	121.3
Na	47.8	103.9	66.3	144.2	1.11	2.4	187.8	408.3
K	0.0	0.0	0.0	0.0	0.00	0.0	0.0	0.0
NH4	0.0	0.0	0.0	0.0	0.00	0.0	0.0	0.0
Ba	0.000	0.0	0.000	0.0	0.000	0.0	0.000	0.0
Sr	0.000	0.0	0.000	0.0	0.000	0.0	0.000	0.0
CO3	0.1	0.2	0.1	0.2	0.00	0.0	0.4	0.7
HCO3	69.3	56.8	96.5	79.1	0.80	0.7	274.7	225.1
SO4	44.7	46.6	62.4	65.0	0.07	0.1	178.5	185.9
Cl	50.5	71.2	70.4	99.3	0.33	0.5	200.9	283.4
F	1.8	4.7	2.5	6.6	0.02	0.1	7.1	18.8
NO3	44.7	36.0	61.6	49.7	2.11	1.7	172.4	139.1
SiO2	3.7		5.2		0.03		14.7
TDS	302.6		420.9		4.7		1195.9
pH	7.0		7.1		5.3		7.6

	Raw water	Feed water	Concentrate
CaSO4/K sp  ×   100	1%	1%	6%
SrSO4/K sp  ×   100	0%	0%	0%
BaSO4/K sp  ×   100	0%	0%	0%
SiO2 saturation	3%	4%	11%
Langelier Saturation Index	−   1.21	−   0.79	0.58
Stiff and Davis Saturation Index	−   1.16	−   0.74	0.60
Ionic strength	0.01	0.01	0.02
Osmotic pressure	2.5   psi	3.4   psi	9.8   psi

A typical RO skid is shown in Figure 2.25. It is a single-pass, two-stage (4:2 array) unit with pressure vessels containing six spiral wound membrane elements (20   cm   ×   100   cm) in series in each vessel. There is room on the backside of the skid to double the number of vessels to make it into an 8:4 array with permeate flow rates approaching 70   m³/h at 75%. The RO high-pressure pump is multi-stage, horizontal, submersible type. The skid includes a control panel and instruments such as conductivity and flow monitors shown on the right-hand side. The end view of a large RO/NF membrane system is shown in Figure 2.26.

Figure 2.25. A typical RO skid showing a 4:2 two-stage membrane array, high-pressure pump, instruments and control panels. The high-pressure RO pump is a horizontal, multi-staged submersible type. Each pressure vessel contains six spiral-wound modules, 20   cm dia. × 100   cm long.

Source: USFilter.

Figure 2.26. End view of a RO/NF membrane plant.

Source: Ultrapure Water.

The effect of colloidal fouling on membrane processes was discussed earlier in this chapter; it is a function of the permeate flux and the solids content of the feed solution. Since colloidal fouling has a strong negative effect on membrane performance, membrane systems are designed by limiting the permeate flux of each element based on the recommendations of the membrane manufacturers. One such plot for a RO membrane with various natural water feeds is shown in Figure 2.27. The data show that as the quality of feed water improves, the recommended permeate flow rate also increases. For example, the membrane flux is linear with pressure in the case of nearly pure water or RO permeate. The figure shows that the flux reaches a plateau at higher pressures for solutions other than pure water due to concentration polarisation (CP) as discussed in Chapter 1. The mid-point of the non-linear curve is the region of critical flux and the optimal operating condition to minimise CP as discussed earlier.

Figure 2.27. Membrane flux characteristics of a spiral wound RO module for various feed waters.

Source: Film-Tec membrane catalogue. For non-pure water solutions the flux reaches a constant, steady state value at higher pressures. For optimal performance, the system should be run below the "critical flux" region, which is the mid-point of the curved part of the curve.

Read full chapter

URL:

https://www.sciencedirect.com/science/article/pii/B9780444633620000021

27th European Symposium on Computer Aided Process Engineering

M.A. Al-Obaidi , ... I.M. Mujtaba , in Computer Aided Chemical Engineering, 2017

Nomenclature

Parameter	Description, units	Parameter	Description, units
A	Area of membrane (m2)	Qp	Total permeate flowrate (m3/s)
Aw	Water permeability constants (m/atm   s)	R	Gas constant (atm   m3/K kmol)
b	The friction factor (atm   s/m4)	Reb	Reynold number at feed channel (-)
Bs	Solute permeability constant (m/s)	Rec	The total recovery (-)
Cb	Bulk solute concentrations (kmol/m3)	Rej	The solute rejection (-)
Cf	Feed concentration (kmol/m3)	Rep	Reynold number at permeate channels (-)
Cm	Dimensionless solute concentration	T	Feed temperature (°C )
Cp	Permeate concentration (kmol/m3)	tf	Height of feed channel (m)
Cr	Retentate concentration (kmol/m3)	tp	Height of permeate channel (m)
Cw	Wall membrane concentration (kmol/m3)	W	Membrane width (m)
Db	Solute diffusion parameter (m2/s)	μb	Feed viscosity (kg/m   s)
E	The energy consumption (kWh/m3)	μp	Permeate viscosity (kg/m   s)
k	Mass transfer coefficient (m/s)	ρw	Molal density of water (kmol/m)
L	The membrane length (m)	ρb	Feed density (kg/m3)
P f(in)	Inlet feed pressure (atm)	ρp	Permeate density (kg/m3)
P f(out)	Outlet feed pressure (atm)	∆ρ b(in)	Inlet and outlet pressure difference (atm)
Pp	Permeate pressure (atm)	∆ρ b(out)	Outlet pressure difference (atm)
Qb	Bulk flowrate (m3/s)	εpump	The efficiency of pump (-)
Qf	Feed flowrate (m3/s)	θ	Parameter defined in Eq. (20)
Qr	Petentate flowrate (m3/s)

Read full chapter

URL:

https://www.sciencedirect.com/science/article/pii/B9780444639653504542

Fundamentals

In The MBR Book (Second Edition), 2011

2.3.7.3 Unsteady-state Operation

Unsteady-state operation can arise from such things as variations in feedwater quality (and so organic load), permeate flow rate (and hence hydraulic load) and aeration rate, which are all known to impact on MBR membrane fouling propensity, along with other dynamic effects ( Table 2.13). In an experiment carried out with a large pilot-scale MBR in which the effects of unstable flow and sludge wastage were assessed (Drews et al., 2006), it was established that the level of carbohydrate in the supernatant before and after each sludge withdrawal increased. Whilst the increase following wastage was thought to be due to the sudden stress experienced by cells due to biomass dilution (which in extreme cases is known to lead to foaming in full-scale plant), increase before sludge withdrawal was attributed to the high MLSS concentration and the resulting low DO level in the bioreactor. It was concluded that unsteady-state operation changed the nature and/or structure (and fouling propensity) of the carbohydrate rather than the overall EPS formation. These findings corroborated results previously reported on effects of transient conditions in feeding patterns: the addition of a pulse of acetate in the feedwater has been shown to decrease significantly the MBR biomass filterability due to the increase in SMP levels produced (Evenblij, Verrecht, van der Graaf, & Van der Bruggen, 2005b). More detailed characterization of the impact of a wide range of unsteady-state conditions on the EPS present in activated sludge has recently been presented (Yang & Li, 2009). Along with changes in DO level, variation in the ratio of monovalent and polyvalent cations present in the feedwater can result in sludge deflocculation, usually leading to increased supernatant SMP levels. In the experiments reported by Van Den Broeck et al. (2010), high monovalent/polyvalent ratios resulted in significant deflocculation and decline in hydraulic performance.

Table 2.13. Examples of Dynamic Effects in MBR Operation (see also Table 3.38)

Determinants	Parameters affected
Flow rate	Ultimate flux and rate of change
Feedwater quality	Ultimate composition and rate of change
MLSS dilution	Dilution factor and rate of concentration change
(Partial) aeration loss	Percentage and period of reduction
Backflush/cleaning loss	Period of loss
Hydraulic shock	Rate and level of flow increase
Saline intrusion	Ultimate concentration factor and rate of concentration change

The effects of starvation conditions on the biological suspension have been assessed by incorporating different substrate impulses in batch tests (Lobos, Wisniewski, Heran, & Grasmick, 2005). Exogenous phases were followed by starvation periods, both characterized by the S/X (substrate to biomass concentration ratio) where high ratios led to multiplication of bacteria cells, whilst at low ratios MLVSS decreased, SMPp production was absent and bacteria lysis ceased. S/X closely relates to F/M ratio (Equation (2.23)), and the low F/M values generally used in MBRs are thus theoretically close to starvation conditions which are in turn likely to be beneficial to MBR operation on the basis of the reduced SMPp production and correspondingly reduced fouling.

The principal period of unsteady-state operation is during start-up when the system is acclimatizing. Cho, Song, Lee, & Ahn (2005b) reported temporal changes in the bound EPS levels when the MBR was acclimatized at three different SRTs (8, 20, 80 days). As expected from general trends described in Section 2.3.6.5, the EPS concentration was lower at the longer SRT (83 vs 26   mgTOC/gSS for SRTs of 8 and 80 days, respectively). An initial latent phase was observed in which EPS concentration did not vary significantly. However, EPS levels increased exponentially after 40 days of operation at an SRT of 8 days, and after 70 days when the MBR was operated at 20 days SRT. No change in EPS levels was observed during the 80 days of operation at 80 days SRT. For another MBR operated at infinite SRT, no significant changes in SMP concentration during 100 days of operation were observed, over which time period the MLSS increased from 1.8 to 4.5   g/L (Jinhua, Fukushi, & Yamamoto, 2006). In a further study, following a latent phase of 30 days, MLSS and SMP levels started to increase significantly and stabilized after 140 days of operation at infinite SRT, whereas EPS levels increased continuously from the start but also stabilized after 140 days (Gao, Yang, Li, Wang, & Pan, 2004a). Nagaoka and Nemoto (2005) observed an increase in MLSS concentration from 4 to 14   g/L over 100 days along with a steady increase in EPS (from 50 to 250   mgTOC/L). There therefore appears to be no distinct pattern regarding foulant species generation and start-up, other than a general trend of more stable foulant levels at longer SRTs.

The generation of MBR foulants arising from changes in salinity has been studied by Reid, Liu, and Judd (2006), and the literature on the CASP effects date back to the 1960s (Ludzack & Noran, 1965; Tokuz & Eckenfelder, 1979). Reports indicate changes in salinity to have a greater impact on biotreatment efficacy, as manifested in the outlet organic carbon concentration, than high salinity levels per se. According to Reid et al., SMP and EPS turbidity, EPSp and SMPc all increased when a shock load of sodium chloride was administered to an MBR in a way designed to mimic saline intrusion in coastal MBRs. As with other studies (Section 2.3.6.5), permeability decline correlated with SMPc.

Finally, seasonal variations of the environment are also expected to affect MBR performances. A long-term study revealed the buffering effect of long SRT on fouling behaviour. Although the fouling reversibility was observed to vary at short SRT of 13 days (i.e. greater fraction of irreversible fouling at high temperature), the impact of temperature variations on fouling was not observed for an SRT of 50 days (Miyoshi et al., 2009).

Read full chapter

URL:

https://www.sciencedirect.com/science/article/pii/B9780080966823100022

Membrane operations in wastewater treatment: complexation reactions coupled with membranes, pervaporation and membrane bioreactors

A. Cassano , ... R. Molinari , in Handbook of Membrane Reactors: Reactor Types and Industrial Applications, 2013

19.2.2 Complexation - ultrafiltration

The separation of solutes with ionic dimensions can be accomplished by using the reverse osmosis operation, but this will result in high operating costs, low permeate flow rate and low ion selectivity. In order to overcome these problems CP - UF, also named polymer-assisted ultrafiltration (PAUF), using a water-soluble polymer as complexing agent, was introduced ( Argurio, 2002). The idea of the PAUF process is that UF can be used for removal of ions from aqueous streams, providing they are preliminarily bound to water-soluble polymers (Juang and Chiou, 2001; Molinari et al. 2008). The unbound ions pass through the membrane, whereas the polymers and their complexes are retained (Fig. 19.2). For this method to be economically feasible, the complexing agent should be regenerated by releasing the metal from the metal complex, for example, by decreasing the pH of the solution and subsequent UF. The retentate obtained by this filtration step is a complexing agent rich phase, that can be recycled at the complexation step, while the permeate is a concentrate solution of metal salts, which can be reused in the industrial process or further processed for metal recovery (Petrov and Nenov, 2004). Besides, in order to achieve a separation process that will be successful, the polymer should meet the following requirements: good solubility, high selectivity, chemical and mechanical stability, low toxicity, high molecular weight with low viscosity and low cost (Geckler and Volchek, 1996).

19.2. Schematic representation of the CP - UF separation technique (particles in the wash solution (black) are retained in the retentate binding to the complexing agent while the others (white) pass in the permeate).

Low molecular weight species, such as metal ions, can be bound to macromolecules by intermolecular forces, mainly ionic interaction and complex binding, or the combination of both.

Formation of complexes is significantly more selective than ionic interactions. An example of this binding mechanism is the complexation reaction among the polymeric agent (PEI), the proton (H⁺) and the metal cation (Cu^{2   +}) that is represented by the equilibrium equations:

[19.3] $\text{PEI}+n{\text{H}}_2\text{O}\rightleftarrows {\text{PEIH}}_n^{n+}+n{\text{OH}}^{-}$

[19.4] $\text{PEI}+a{\text{Cu}}^{2+}\rightleftarrows {\text{PEICu}}_a^{2a+}$

where $0\le n\le \overline{n}$ and $0\le a\le \overline{a}$ with $\overline{n}$ equal to the number of monomers contained in a single polymeric chain and ā represents the maximum complexation ratio of the polymers with copper ions ( $\overline{a}=\overline{n}/4$ for PEI-Cu complex because of Cu^{2   +} tetra-coordination with four nitrogen of PEI).

The water-soluble polymers polyacrylic acid (PAA) and polyacrylic acid sodium salt (PAASS) interact with copper cation by ion-exchange reactions described by the following equations:

[19.5] $\text{PAA}+n{\text{Cu}}^{2+}\rightleftarrows {\text{PAACu}}_n+2n{\text{H}}^{+}$

[19.6] $\text{PAASS}+m{\text{Cu}}^{2+}\rightleftarrows {\text{PAACu}}_m+2m{\text{Na}}^{+}$

The ionic interaction mechanism of the above equations has low selectivity, and the disadvantage to release another ion (H⁺ or Na⁺ in this case) in the feed solution, in order to remove an ion from an aqueous solution another one must be released. In contrast, reactions such as [19.4] does not present this disadvantage.

Read full chapter

URL:

https://www.sciencedirect.com/science/article/pii/B9780857094155500196

High-Salinity Pressure Retarded Osmosis Using Seawater Reverse Osmosis Brine

Sangho Lee , ... Yong-Gyun Park , in Membrane-Based Salinity Gradient Processes for Water Treatment and Power Generation, 2018

11.5.2 Draw Solution Concentration

The concentration of draw solution (c _draw ) is an important factor affecting the flux and power density of PRO systems. As reported previously [14,32] , the permeate flow rate increases as c _draw increases, which is attributed to an increase in the osmotic pressure. This also results in an increase in the power density of the PRO membrane module. However, the flux behavior is nonlinear with respect to c _draw , due to the effect of external concentration polarization on the feed water side [26]. An increase in c _draw leads to an increase in both flux and external concentration polarization.

An example of the effect of draw solute concentration on the power density is shown in Fig. 11.4. These are the results from a pilot-scale study using the 4-inch PRO membrane modules [8]. Total dissolved solids (TDS) of the draw solution ranged from 50,000 to 70,000   mg/L, while that of the feed solution was 400   mg/L. As expected, the power density increases by 4.5%–35% with an increase in TDS of the draw solution from 50,000 to 70,000   mg/L. The maximum power densities were 4.2   W/m² for the draw solution of 50,000   mg/L and 5.8   W/m² for the draw solution of 70,000   mg/L, respectively. These power densities are close to the value that allows the economic feasibility for a PRO system (~5   W/m²) [18,19].

Figure 11.4. Effect of draw solution concentrations on power density of PRO module.

Read full chapter

URL:

https://www.sciencedirect.com/science/article/pii/B9780444639615000110

26th European Symposium on Computer Aided Process Engineering

A. Sharma , ... M. Fikar , in Computer Aided Chemical Engineering, 2016

3 Parameter Estimation

In this section, the parameters of limiting flux model and the parameters of the four fouling models described above are estimated. Several experiments were performed and one of the experimentally obtained permeate flow rate data w.r.t. increasing concentration of lactose and time as depicted in Figure 2 is used here to perform the estimation. The minimization of the sum of squared differences between experimental flux data (J _exp), and estimated flux model (J _est) can be formulated as:

Figure 2. Comparison of estimated four fouling models, limiting flux model (with no fouling), and experimental data.

(7a) $\underset{K,k,c_{\text{lim}}}{min}{\displaystyle \sum _{j=1}^m{\left(J_{j,exp}-J_{j,\mathrm{e}\mathrm{s}\mathrm{t}}\right)}^2}$

(7b) $\mathrm{s}.\mathrm{t}\text{.}$

(7c) $\frac{\mathrm{d}{c}_1}{\mathrm{d}t}=\frac{c_1^2}{c_{1,0}{V}_0}AJ,c_1\left(0\right)=40\left[\text{g}/\mathrm{L}\right]\text{,}$

(7d) $J_0=k\text{ln}\frac{c_{\text{lim}}}{c_1},J=J\left(J_0,K,n,t\right),J_{j,\mathrm{e}\mathrm{s}\mathrm{t}}=J\left(J_0,K,n,t_j\right)\text{,}$

where m is the number of data points, and J is the permeate flux defined either by (5) or by (6). The Eq. (7c) is derived from (2a), by replacing V(t) from (3). The volume of the processed solution in the beginning of the operation is 0.03   m³. Based on technological considerations, the three estimated parameters (K, k and c _lim) are expected to lie within the intervals K ∈ [0, 1000] units, k ∈ [0,10]m/h, c _lim ∈ [0,1500]g/L. The experimental measurements show that the flow rate of permeate decreases with time, because of the gel-polarization layer formed on the membrane surface, and due to the fouling of membrane.

Non-linear least-squares estimation was performed to identify the values of the parameters k, c _lim of the limiting flux model (4) and the fouling rate constant K for all the four fouling models. The linear least-squares method (Foley, 2013) was also used to estimate the limiting flux model parameters k and c _lim assuming no fouling. A non-linear estimation of the parameters (k, c _lim) of limiting flux model without fouling was also done for comparison, and they were estimated to be: k  =   0.0066   m/h and c _lim  =   880.97   g/L. These values as seen in Table 1 are analogous to the linearly estimated limiting flux model. All the optimization problems were solved in MATLAB using the SQP solver implemented in the function fmincon. MATLAB function ode45 was used for numerical solution of the initial value problem (7c) – (7d).

Table 1. Comparison of estimated values of K, k and c _lim for different fouling models, and limiting flux model, along with the value of least squares criterion, $f={\sum }_{j=1}^m{\left(J_{j,exp}-J_{j,\mathrm{e}\mathrm{s}\mathrm{t}}\right)}^2$ .

model (n)	K	k  ×   10−   2[m/h]	c lim[g/L]	f  ×   10−   5[m/h]
cake filtration (0)	494.14 [s/m2]	1.30	210.43	3.43
intermediate blocking (1)	33.47 [1/m]	0.76	880.89	1.36
internal blocking (1.5)	$2.59\left[1/\sqrt{\mathrm{s}}\right]$	0.74	880.98	1.91
complete blocking (2)	0.19 [1/s]	0.72	880.97	2.55
limiting flux (–)	–	0.66	880.97	5.98

Figure 2 shows the comparison between experimental data, limiting flux model, and the four fouling models. It can be observed that the performance of the limiting flux model is the worst as it does not account for fouling. On the other hand, all four fouling models fit the data reasonably well. The cake filtration model with n  =   0 is estimated to be linear w.r.t. time as seen from the figure, and hence does not fit the experimental data with high precision. The other three fouling models are of non-linear nature and all fit the experimental data with satisfactory precision. This similarity of the models suggests that the fouling behavior could occur due to nanofiltration being a higher pressure based separation process. It is a well-known phenomenon that the fouling in the form of pore blocking increases with increasing pressure for membrane processes operated in cross-flow mode, and higher pressures tend to foul the membrane internally rather than externally on the surface due to higher sweep-off in-flow rate. The cake filtration fouling model, on the other hand, states fouling on the surface of the membrane by forming a layer of solutes, which is quite prominent in dead-end membrane separation rather than in cross-flow filtration. The other three fouling models account for blocking of membrane pores by solutes too, and hence fit the experimental data more precisely.

Table 1 provides estimated values of all parameters. The value of the objective function qualifies the intermediate fouling model (Figure 2) as the best fit for the experimental case studied here. The study done on nanofiltration of water in Chang et al. (2011) suggested the same model defining the behavior of fouling. Note also the comparison of different values for the parameters k, and c _lim of the limiting flux equation with the cake filtration fouling model, to other three models. This also points to appropriateness of the cake filtration model. On the other hand, the limiting flux parameters estimated for other three fouling models are in a very close proximity of linearly and non-linearly estimated limiting flux model without fouling.

Read full chapter

URL:

https://www.sciencedirect.com/science/article/pii/B9780444634283500308

Desalination technologies and their working principles

Marc Rosen , Aida Farsi , in Sustainable Energy Technologies for Seawater Desalination, 2022

2.2.4.1 Specific energy consumption of RO process

The SEC of a RO process can be defined as the energy consumption per volume of permeate. SEC mathematically is a function of the pressure at the feed and permeate sides and also the recovery ratio (ratio of permeate flow rate divided by feed flow rate). The recovery ratio depends on the feed salinity, temperature, efficiencies of devices (e.g., pumps, energy recovery devices (ERDs)), design configuration (single-stage, two-stage, etc.), pretreatment and brine recovery considerations. The higher is the feed salinity, the higher is the osmotic pressure, which results in a higher SEC for RO desalination. In addition, temperature affects membrane physiochemical properties. As the temperature of the feed solution increases, both water and salt permeability coefficients and osmotic pressure increase. Thus, increasing the feed temperature throughout the range of interest (e.g., 15–40°C) results in inconsistent effects on the SEC. For instance, for a low-salinity fluid-like brackish water, increasing the feed temperature from 15°C to 40°C results in a decrease in the SEC of the RO process, while for higher feed salinities, increasing the feed temperature up to only 25°C might be beneficial in terms of the SEC of the RO process ( Koutsou, 2020).

The SEC of industrial seawater RO plants is typically within the range of 2.5–4 kWh/m³ (Voutchkov, 2018). The recovery ratio varies from 50% to 85% for the brackish water RO and 35% to 50% for seawater RO (Greenlee, 2009). The feed pump pressure is typically 6–30 bar for brackish water and 55–80 bar for seawater (Okamoto, 2019). The salt rejection in both brackish water and seawater systems is higher than 95%. The recovery design parameter should be selected in a balanced fashion, to balance the purity of the permeate and the concentration of the brine stream. A higher recovery ratio indicates a lower brine volume that needs to be discharged at the expense of a lower purity of permeate.

Read full chapter

URL:

https://www.sciencedirect.com/science/article/pii/B9780323998727000097

Hybrid membrane systems – applications and case studies

Rajindar Singh , in Hybrid Membrane Systems for Water Purification, 2005

3.2.1 Seawater desalination

MSF desalination had been the mainstay for producing potable water from seawater in western Asia and the arid regions of the Mediterranean since the early 1960s. During the last 20 years, however, SWRO membrane desalination has made significant inroads gradually becoming the dominant technology. ⁴² For example, SWRO plants with total capacity exceeding 100,000 m³/day (>26 MGD [million gallons/day]) provide more than 60% of water needs in the Mediterranean nation of Malta (population ~0.5 million). ⁴³ SWRO plants exceeding 200,000 m³/day (~53 MGD) are now commonplace. In the United States, SWRO desalination is being increasingly used in California and Florida. ⁴⁰ , ⁴¹

Membrane desalting costs have dropped nearly 50% in the last 15 years due to higher performance membranes, improved process controls, and lower energy consumption with the deployment of energy recovery turbines. Process flow scheme of a typical hybrid SWRO plant with 35,100 mg/l feed water TDS operating at 60 bar g and 50% recovery is illustrated in Figure 3.34. The complexity of the seawater desalination process increases with the product water requirements: TDS <400 mg/l; chloride level <100 mg/l; and boron level of <0.5 mg/l TDS. ⁴¹

Figure 3.34. A seawater RO plant process flow schematic. The membrane plant uses boreholes for raw seawater intake. Sulphuric acid and 5.0 μm cartridge filter are used for pretreatment. Post-treatment includes decarbonation, lime addition for pH adjustment and sodium hypochlorite injection for chlorination. The plant includes an automatic system for flushing the membranes.

Source: Rahman

A new potential technology for desalination is membrane distillation (MD), which is a thermally-driven membrane process that uses hydrophobic microporous membranes similar to the ones used in osmotic distillation discussed earlier in this chapter. MD combines the advantages of conventional thermal processes with those of membrane processes. However, fouling affects the performance quite adversely. ⁴⁴ One possible solution is to use an integrated membrane system using MF/UF pretreatment followed by RO and MD. Since, RO seawater desalination product water dissolved solids content is typically 300–400 mg/l (as compared to about 10 mg/l with thermal distillation), post-treatment of RO product water by MD is an attractive option. A product water recovery of 87.6% was achieved. ⁴⁵

Case Study I. The process flow diagram of a SWRO plant is shown in Figure 3.35. The plant design and operating conditions are as follows:

Figure 3.35. Engineering process flow diagram of a 42 m³/h seawater RO membrane plant integrated with a hydraulic energy recovery turbine.

Source: Desalination 82 (1991), 31

• Feed water flow rate:	94 m3/h at 69 bar g
• Permeate flow rate:	42 m3/h
• Product water recovery:	45%
• Feed water TDS:	38,000 mg/l
• Product water TDS:	325 mg/l at > 99% salt rejection
• Energy consumption:	5.8–6.0 kWh/m3

The feed water source is a seawater infiltration well located 100 m from the shoreline. The well is dosed with sodium bisulphite (SBS) for disinfection (30–40 ppm). Seawater is transferred from the well to three media filters operating in parallel (one filter on standby) by two centrifugal pumps. Anti-scalant sodium hexametaphosphate (SHMP) is injected upstream of the 5.0 μm cartridge microfilter (MF) in the media filters effluent to prevent precipitation of magnesium sulphate. The SHMP dosage is based on a 5–10 ppm level in the RO feed water. Coagulant dosage upstream of the sand filters is 1–2 ppm, if required. The silt density index (SDI), an indicator of colloidal fouling, was consistently in the range of 0.5–1.0, indicating the feed seawater is very clean. This is because the seawater feed intake is a well.

The RO feed water is transferred by the high-pressure pumps (one pump is in operation, and the other is on standby) to the RO membrane unit. The high-pressure, multistage, submersible centrifugal pump is connected to an energy recovery turbine (ERT). The pump-motor-turbine-set enables the recovery of energy from the high-pressure brine reject stream. The high-pressure brine drives the turbine (reverse-running pump) mounted on a common shaft with the electric motor and the RO high-pressure pump. The high-pressure feed seawater is piped to the RO vessel rack. The membrane unit is a two-stage array (12:6); stage one consists of 12 pressure vessels in parallel, and stage two contains six pressure vessels also in parallel. The rejected brine is piped to the ERT from which it discharges to the sea.

Each pressure vessel contains six polyamide TFC seawater membrane (Film-Tec 30) spiral-wound elements, 8-in. diameter × 40 in. long, producing 1000 m³/day (~0.25 MGD). The feed seawater TDS is about 38,000 mg/l. Since 45% is recovered as product water with a TDS of 325 mg/l, the TDS of the rejected brine is 69,000 mg/l, and the osmotic pressure of the rejected brine stream is nearly 54 bar. This means the effective driving force at the end of the RO unit is only 16 bar, since the feed pressure is 70 bar g.

Case Study II. The process and instrumentation diagram (P&ID) of a SWRO system is shown in Figure 3.36. The one-pass RO unit is a single-stage array (6:0) consisting of six parallel pressure vessels with each pressure vessel containing six spiral-wound TFC membrane elements connected in series. ⁴² The permeate flows to the product water storage tank. High-pressure reject flows to the hydraulic ERT before it flows back to the sea. The ERT transfers the recovered energy to the high pressure RO pump. The system design conditions are as follows:

Figure 3.36. A P&amp;ID of a dead-end cartridge microfiltration pretreatment system for a seawater RO plant.

Source: Rahman

• Feed water flow rate:	60 m3/h at 75 bar g
• Permeate flow rate:	21 m3/h
• Product water recovery:	35%
• Reject flow and pressure:	39 m3/h at 72 bar g
• Product water TDS:	300 mg/l at >99% salt rejection.

The feed water pretreatment for this SWRO plant in the Mediterranean is minimal; multimedia filtration for removing particulate matter larger than 20.0 μm, acid injection for reducing pH to less than 6.0 to prevent calcium carbonate scaling, sodium metabisulphite injection for dechlorination, and 5.0 μm cartridge filtration for protecting the RO modules from fouling. Post-treatment includes a product water drawback tank to protect the membranes after shutdown, and addition of lime solution to the product water for raising the pH to 7.5–8.0 to prevent corrosion of downstream piping and equipment since the pH of the permeate is typically in the range of 5–6.

Figure 3.36b. A P&ID of a 94 m³/h seawater RO one-pass single-stage membrane system.

Source: Rahman

Read full chapter

URL:

https://www.sciencedirect.com/science/article/pii/B978185617442850004X

monahandancle.blogspot.com

Source: https://www.sciencedirect.com/topics/engineering/permeate-flow-rate

Share this post