170 Solid/Liquid Separation: Scale-up of Industrial Equipment

4.3.1.9 Diafiltration
Diafiltration is a process used to reduce, remove or exchange salts and
other small molecules from a process fluid. The process may be con-
ducted continuously or discontinuously. In discontinuous diafiltration,
the process fluid is diluted typically by a factor of 2 with the
replacement fluid (buffer). It is then brought back to the starting level
by concentration. This process may be repeated several times. Each
subsequent step reduces the concentration of the smaller passing
molecules (e.g. salts) on the upstream side of the TFF membrane
device.
Continuous diafiltration requires addition of diafiltration solution at
the same rate as the filtrate flow rate. The process volume remains
constant. The addition of a diafiltration solution equal to the volume of
the process fluid when the diafiltration started is referred to as one
diafiltration volume (1 DV).
The number of diafiltration volumes required depends on how much
salt or other small molecules need to be removed and the permeability
of the molecule through the membrane (for either continuous or
discontinuous diafiltration).
In a process that requires both concentration and diafiltration,
diafiltration can be performed either before starting to concentrate,
after completing concentration or part way through concentration and
then completing the concentration after diafiltration.
Diafiltration, when a stage in a larger process operation, is often
performed after other stages as the final volume is usually reduced. For
example, if there is 10 litres of process fluid that must be concentrated
5 times and diafiltered 5 times, then if diafiltration is performed first, it
will require the passing of 50 litres through the membrane but only 10
litres if diafiltration is performed after the concentration. However, the
flux for each case will be different. The flux for the concentrated
solution will be lower than for the starting process fluid. It is necessary
to know what the fluxes are because it may take longer to diafilter the
smaller volume. It should be noted that in some cases diafiltration may
be done first to remove components that would cause the product to
denature or precipitate when concentrated if they were not removed.
To determine the optimum concentration at which to perform
diafiltration we first obtain the CQ value as discussed above. This is a
theoretical maximum concentration for the fluid, and does not
represent a working condition. It has been determined (Ng et al, 1976)
that the effective optimum diafiltration concentration is given by the
following equation;
 4 • Membrane filters - microfiltration and ultrafiltration 171

C . = ^ = - ^ = 0.368Q
^ e 2.718 ""
For many applications this optimum concentration to carry out diafilt-
ration is never reached but the value should be a target to approach.

4.3.1.10 Cleaning and storage

TFF devices and membranes can be reused many times. Manufacturers
recommendations for chemicals, maximum concentrations and
temperatures should be followed to recover the membranes. Typically
caustic or acid solutions are used to remove organic and inorganic
fouling materials from the surface of the separation media. These
techniques are used widely in the high volume applications such as
water processing and will be discussed later. For some devices, steam-
in-place can also be considered for certain critical applications (such as
virus production).

4.3.1.11 Scale-up
As with all scale-up procedures, it is vital to ensure that sample fluids
are representative of the process fluids. However, with TFF systems it
is important to consider the following testing conditions:

• The sample volume to membrane area ratio should be close to that

expected for the final process
• Increase membrane area in direct proportion to the increase in
sample volume
• Flow path lengths must be constant
• Increase volumetric flow rate in direct proportion to the increase in
membrane area
• Operate the process at the TMP developed at smaller scale.

If these conditions are met, then the operating parameters developed and
optimised for the process during the bench or pilot scale can be scaled
directly to a full-scale operation. Filtrate flux and process time will match
those determined in the small-scale test work. For devices such as hollow
fibres and to some extent ceramics, where the path length tends to
increase with increasing membrane area, further tests are normally
required at intermediate scale to determine optimum operating conditions.
As in the case of the DFF membranes discussed earlier, to simplify
scale-up, manufacturers often offer several different cassette formats.
For example, the Centramate^^ (area = Ift^) has a 17 cm path length.
172 Solid/Liquid Separation: Scale-up of Industrial Equipment

This scales directly to the Centrasette'^^ (area = 5ft^), which also has a
path length of 17 cm.
To further demonstrate this scale-up principle, Figure 4.29 shows the
Centramate^WCentrasette^^ range of holders. It is possible to run trials
with a 0.093 m^ membrane area on the Centramate^^ and then scale up
directly to a maximum installed membrane area of 20 to 80 m^ on the
Centrasette^^ and Centrastak^^ holder.

Figure 4 . 2 9 Centramate™/Centrasette™ C5/Centrasette™ CIO holders

Table 4.4 shows actual data obtained for three different volume stages
of a scale-up evaluation. From the data, it can be seen that despite a
membrane area increase from 0.1 m^ to 8 m^, the flux was directly
scaleable, remaining constant at about 145 LMH.
Table 4 . 4 Scale-up volumes and membrane areas

Scale (1) 10 80 600

Flux (LMH) 146 144 144
Area (m^) 0.1 1 8

4.3.1.12 Calculation of membrane surface area

Trial data as determined above can be used to calculate the required
membrane area to process any volume of sample in a specified time.

J
At
where J is the filtrate flux (LMH), V is the volume filtered (litres), t is
the filtration time (hours), and A is the filter area (m^).
 4 • Membrane filters - microfiltration and ultrafiltration 173

1. Perform optimisation on a small sample to determine average flux

for each step. The example below is for a three stage process
consisting of initial concentration -» diafiltration -^ final
concentration.
V
J _ stepX

V
J _ ^ step!

V
J step?)

2. Calculate membrane area required based on the desired processing

time:

t,j

^V V V ^
A-^ ^ StepX ^ step! ' step3

\ ^step 1 ^step! ^step3 J

I -T' I1 l~ I 'J "l L -J

Example Determine the membrane area required to process a 100 litre

batch in 4 hours. The process consists of an initial concentration step
(2x), followed by diafiltration (5 DVs) and then a final concentration
step (2.5x) to reach the desired final VCF of 5x. Small scale tests were
performed on a sample of material to determine the average fluxes
below:

Small-scale test data:

Step 1 Concentrate 2x from 100 litres to 50 litres:
Filtrate Volume = 50 litres
Flux: 71 LMH

Step 2 Diafilter using 5 DV, volume increases to 50 x 5 litres:

Filtrate Volume = 250 litres
Flux: 64 LMH
174 Solid/Liquid Separation: Scale-up of Industrial Equipment

Step 3 Concentrate 2.5x from 50 litres to 20 litres:

Filtrate Volume = 30 litres
Flux: 38LMH

50 250 30
= 1.35 m '
4 7 1 ^ 64 ^ 3 8

Therefore the membrane area required is 1.35 m^.

4.3.2 industrial MF
Water-based industrial wash fluids containing detergents are used to clean
components before assembly in many industries including the automotive
and aerospace industries. With fresh water and waste costs increasing
year by year, recycling fluids for reuse is an important process to save and
reduce costs while improving the wash cleaning performance.
Tangential flow MF systems are used to recycle the water-based wash
fluid by removing the contaminants from the fluid without stripping
the original detergent from the process. The contaminants in the
process are oil, including free and emulsified oil, bacteria, fungus and
suspended solids; all these can adversely affect the wash performance.

4.3.2.1 System scale-up

The scaling up process aims to allow accurate sizing of a full-scale
system to achieve control of the contaminant level in the wash fluid.
The consequences of errors or inaccurate scale-up could be cata-
strophic. An undersized system could exhibit lower than expected
permeate fluxes, high contaminant levels or increased waste fluid
volumes while an oversized system would be too expensive resulting
in poor cost payback.
Therefore small-scale trials that can be accurately scaled are required
to assess the application and size a production unit for a water-based
industrial wash system.
Initial laboratory feasibility assessment should be undertaken to
determine if the fluid can be processed at a reasonable permeate flux to
allow a system to be sized and evaluated for payback analysis to be
completed.
The laboratory tests will also assess if any interactions exist between
additives in the cleaning fluids and the microfiltration membranes. The
cleaning additives should pass through the membrane and the cleaning
 4 • Membrane filters - microfiltration and ultrafiltration 175

fluids properties should be unaffected by the tangential flow filtration

system. The fluid properties that should be checked are:

• Fluid pH
• Detergent/additive concentration
• The corrosion protection properties of the fluid on components
• The foaming properties of the fluid.

The optimum membrane rating should be determined as the use of

finer rated membranes such as UF can result in depletion of the
surface-active additives in some cases, by either physical or chemical
means. For example, anti-foams, these can be silicone, calcium or oil-
based or other surface-active agents, can affect the filtration mechanisms
dramatically.
Preparation for testing should include a detailed site assessment on the
washes current performance and fluid usage. Then the type and volume
of sample should be established. Often, a process will be sensitive to
time or other operational factors. In this case a sample of fluid from the
wash should be taken when the system is operational rather than static.
Again, the laboratory test results will only be as accurate as the sample
taken; this should be as representative as possible of the fluid quality
from the wash.
The fluid, being water-based, must be transported quickly to ensure
extremes of temperature do not affect the fluid quality. For example, in
hot ambient conditions ~30°C bacteria and fungi can rapidly multiply,
affecting the sample properties. Conversely, very cold conditions could
cause the sample to freeze thereby destabilising oil emulsions.

4.3.2.2 Laboratory test equipment

The laboratory test apparatus consists of the equipment arranged in
similar fashion to that described in above. The schematic layout is
again broadly the same as that presented in Figure 4.18 while the
apparatus is illustrated photographically in Figure 4.30.
The pump recirculates flow of the sample across the MF membrane
surface, usually delivering a velocity of between 1 and 6 m s"^
Typically, 1 to 3 m s~^ would be used for polymeric membranes and 3
to 6 m s~^ for inorganic or ceramic membranes.
Flow and pressure measurement can be manual at this stage of testing
using measuring cylinder and stopwatch to measure permeate flows
and calibrated gauges to determine pressures.