Designation: D6908 − 06 (Reapproved 2010)

Standard Practice for

Integrity Testing of Water Filtration Membrane Systems1
This standard is issued under the fixed designation D6908; the number immediately following the designation indicates the year of
original adoption or, in the case of revision, the year of last revision. A number in parentheses indicates the year of last reapproval. A
superscript epsilon (´) indicates an editorial change since the last revision or reapproval.

1. Scope D5173 Test Method for On-Line Monitoring of Carbon

1.1 This practice covers the determination of the integrity of Compounds in Water by Chemical Oxidation, by UV
water filtration membrane elements and systems using air Light Oxidation, by Both, or by High Temperature Com-
based tests (pressure decay and vacuum hold), soluble dye, bustion Followed by Gas Phase NDIR or by Electrolytic
continuous monitoring particulate light scatter techniques, and Conductivity
TOC monitoring tests for the purpose of rejecting particles and D5904 Test Method for Total Carbon, Inorganic Carbon, and
microbes. The tests are applicable to systems with membranes Organic Carbon in Water by Ultraviolet, Persulfate
that have a nominal pore size less than about 1 µm. The TOC, Oxidation, and Membrane Conductivity Detection
and Dye, tests are generally applicable to NF and RO class D5997 Test Method for On-Line Monitoring of Total
membranes only. Carbon, Inorganic Carbon in Water by Ultraviolet, Persul-
fate Oxidation, and Membrane Conductivity Detection
1.2 This practice does not purport to cover all available D6161 Terminology Used for Microfiltration, Ultrafiltration,
methods of integrity testing. Nanofiltration and Reverse Osmosis Membrane Processes
1.3 The values stated in SI units are to be regarded as D6698 Test Method for On-Line Measurement of Turbidity
standard. No other units of measurement are included in this Below 5 NTU in Water
standard. E20 Practice for Particle Size Analysis of Particulate Sub-
1.4 This standard does not purport to address all of the stances in the Range of 0.2 to 75 Micrometres by Optical
safety concerns, if any, associated with its use. It is the Microscopy (Withdrawn 1994)3
responsibility of the user of this standard to establish appro- E128 Test Method for Maximum Pore Diameter and Perme-
priate safety and health practices and determine the applica- ability of Rigid Porous Filters for Laboratory Use
bility of regulatory limitations prior to use. F658 Practice for Calibration of a Liquid-Borne Particle
Counter Using an Optical System Based Upon Light
2. Referenced Documents Extinction (Withdrawn 2007)3
2.1 ASTM Standards:2
3. Terminology
D1129 Terminology Relating to Water
D2777 Practice for Determination of Precision and Bias of 3.1 Definitions:
Applicable Test Methods of Committee D19 on Water 3.1.1 For definitions of terms used in this practice, refer to
D3370 Practices for Sampling Water from Closed Conduits Terminologies D6161 and D1129.
D3864 Guide for On-Line Monitoring Systems for Water 3.1.2 For description of terms relating to cross flow mem-
Analysis brane systems, refer to Terminology D6161.
D3923 Practices for Detecting Leaks in Reverse Osmosis 3.1.3 For definition of terms relating to dissolved carbon
and Nanofiltration Devices and carbon analyzers, refer to D5173, D5904 and D5997.
D4839 Test Method for Total Carbon and Organic Carbon in 3.1.4 bubble point—when the pores of a membrane are
Water by Ultraviolet, or Persulfate Oxidation, or Both, and filled with liquid and air pressure is applied to one side of the
Infrared Detection membrane, surface tension prevents the liquid in the pores
from being blown out by air pressure below a minimum
1
This practice is under the jurisdiction of ASTM Committee D19 on Water and pressure known as the bubble point.
is the direct responsibility of Subcommittee D19.08 on Membranes and Ion 3.1.5 equivalent diameter—the diameter of a pore or defect
Exchange Materials.
Current edition approved May 1, 2010. Published May 2010. Originally
calculated from its bubble point using Eq 1 (see 9.3). This is
approved in 2003. Last previous edition approved in 2006 as D6908 – 06. DOI: not necessarily the same as the physical dimensions of the
10.1520/D6908-06R10. defect(s).
2
For referenced ASTM standards, visit the ASTM website, www.astm.org, or
contact ASTM Customer Service at [email protected]. For Annual Book of ASTM
3
Standards volume information, refer to the standard’s Document Summary page on The last approved version of this historical standard is referenced on
the ASTM website. www.astm.org.

Copyright © ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959. United States

1
 D6908 − 06 (2010)
3.1.6 integrity—measure of the degree to which a mem- upper end of the UF pore size range (0.01 µm and larger pore
brane system rejects particles of interest. Usually expressed as sizes) due to insignificant or inconsistent removal of TOC
a log reduction value (LRV). material by these membranes.
3.1.7 log reduction value (LRV)—a measure of the particle 4.2 These methods may be used to identify relative changes
removal efficiency of the membrane system expressed as the in the integrity of a system, or used in conjunction with the
log of the ratio of the particle concentration in the untreated equations described in 9.4, to provide a means of estimating the
and treated fluid. For example, a 10-fold reduction in particle integrity in terms of log reduction value. For critical
concentration is an LRV of 1. The definition of LRV within this applications, estimated log reductions using these equations
Standard is one of many definitions that are used within the should be confirmed by experiment for the particular mem-
industry. The user of this standard should use care as not to brane and system configuration used.
interchange this definition with other definitions that poten- 4.3 The ability of the methods to detect any given defect is
tially exist. The USEPA applies the LRV definition to patho- affected by the size of the system or portion of the system
gens only. tested. Selecting smaller portions of the system to test will
3.1.8 membrane system—refers to the membrane hardware increase the sensitivity of the test to defects. When determining
installation including the membrane, membrane housings, the size that can be tested as a discrete unit, use the guidelines
interconnecting plumbing, seals and valves. The membrane can supplied by the system manufacturer or the general guidelines
be any membrane with a pore size less than about 1 µm. provided in this standard.
3.1.9 multiplexing—the sharing of a common set of 4.4 The applicability of the tests is largely independent of
physical, optical, and/or electrical components across multiple system size when measured in terms of the impact of defects on
system sample points. Two approaches of multiplexing are the treated water quality (that is, the system LRV). This is
considered in this practice: sensor multiplexing and liquid because the bypass flow from any given defect is diluted in
multiplexing. Sensor multiplexing monitors a unique sample proportion to the systems total flowrate. For example, a
with a dedicated sensor. Sensors are linked to a centralized 10-module system with a single defect will produce the same
location, where data processing and the determinative mea- water quality as a 100-module system with ten of the same size
surement is performed. Liquid multiplexing uses a common defects.
instrument to measure multiple process sample streams in a
sequential manor. Samples are fed to the common analyzer via 5. Reagents and Materials
a system of a manifold, valves and tubing. 5.1 Reagents—As specified for the TOC analyzer in ques-
3.1.10 relative standard deviation (RSD)—a generic con- tion. D5173 lists requirements for a variety of instruments.
tinuous monitoring parameter used to quantify the fluctuation
5.2 Soluble Dye Solution—Use FD&C or reagent grade dyes
of the particulate light scatter baseline from a laser-based
such as FD&C Red #40, dissolved in RO permeate, or in
incident light source. As an example, the RSD may be
ASTM Reagent Grade Type IV water.
calculated as the standard deviation divided by the average for
a defined set of measurements that are acquired over a short 5.3 Light Scatter Standards—See Test Method D6698 for
period of time. The result is multiplied by 100 to express the the selection of appropriate turbidity standards. In addition,
value as a percentage and is then reported as % RSD. The polystyrene latex standards of a defined size and concentration
sample monitoring frequency is typically in the range of 0.1 to may be used in place of a turbidity standard as long as count
60 seconds. The RSD parameter is specific for laser-based concentration is correlated to instrument response.
particulate light-scatter techniques which includes particle 5.4 Light Obscuration Standards—Standards that are used
counters and laser turbidimeters. The RSD is can be treated as for the calibration of particle counters, namely polystyrene
an independent monitoring parameter. Other methods for RSD latex spheres should be used. Consult the instrument manufac-
calculations may also be used. turer for the appropriate type and size diameter of standards to
3.1.11 UCL—a generic term to represent the aggregate be used.
quantity of material that causes an incident light beam to be
scattered. The value can be correlated to either turbidity or to 6. Precision and Bias
specific particle count levels of a defined size. 6.1 Neither precision nor bias data can be obtained for these
test methods because they are composed of continuous deter-
4. Significance and Use minations specific to the equipment being tested. No suitable
4.1 The integrity test methods described are used to deter- means has been found of performing a collaborative study to
mine the integrity of membrane systems, and are applicable to meet the requirements of Practice D2777. The inability to
systems containing membrane module configurations of both obtain precision and bias data for methods involving continu-
hollow fiber and flat sheet; such as, spiral-wound configuration. ous sampling or measurement of specific properties is recog-
In all cases the practices apply to membranes in the RO, NF, nized and stated in the scope of Practice D2777.
and UF membrane classes. However, the TOC and Dye Test
practices do not apply to membranes in the MF range or the

PRACTICE A—PRESSURE DECAY AND VACUUM DECAY TESTS

2
 D6908 − 06 (2010)

NOTE 1—The last example also represents the vacuum decay test when a partial vacuum is applied to one side of the membrane.
FIG. 1 Various Configurations for the Pressure Decay Test

7. Scope decay on the isolated side of the membrane. The results of both
7.1 This practice covers the determination of integrity for the PDT and VDT are a direct measure of the membrane
membrane systems using the pressure decay test (PDT) and system integrity.
vacuum decay test (VDT). 8.2 Limitations and Applications—The tests are limited to
7.2 The tests may be used on membranes in all classes, RO monitoring and control of defects greater than about 1 to 2 µm
through MF, and are suitable for hollow fibers, tubular and flat (see 9.3, Selection of Test Pressure).
sheet (such as spiral wound) configurations. However, the PDT 8.2.1 The tests can be applied in various forms provided a
is most commonly employed for in-situ testing of UF and MF differential pressure below the bubble point is established
systems and the VDT for testing NF and RO elements and across a wet membrane with air on the relative high pressure
systems. See Practice D3923. side of the membrane. Some examples are included in Fig. 1.
8.2.2 Both the PDT and VDT are described here in their
8. Summary of Practice most common forms. In the case of the PDT this is with one
8.1 Principles—The tests work on the principle that if air side of the membrane pressurized with air and the other filled
pressure is applied to one side of an integral, fully wet with liquid vented to atmosphere. In the case of the VDT, air is
membrane at a pressure below the membrane bubble point, typically present on both sides and vacuum is applied to the
there will be no airflow through the membrane other than by permeate side.
diffusion through liquid in the membrane wall. If a defect or
leak is present then air will flow freely at this point, providing 9. Procedure
that the size of the defect is such that it has a bubble point 9.1 Pressure Decay Test (PDT)—The pressure decay test
pressure below the applied test pressure. The configurations for can be carried out by pressurizing either side of the membrane
applying air and water are shown in Fig. 1. (see Fig. 1). For complete wet-out of all the membrane in the
8.1.1 Air based tests are means of applying air, at a pressure system, the system should be operated at its normal pressure
below the membrane bubble point, to one side of a wet before the test is performed. The steps involved in the PDT are:
membrane and measuring the air flow from one side to the 9.1.1 Drain the liquid from the side of the membrane to be
other. Air flow can be measured directly, but more commonly, pressurized (referred to here as the upstream side).
it is derived from pressure or vacuum decay. In the PDT air 9.1.2 Open the downstream side of the membrane system to
flow is measured as the rate of pressure decay when one side atmosphere. This ensures air that leaks or diffuses is free to
of a membrane system (either the feed or filtrate side) is escape without creating backpressure, and establishes the
isolated and pressurized with air. In the VDT an air pressure downstream pressure as atmospheric pressure.
differential is generated by isolating one side of a wet mem- 9.1.3 Isolate and pressurize the upstream side with air to the
brane and applying a partial vacuum with atmospheric pressure test pressure. Then isolate the air supply. Do not exceed the test
on the other side. Air flow is measured as the rate of vacuum pressure as this could lead to blowing out smaller pores than

3
 D6908 − 06 (2010)

FIG. 2 Connection Arrangement for the VDT

intended resulting in a higher PDT. Record this pressure as determination of the diffusive flow, such as laboratory mea-
Ptest,max, the maximum test pressure. surements or by measuring the PDR on a system confirmed
9.1.4 After allowing time for the decay rate to stabilize suitably integral by other means. In such cases, the measured
record the initial pressure, Pi, and commence timer.4 PDR result is corrected as follows:
9.1.5 After at least 2 min, record the final pressure, Pf, and PDRcorrected 5 PDRmeasured 2 PDRdiffusion
the time taken for the pressure to decay from Pi to Pf (t). The
time period can be extended in order obtain a more accurate where:
result if the pressure decay rate is slow.5 PDRdiffusion = PDRmeasured for the integral system, at the
9.1.6 Calculate the Pressure Decay Rate (PDR) as follows same PTest and temperature.
and record the result along with the test conditions
(temperature, average test pressure Ptest,avg and maximum 9.1.8 For most practical applications of the test sufficient
pressure Ptest,max): accuracy can be obtained by taking the conservative approach
and assuming that all the pressure decay is related entirely to
Pi 2 Pf
PDRmeasured 5 leaks (PDRdiffusion = 0).
t
9.2 Vacuum Decay Test—The VDT is conducted with air on
where:
both sides of the membrane. For complete wet-out of all the
PDRmeasured = measured pressure decay rate, kPa/min at the membrane in the system, the system should be operated at its
average test pressure, Ptest,ave = Pi + Pf / 2, normal pressure before the test is performed. The steps
Pi = initial pressure, kPa gauge,
involved in the VDT are:
Pf = final pressure, kPa gauge,
t = time taken for pressure to decay from Pi to Pf, 9.2.1 Drain the liquid from the feed side of the membrane
mins, and (referred to here as the upstream side), and let it remain open
Ptest,max = maximum test pressure given as the pressure to the atmosphere. For membrane devices placed horizontally,
at the start of the test, kPa. the feed and exit ports must be located on the bottom of the
device housings in order for this to work.
9.1.7 The PDR will result from diffusion through the
membrane wall, as well as leaks through defects, damaged 9.2.2 Use the equipment connected in this order (see Fig. 2):
membranes, or seals. The diffusive component of the airflow is a vacuum pressure gauge, an isolation valve, a water trap that
not related to the integrity, so a more accurate estimate of the will not buckle at vacuum, and a vacuum pump, to the
nondiffusive pressure decay can be obtained by subtracting the permeate manifold that serves one or more membrane devices.
diffusive flow from the measured flow. The diffusive compo- Addition of another isolation valve (B) at the permeate header
nent can be estimated either by calculation or experimental allows easy connection of the equipment without disrupting
operation of the membrane system.
4
The pressure decay rate at the start of the test is usually quite high due to
9.2.3 Open isolation valves A and B and run the vacuum
displacement of some of the liquid in the membrane wall. The time taken for the pump to evacuate the permeate side until the pressure gauge
decay rate to stabilize will be different for different systems, but may take up to 3 shows a stable vacuum. The water removed during this
min. operation is collected in the water trap. Close isolation valve A.
5
Due to the nonlinear decay in pressure with time and the desire to simplify the
equations by using the first order approximation for decay rate, the maximum time Start the stopwatch and record the initial vacuum (Pi). The test
should be such that Pf is no more than 10 % lower than Pi. vacuum can be selected using the guidelines in 9.3.

4
 D6908 − 06 (2010)
9.2.4 After the determined time (60 s is a typical time, 120, where:
180 or 300 s will yield a more sensitive test) record the final ∆Ptest,max = the maximum differential test pressure applied
pressure (Pf) and the time (t) for reaching this value.5 across the membrane. This is the Ptest,max re-
9.2.5 Calculate the Vacuum Decay Rate (VDR) as follows: corded during the test corrected for any static
Pf 2 Pi
head contribution,
VDRmeasured 5 γ = surface tension at the air-liquid interface,
t
θ = liquid-membrane contact angle, and
where: d = equivalent diameter of the smallest defect in-
VDRmeasured = measured vacuum decay rate, kPa/min at the cluded in the test.
average test pressure, Ptest,ave = Pi + Pf / 2, 9.3.1 For the theoretical case of a perfectly hydrophilic
Pi = initial vacuum, kPa gauge, membrane, the contact angle is zero, and assuming water at
Pf = final vacuum, kPa gauge, 25°C (surface tension 72 dynes/cm), Eq 1 simplifies to Eq 2,
t = time taken for vacuum to decay from Pi to Pf, with d in micrometres and Ptest,max in kilopascal:
mins, and
Ptest,max = maximum test vacuum given as the pressure 288
d5 (2)
at the start of the test, kPa. ∆P test,max

9.2.6 The VDR will result from diffusion through the 9.3.2 Fig. 3 shows the relationship between test pressure
membrane wall, as well as leaks through defects, damaged and equivalent defect diameter expressed by Eq 1 and assum-
membranes, or seals. The diffusive component of the airflow is ing a surface tension of 72 dynes/cm. The solid line represents
not related to the integrity, so a more accurate estimate of the Eq 2; that is, the conservative situation of cosθ = 1. In practice
nondiffusive vacuum decay can be obtained by subtracting the most membranes used in water treatment have a contact angle
diffusive flow from the measured flow. The diffusive compo- greater than zero, which is represented by the shaded region
nent can be estimated either by calculation or experimental under the solid line in Fig. 3. If the contact angle is known or
determination of the diffusive flow, such as laboratory mea- can be determined, Eq 1 may be used. However, if the contact
angle is not known, a conservative estimate of the test pressure
surements or by measuring the VDR on a system confirmed
required can be made by applying Eq 2.
suitably integral by other means. In such cases, the measured
VDR result is corrected as follows: 9.3.3 The test pressure is usually selected to ensure that the
minimum defect diameter picked up by the test is smaller than
VDRcorrected 5 VDRmeasured 2 VDRdiffusion contaminates or particles of interest. For example, Eq 2
where: indicates that a test pressure of 100 kPa would include all
VDRdiffusion = VDRmeasured for the integral system, at the defects larger than or equal to 3 µm. A lower pressure could be
same Ptest and temperature. used for less hydrophilic membranes. For example, if the
contact angle is 60 degrees (typical for polypropylene,
If VDRdiffusion is unknown, the conservative approach is to polysulfone, or PVdF) Eq 1 indicates that defects of 3 µm
set VDRdiffusion = 0. would be included at a test pressure of 50 kPa. An even lower
test pressure may be used for larger defects, such as for
9.3 Selection of Test Pressure—The test pressure selected
example detection of broken fibers in a hollow fiber system.
determines the minimum equivalent diameter of a defect that
9.3.4 In practice the applied test pressure is rarely more than
can contribute to the pressure or vacuum decay rate. The
300 kPa, which is usually sufficient to include defects smaller
relationship between the test pressure and the equivalent defect
than most pathogens of interest. At this pressure limit the test
diameter is given by Eq 1. Defects smaller than this will be too
is not suitable for direct validation of virus rejection as these
small for the bubble point to be overcome and thus will not
particles are very small (typically less than 0.01 µm) with a
contribute to airflow. Larger defects will allow airflow as the
corresponding test pressure of several thousand kilopascals.
bubble point will be exceeded by the applied test pressure.
Details on the derivation of this equation and its use in 9.4 Interpreting PDR and VDR Results as Log Reduction
determining maximum pore size for membranes can be found Values—Both the PDR and the VDR are measurements of the
in Method E128.6 airflow from one side of the membrane to the other under a
known set of test conditions (temperature and pressure). This
4γcosθ
d5 (1) information can be used to estimate the flow of liquid through
∆P test,max
the same defects during filtration conditions. This provides an
estimate of the membrane bypass flow and thereby an estimate
of the log removal of particles for the system. One approach is
6
Eq 1 is often modified to include a correction factor referred to as the pore based on the Hagen-Poiseuille law, which assumes laminar
shape factor or the Bechold Constant. This is a value < 1 and takes into account the
flow through cylindrical defects. Whilst this method provides a
irregular shape of membrane pores. For the purpose of this practice the shape factor
is assumed to be 1 as this is the most conservative position, and the shape of any useful estimate, its applicability is limited to small fibers (<
particular defect detected by these tests is not known. 400 µm ID) where the criteria for laminar flow are more closely

5
 D6908 − 06 (2010)

NOTE 1—The solid line represents Eq 2.

FIG. 3 The Relationship Between Test Pressure and Equivalent Defect Diameter (Eq 1, Water at 25°C)

approximated. The method is described in 9.4.1 and a detailed ƒ2 = pressure correction factor = Pu,test2 − Pd,test2 / 2Patm
derivation, along with the assumptions required, is contained in TMP,
Appendix X1. An alternative method is to experimentally Qfilt = filtrate flowrate (m3/s),
measure the relationship between liquid and air flows for the Pu,test = upstream pressure during the PDT or VDT =
worst case failure mode. This is typically a broken fiber at the Ptest,avg for PDT and Patm for VDT, (kPa absolute),
pot for most hollow fiber MF or UF systems. This approach, Pd,test = downstream pressure during the PDT or VDT =
described in 9.4.3, assumes that all the measured gas flow is Patm for PDT and Ptest,avg for VDT, (kPa absolute),
due to “worst case” failures and so provides a conservative Patm = atmospheric pressure (kPa absolute),
estimate of bypass flow and LRV for the system. While these CF = concentration factor. This represents the increase
approaches have been applied in practice, data covering a in the contaminant concentration that could occur
range of different membrane configurations, test conditions, on the upstream side of the membrane relative to
and fiber diameters are not yet available. Regardless of the the feed water concentration due to the operating
chosen method the relationship between integrity test results mode. This would typically be equal to 1 for
and LRV should be verified by experiment in the field on the dead-end systems, but could be higher for cross
particular membrane and configuration used. flow or feed and bleed modes,
9.4.1 The Laminar Flow Approach Using the Hagen- PDR = pressure decay rate (kPa/s),
Poiseuille (H-P) Law—This approach assumes laminar flow VDR = vacuum decay rate (kPa/s),
through cylindrical defects and is most suitable for small TMP = transmembrane pressure during filtration (kPa),
diameter fibers (200 to 400 µm lumen diameter). A detailed Vsystem = volume pressurised (or under vacuum) during test
derivation along with key assumptions is contained in Appen- (m3),
µwater = the viscosity of the liquid during filtration (Pa·s),
dix X1. The equations required to convert the PDR and VDR
µair = the viscosity of the air during the test (Pa·s), and
results obtained using the method described here to a log LRVe = estimated log reduction value.
reduction value, are given below as Eq 3 and 4 respectively:
For PDR: 9.4.2 Example Calculation of the Log Reduction of Par-
ticles from the PDT Using the H-P Approach—Estimate the
LRVe 5 log10 S Q filt P atm
ƒ ƒ
CF·PDT·V system 1 2 D (3) LRV for a membrane system operating at a filtrate flowrate of
50 L/s and a transmembrane pressure of 70 kPa. The water
and for VDR: temperature is 20°C, and the PDR for the system is 2.5 kPa/min
LRVe 5 log10 S Q filt P atm
ƒ ƒ
CF·VDT·V system 1 2 D (4)
at 100 kPa test pressure and 27°C. The system is operating in
dead-end mode so CF = 1. The viscosity of water at 20°C is
1.00 × 10-3 Pa·s and air at 27°C is 1.84 × 10-5 Pa·s. The
where: pressurized system volume during the PDT is 400 L.
ƒ1 = viscosity correction factor = µwater / µair, First calculate ƒ1 and ƒ2:

6
 D6908 − 06 (2010)

FIG. 4 PDR Values

µ water 1.00 3 1023 (3) Evaluate the system LRV using the following:
ƒ1 5 5 5 54.35
µ air 1.84 3 1025 (a) Measure the PDR (or VDR) for the system. Calculate
2 2
P u,test 2 P d,test ~ 201.3 kPa! 2 ~ 101.3 kPa! 2
2
the gas flow using Eq 5 (for PDT) or Eq 6 (for VDT). Note that
ƒ2 5 5 5 2.13
2P atm TMP 2·101.3 kPa·70 kPa these are the equations derived as Eq X1.4 and X1.5 in
Estimate the LRV from Eq 3 as follows: Appendix X1.

LRVe 5 log10 S Q filt P atm

ƒ ƒ
CF·PDT·V system 1 2 D Q G,atm 5 PDR
V system
P atm
(5)

5log10 S 23 3
50 3 10 m /s·101.3 kPa
1·2.5/60 kPa/s·400 3 1023 m 3
·54.35·2.13 D Q G,atm 5 VDR
V system
P atm
(6)
54.5 (b) Calculate the equivalent number of broken fibers for
Note that from Eq 2 the test pressure of 100 kPa equates to the system (see Fig. 4) as:
a minimum defect size of 2.9 µm (conservatively). So the LRV Q G,atm
of 4.5 calculated above is the minimum LRV for particles N equivalent 5 (7)
Q G,atm,fiber
greater than 2.9 µm diameter. (c) Calculate the liquid bypass flow, Qbypass by multiply-
9.4.3 Experimental Approach to Correlating Test Results ing the equivalent number of broken fibers by the flow per fiber
and System LRV Using Equivalent Number of Broken Fibers— at the operating TMP (from the data generated in step 2):
This approach relies on measuring the relationship between gas
flow and bypass flow for “worst case” defects for hollow fiber Q bypass 5 N equivalent 3 Q L,fiber (8)
systems, and assuming that all bypass will be through such Eq 5 can be written for an individual fibre as QG,atm,fiber =
defects. This approach provides a conservative estimate of PDRfiberVsystem / Patm where PDRfiber is the pressure decay rate
LRV that can be applied to most membrane diameters and corresponding to QG,atm,fiber. Combining with Eq 7 and 8 gives:
configurations. For hollow fiber membrane systems the worst
case failure will usually be a fiber that is cut cleanly at the PDRcorrected
Q bypass 5 ·Q L,fiber (9)
fiber-pot interface. This provides the shortest bypass path and PDRfiber
the largest possible diameter. The steps involved are:
(d) Calculate the estimated LRV using Eq 10 (also Eq
(1) Experimentally determine the gas flow through a single
X1.2):
fiber, cut at the pot, at the selected test pressure (call this
QG,atm,fiber). Preferably this is carried out in field tests using one
or more modules of the full-scale design, or alternatively in a
LRVe 5 log10 S Q filt
Q bypass D (10)

laboratory using the same membrane fiber and potting materi-

als. Substituting Eq 9 into Eq 10:
(2) For the same configuration determine the water flow
through the lumen (QL,fiber) at a range of pressures to establish
LRVe 5 log10 S PDRfiber·Q filt
PDRcorrected·Q L,fiber D (11)

the bypass flow vs TMP curve for a single fiber. This can be
done experimentally using short fiber lengths in the laboratory, A similar derivation for VDT gives:
or by theoretical calculation combined with experimental
determination of friction factor (for turbulent flow). LRVe 5 log10 S VDRfiber·Q filt
VDRcorrected·Q L,fiber D (12)

7
 D6908 − 06 (2010)
The values for QG,atm,fiber and QL,fiber can be calculated using hollow cylinders at a filtration TMP of 50 kPa, including
known hydraulic formulae (such the Darcy-Weisbach equa- allowance for both ends of the cut fiber, gives:
tions) including consideration of entrance and exit losses, πd 4 TMP
however for nonlaminar flow situations solving these requires Q L,fiber 5
128Lµ
an iterative approach as well as establishing values for surface π ~ 250 3 1026 m ! 4 ·50 3 103 Pa 1000 L 3600 s
roughness which must be experimentally determined. When 5 · ·
128·1.62 3 1023 Pa·s m3 h
using theoretical calculation of QL,fiber, consideration should
also be given to flow through the free end of the cut fiber as S
·
1
1
1
0.125 m 1.035 m D 5 0.095 L/h
well as the pot, although in most cases this will be small
compared to the flow through the pot. Checking Reynolds number confirms this is laminar flow
9.4.4 Example of the Experimental Method Using the and hence the equation is valid. An allowance for entrance and
Equivalent Number of Broken Fibers—The following example exit losses could be made, however, given the low Reynolds
is taken from data presented in Kothari and St. Peter.7 The number this correction will be minor and the value as calcu-
filtration unit is a hollow fiber microfilter using membranes lated above is conservative.
with an internal diameter of 250 µm. Results from a study Step 3. Calculate the Relationship Between PDR and By-
looking at the impact on PDR of cutting fibers are presented. pass Flowrate—Using Eq 11 gives:
Fibers were cut near the pot, giving a cut fiber length of
approximately 125 mm, with the long end of the fiber approxi-
LRVe 5 log10 S PDRfiber·Q filt
D
PDRcorrected·Q L,fiber
mately 1035 mm. Temperature is assumed to be 5°C (viscosity
1.62 × 10-3 Pa·s), with a filtrate flow of 120 000 L/h. Data up
5log10 S 0.0702 3 120 000 L/h
PDRcorrected 3 0.095 L/h D
to 400 cut fibers is presented, although only the data up to 40
cut fibers is used here as the test pressure was reasonably
5log10 S88 674
D
PDRcorrected
constant between tests at an average of 100 kPa. 54.95 2 log10 ~ PDRcorrected!
54.95 2 log10 ~ PDRmeasured 2 0.72!
Number of PDR PDR Starting
Cut Fibers (kPa/min) Pressure (kPa) The estimated LRV’s using the above equation are tabulated
0 0.69 101.8
1 0.76 101.7 below for varying numbers of cut fibes. The LRV’s calculated
2 0.90 101.6 according to the H-P method (as described in 9.4.1) are also
6 1.10 100.9 included for comparison. The difference between the two
12 1.58 100.3
24 2.41 98.4 methods of estimating the LRV is small in this case (0.05 to 0.1
40 3.51 95.8 log). As the fiber diameter increases the limitations of the
Step 1. Determine the Relationship Between Gas Flow and assumptions involved in the H-P method will become greater,
Fibers Cut at the Pot—In order to do this the above PDR and the experimental approach might be more suitable. Particle
values are plotted producing the graph shown in Fig. 4. The count data are also included to indicate the difficulty of using
slope of the line of best fit represents the change in pressure conventional water quality methods to verify integrity at these
decay for each cut fiber, and the intercept represents the gas levels.
flow due to diffusion only (at 100 kPa test pressure). This could LRVe
Total
be converted to a gas flow using Eq 5, however for this PDT Equivalent LRVe
No. of Particle
(kPa/ Broken Fibers H-P Method
example it is more useful to leave it as a PDR per cut fiber. Cut Fibers
min) Method (see (see 9.4.1)
Count
(counts/mL)
Step 2. Determine the Liquid Flowrate Through a Single 9.4.3)
Broken Fiber at the Pot—In this case we will calculate the 0 0.69 1.40
1 0.76 6.37 6.47 1.07
flowrate from theory, although it could also be determined by 2 0.90 5.70 5.80 7.50
laboratory measurement. Using Eq X1.7 for laminar flow in 6 1.10 5.37 5.46 2.60
12 1.58 5.01 5.10 3.00
24 2.41 4.72 4.79 1.30
7
Kothari, H., St. Peter, E., “Utility Perspective on Regulatory Approval for 40 3.51 4.50 4.55 2.30
Microfiltration Treatment Facilities in Wisconsin,” Proceedings of AWWA Annual
Conference, June 11-15 2000, Denver, CO.

PRACTICE B—USE OF TOTAL ORGANIC CARBON ANALYZERS FOR MONITORING INTEGRITY OF RE-
VERSE OSMOSIS OR NANOFILTRATION MEMBRANE SYSTEMS

10. Scope 11. Summary of Practice

10.1 This practice is applicable where the membrane system 11.1 Carbon Analysis Summary—There are two processes
and water source will allow the monitoring of TOC both involved in TOC analysis—first dissolved carbon is oxidized to
upstream and downstream of the system, and at least order of CO2 and then the concentration of CO2 is detected and the
magnitude difference from the feed can be measured in the result is interpreted using a customized calibration curve. To
permeate (product) water. See D4839. eliminate interference from inorganic carbon (carbonate,

8
 D6908 − 06 (2010)
bicarbonate, and dissolved CO2) the sample is split into two tion is at least one order of magnitude. TOC monitoring, as a
streams. Both streams are acidified to convert inorganic carbon tool for monitoring integrity, is used to identify relative
(IC) to CO2, and one stream is treated further to oxidize the changes in the integrity of a system. The sensitivity of the
organic carbon to CO2. The samples are sent to separate CO2 method is dependent on:
detectors—one for IC and one for Total Carbon (TC). TOC is 12.1.1 The capabilities of TOC instrument,
the difference between the TC and IC results. D5173 and 12.1.2 The size of the system as measured by permeate flow,
D5997 give detailed descriptions of the various techniques and
used to perform on-line monitoring of carbon compounds in 12.1.3 The change in permeate TOC concentration that
water. Instruments using these methods require approximately corresponds to a significant leak.
six minutes to analyze one sample. 12.2 TOC analyzers are affected by conditions outlined
11.2 Sampling from the Permeate Stream—Practices D3370 below. For interference specific to a particular analyzer, contact
describes standard practices for sampling water from closed the manufacturer. A baseline permeate TOC level must be
conduits. A side stream from the permeate line is diverted to established within the limits of the instrument that is still
the TOC analyzer. The length of this line should be as short as significantly different from the challenge or average feed
possible. Most analyzers have a flushing cycle between concentration by one order of magnitude.
samples and by-pass during analysis, which is diverted to 12.3 The size of the system monitored by one sample point
drain. The volume of sample is very small compared to the should be determined using a risk/cost analysis. The risk is the
by-pass flow (as little as 0.35 mL/min versus 30 to 220 mL/min potential for harm or legal action if there is a leak in the
for flush). system. The cost is the price of additional sample points or
11.3 Establishing Baseline Data—When the system has additional analyzers.
stabilized after start-up, the feed, permeate and concentrate
12.4 The change in permeate TOC concentration corre-
streams are analyzed for TOC concentration. If the instrument
sponding to a significant leak (as defined by the risk/cost
used can handle the range in concentrations, with different
analysis) will depend on the volume of permeate produced by
calibration curves, then it is best to use the same instrument as
intact membrane in the monitored unit.
will be used for integrity monitoring. The instrument can be
used off line in grab sample mode for these tests. It is important 12.5 When determining the size that can be tested as a
to perform enough repeat sample analyses to ensure the sample discrete unit, consider the change in TOC concentration
lines are completely filled with the test solution. Testing the expected from a leak that should initiate action. The change
permeate sample first will make this task easier. Sample size should be greater than 3 standard deviations of the average
should be large enough to reflect normal variations due to product concentration measured for that system. Fig. 5 shows
temperature and time of day. change in permeate TOC concentration in an RO system with
11.4 Concentrate Sampling—The concentrate stream is different types of damage. The feed and concentrate concen-
tested to determine the system’s mass-balance. It may be that trations were approximately 5 and 10 mg/L, respectively.
organic carbon is adsorbing to the membrane. If so, there may
13. Interferences
be break-through later on when all adsorption sites are taken up
and a new permeate baseline will be necessary. 13.1 Changes in Inorganic Carbon Concentration—
11.5 TOC Monitoring—Follow instructions for the particu- Instability in the pretreatment acidification process can cause
lar TOC analyzer in service. Be sure to keep the power on, fluctuations in the inorganic carbon concentration of the
chemicals fresh, pre-filters clean and UV or IR sources in good permeate stream. If adjustment is not made in the acidification
working order. Become familiar with the data output for your process to drive off excess IC, then the TOC results will be
analyzer. It should provide the time, alarms, cause of the alarm, high.
alerts when analysis conditions have been changed and a 13.2 Changes in Background Conductivity—Changes in
description of the new conditions. View permeate TOC con- sample background conductivity will corrupt the comparison
centration on a graph with the feed and permeate baseline of CO2 conductivity with the calibration curve. Since TOC
concentrations marked. analyzers can be much more sensitive than conductivity
11.5.1 Decision Point—A decision point must be estab- sensors, breaches in integrity should be detected due to
lished for your particular process depending on the degree of increase in TOC concentration before there is a significant
risk associated with a breach of integrity. change in permeate conductivity.8
11.5.2 Variability—Process fluctuations, temperature, 13.3 Particulates—Particles suspended in the water stream
changes in chemical cartridges, fouling of the TOC analyzer may cause blockage in the monitor over time.
inlet pre-filter, changes in flow to the analyzer can all affect the
TOC analysis. The degree of variability depends on the process 14. Apparatus
and operation of the analyzer. The decision point should not be
reached due to normal process variability. 14.1 D5173 shows block diagrams of several designs of
on-line TOC analyzers that have been introduced successfully.
12. Significance and Use
12.1 TOC Monitoring can be used effectively when the
difference between average feed and product TOC concentra-

9
 D6908 − 06 (2010)

TOC concentration during damage events. TOC does detect damage reliably. Value for damage event B is from one sample.

NOTE 1—Error bars indicate 3 standard deviations from the average (Chapman and Linton).8
FIG. 5 Change in TOC Concentration with Different Types of Damage

15. Interpretation of Results trations. If permeate concentration exceeds three standard

15.1 Permeate and feed (or average of feed and concentrate) deviations from the average, check the system to determine the
TOC concentrations should be plotted over time. Using the cause (see Fig. 6).
feed concentration will provide the more conservative bench-
mark and simplify the procedure.
15.2 When the system has stabilized after start-up, calculate
the standard deviation of the permeate and feed TOC concen-

PRACTICE C—SOLUBLE DYE TEST

10
 D6908 − 06 (2010)

FIG. 6 Process Monitoring Chart Displaying Upper Control Limits Plotted with Monitoring Data During a Fiber Cutting Study

16. Scope of integrity.9 Alternatively, calculate the LRV from the feed and
16.1 This guide is applicable to RO and NF membrane permeate dye values (as described in Section 21), to assure the
systems, including those with spiral, tubular or flat sheet required removal is achieved.
configuration elements. The guide describes the application of 17.3 Plumbing connections and operational considerations
two soluble dyes, Red Dye # 40 and Rhodamine WT. Both should allow the system to be run 30 min in recirculation
dyes have a molecular weight of approximately 500. See mode, or alternately with continuous liquid dye injection for up
Practice D3923. 30 min and when introduction of a soluble dye will not
interfere with operation of the system for its application. The
17. Summary of Practice dye chosen must be rejected (retained) by the intact membrane
17.1 This test works on the principle that a dissolved dye in the system.
that is nearly completely rejected by an intact membrane
18. Apparatus
element will pass through a membrane or seal defect into the
permeate at an increased rate that indicates a leak that is 18.1 Feed Tank—For batch (recirculation) tests, a feed tank
capable of passing significant amounts of microbial material. of sufficient volume relative to the system size to allow
operation in recirculation mode, such as the system’s clean-in-
17.2 A solution of controlled concentration of a dye, known
place (CIP) tank connected to the feed and outlet piping
to be rejected at a rate of 99.0 % or greater (≥ 2 log) by the
system. Alternately, for flow-through tests, a system with a
membrane, is circulated through the system under standard
chemical feed pump calibrated to allow a controlled amount of
operating conditions as recommended by the manufacturer.
dye plumbed in prior to the high pressure pump can be used.
The concentration of the dye in the permeate and in the feed is
measured with a spectrophotometer for dyes that adsorb light 18.2 Spectrophotometer—The spectrophotometer must be
maximally at a specific wavelength or with a fluorometer for capable of measuring at a wavelength best for the absorption
fluorescing dyes that adsorb at one wavelength and emit at a spectra for the dye of interest.
second wavelength. A leak, or loss of integrity, will be 18.3 Fluorometer—The fluorometer shall be capable of
indicated by increased dye passage, as measured by a critical measuring Rhodamine WT with a minimum detection limit of
percent increase in the permeate concentration. The membrane
or system supplier may have a specific dye passage specifica- 9
Chapman and Linton found that a response greater than 0.53 µg/L was
tion that indicates loss of integrity—consult the supplier. For significant and could be differentiated from the baseline. Therefore, a feed
RO systems tested with FD&C Red Dye # 40, a passage greater concentration of 5 mg/L and a permeate concentration of 5 µg/L would correspond
than 0.2 % of the feed concentration is known to indicate a loss to a 3 log reduction (LRV) of dye.

11
 D6908 − 06 (2010)
10 nanograms per litre (ng/L) in clean water, using excitation solution, both the active concentration and specific gravity of
wavelength of 550 nm and emission wavelength of about 570 the Rhodamine WT must be accounted for. A100 µg/L concen-
to 700 nm. One fluorometer suitable for this purpose is the tration of active Rhodamine WT is equivalent to 0.32 mL of
Turner Designs model TD-700. 21.3 % active Rhodamine solution in 1 gal of water.
19.2.3 Recirculation Mode—Calculate the tank plus system
19. Reagents hold-up volume, and mix a solution of active dye to achieve a
19.1 Non-fluorescent Dyes: total system volume concentration of 100 µg/L.
19.1.1 Dye Feed Solution—For all RO systems and those 19.2.4 On-stream Mode—To achieve a feed dye concentra-
NF systems where the membrane is known to have a pore size tion of 100 µg/L, inject a 0.01 % (100 mg/L) solution of dye at
that retains molecules larger than 400 Daltons, FD&C Red Dye a rate of 3.2 gal per hour for every 100 gallons per minute of
#40 is suggested. If another dye is chosen, it must be miscible system feed flow. This injection rate will change if there is
in water, stable in the mid pH range, not adsorbed by the internal recycle of the concentrate stream back to the feed (the
membrane and nontoxic. Its molecular weight must also be test should be run with no internal recycle if possible). If run
appropriate for the membrane being tested. Check with the with recycle, the dye concentration in the concentrate stream
membrane supplier for suitable dye choices. should be calculated assuming 100 % rejection and used to
19.1.2 Recirculation Mode—Calculate the tank plus system recalculate the required dye concentration in the raw feed.
hold-up volume, and mix a solution of dye to achieve the 19.2.5 Calibration Curve—Prior to integrity testing, two
desired total system volume concentration (from 50 to 100 calibration curves of dye concentration to fluorescence must be
mg/L is recommended). developed: low (permeate) level curve (range of 10 ng/L 300
19.1.3 On-stream Mode—To achieve a feed dye concentra- ng/L) and high (feed) level (10 µg/L to 1 mg/L).
tion of 50 mg/L, inject a 1 % solution of dye at a rate of 3 20. Procedure
gallons per hour for every 100 gallons per minute of system
20.1 The system should be running under manufacturer
feed flow. This injection rate will need to be lowered if there is
recommended operating conditions or at conditions that best
internal recycle of the concentrate stream back to the feed (the
simulate the normal production mode of the membrane plant
test should be run with no internal recycle if possible). If run
for a period long enough that the performance of the membrane
with recycle, the dye concentration in the concentrate stream
system (as measured by flow and salt rejection) is at equilib-
should be calculated assuming 100 % dye rejection and used to
rium. For batch tests, add the appropriate volume of dye to the
recalculate the required dye concentration in the raw feed.
feed tank ensuring good mixing in the tank and recirculate the
19.1.4 Calibration Curve—Prior to integrity testing, a cali- feed solution to obtain a steady state dye concentration in the
bration curve of the test dye concentration to absorbance system. For on-stream tests, inject the dye using a metering
should be developed over the range of 1 µg/L to 1 mg/L. The pump to obtain the target concentration of dye in the membrane
proper wavelength must be determined for the chosen dye’s feedwater as indicated in Section 19. Inject the dye upstream of
spectrophotometric measurement. the membrane feed (high pressure) pump at a location that will
19.2 Fluorescent Dye: ensure adequate mixing of the dye with the feedwater.
19.2.1 Dye Feed Solution—For all RO systems and those 20.2 Determine the appropriate sample points, especially
NF systems where the membrane is known to have a pore size the sections of the system where permeate will be sampled. The
that retains molecules larger than 400 Daltons, Rhodamine WT amount of membrane contributing to each permeate sample
fluorescent dye may be used unless incompatible with the determines the sensitivity of the test, since any given leak is
membrane( check for compatability with the membrane manu- diluted by the permeate volume from the nonleaking mem-
facturer). Rhodamine WT has low adsorbability on most solid brane. It is recommended that the permeate from each housing
surfaces, is widely used in the water treatment industry as a (in a membrane train) be sampled to provide the maximum
tracer compound, and has been approved for use in drinking sensitivity.
water by the U.S. EPA provided that the concentration not
exceed 0.1 µg/L (100 ng/L) and exposure be brief and 20.3 Non-Fluorescent Dyes—Allow the system to equili-
infrequent. At this time, no other fluorescent dyes are approved brate for 15 min, or the time recommended by the
for use with drinking water. Based on studies conducted by the manufacturer, while maintaining constant flow, pressure, and
American Water Works Association Research Foundation temperature conditions. At the end of this period, collect
(AwwaRF), the minimum practical quantitation limit for Rho- 100-mL samples of the feed, concentrate, and permeate stream
damine WT in membrane permeates is 20 ng/L. Based on this in clean test tubes or cuvettes from the composite system and
level, a maximum permissible concentration of 100 ng/L in the from each individual housing for which integrity is to be
membrane feedwater, a desired LRV challenge level of 3.5 logs monitored. Measure and record the absorbance of the feed,
and an assumed LRV of 3 by NF and RO membranes, a feed concentrate, and permeate samples using a spectrophotometer
solution concentration of 100 µg/L is required. at the correct wavelength for the dye used (502 nm for FD&C
19.2.2 Dosing Solution Preparation—Commercially- Red Dye #40) and correlate absorbance to a dye concentration
available Rhodamine WT solutions have a specific gravity of value using the calibration curve. Calculate the percent dye
1.2 and are typically 21.3 % active, meaning 21.3 parts of passage using Eq 13.
active Rhodamine WT in 100 parts of water. To obtain a given 20.4 00-mL samples of the feed, concentrate, and permeate
concentration of active Rhodamine WT in a membrane feed stream in clean test tubes or cuvettes from the composite

12
 D6908 − 06 (2010)
system and from each individual housing for which integrity is Cf = dye concentration of the feed.
to be monitored. Measure and record the absorbance of the
Passage of 1 % and 0.1 % would correspond to LRV values
feed, concentrate, and permeate samples using a spectropho-
of 2 and 3, respectively.
tometer at the correct wavelength for the dye used (502 nm for
FD&S Red #40) and correlate absorbance to a dye concentra- 21.2 Because the concentrate stream from one stage is the
tion value using the calibration curve. Calculate the percent feed to an additional stage in series, the dye concentration of
dye passage using the equation in Section 21. the feed to the downstream stage will be higher. To calculate
the dye concentration of the feed to the downstream stage,
20.5 Fluorescent Dye—Allow the system to operate for 15
assume the feed concentration to each stage in a given array is
min, or the time recommended by the manufacturer, while
equivalent. Recalculate the dye concentration for the feed to
maintaining constant flow, pressure, and temperature condi-
each separate stage by reducing the feed volume by the
tions. At the end of this period, collect 100-mL samples of the
approximate volume of the permeate removed in the preceding
feed, concentrate, and permeate stream in clean test tubes or
stage while holding the mass of dye in that feed solution
cuvettes from the composite system and from each individual
constant. This higher concentration of dye will enter the
housing for which integrity is to be monitored. Measure and
downstream stage. Alternately, one can assume the feed to each
record the fluorescence of the feed, concentrate, and permeate
stage in series is equivalent to the feed to the entire system.
samples using a fluorometer and excitation and emission
This will increase the safety factor of the test but may give a
wavelengths as described in Section 19 and calculate dye
false indication of a leak.
concentration using the appropriate calibration curve. Calcu-
late the percent dye passage using the equation in Section 21. 21.3 If a leak is detected measure the composite permeate
from each stage to determine the stage with the breach of
21. Calculations integrity. Then measure permeate flow from individual hous-
21.1 To calculate dye passage, use the equation: ings within the suspect stage to isolate the leaking element(s).10
Cp
Dye Passage ~ % ! 5 ·100 (13)
Cf

where:
10
The elements in the housing may be individually tested with a similar
Cp = dye concentration of the permeate, and procedure from Practice A to determine which have lost integrity.

PRACTICE D—USE OF CONTINUOUS MONITORING PARTICULATE LIGHT SCATTERING METHODS TO

MONITOR MEMBRANE INTEGRITY

22. Scope these two technologies is possible and can increase sensitivity
22.1 This guide is applicable to MF, UF, NF, and RO and facilitate defect identification.
membrane systems, including those with spiral, tubular or flat 22.4 This method provides for the continuous monitoring of
sheet configuration elements. The feedwater must have a each sample point. Flow through the sample point (that is, the
particulate concentration that is at least an order of magnitude sensor) is continuous, and the measurement frequency is at
in higher concentration than that which is detected in the least every 15-min during filtration.
filtrate or permeate stream. This practice is applicable for those
23. Summary of Practice
membrane systems and that will allow for the monitoring of the
water source and in the filtrate or permeate stream as it exists 23.1 This practice works on the principle that an integral
a membrane module. membrane will not allow the passage of particles greater than
1µm into the filtrate or permeate stream. The detection of a
22.2 For turbidity techniques, the feedwater should have a
sudden increase (or spike) of particles greater than 1-µm is
turbidity of at least 0.5 NTU (500 mNTU) and for particle
indicative of some level of integrity loss (that is, a breach). In
counting techniques, the feedwater should have a count con-
the event of an integrity loss, the measurement baseline will
centration of at least 1 particle per mL greater than the defined
increase in response and in its amplitude of fluctuation.
size threshold that is being applied.
23.2 This practice provides a continuous stream of informa-
22.3 This practice provides for the use of two different laser tion that is related to the quality of the filtrate stream as it exits
based particulate detection technologies: optical particle count- the membrane module. The decrease in the quality of the
ing (light obscuration designs) and laser turbidity (light filtrate stream is a reflection of the module’s integrity, and is
scattering).11,12 Liquid or sensor multiplexing with either of signified by an increase in its particulate content. The practice
can be used to detect a change relative to an integrity loss. A
11
Carr, M., et al, “Membrane Integrity Monitoring with Distributed Laser
quantitative change requires the calibration of the specific
Turbidity,” Journal American Water Works Association , 95: 6, 2003, p. 94. monitoring parameter’s response against a defined turbidity or
12
United States Environmental Protection Agency, “Draft of the National particle count standard.
Primary Drinking Water Regulations: Long Term 2 Interim Enhanced Surface Water
Treatment Rule” 40 CFR Parts 141 and 142, Vol 68, No. 154, 2003, pp. 47661 and 23.3 The sensitivity of each of the technologies can vary
47690. depending on the feedwater conditions, pretreatment, the type

13
 D6908 − 06 (2010)
of membrane material, module design, and rack design. Sen- 24. Significance and Use
sitivity can be determined using fiber cutting studies on a pilot 24.1 The integrity test methods described are used to
scale plant that utilizes a representative, full-scale module. monitor the performance of a membrane based on its ability to
During this study, a correlation between instrument param- consistently remove particles that are at least 1 µm in diameter.
eter’s response (that is, laser turbidity value, RSD value, or
particle counts greater than a specific threshold diameter) to the 24.2 The test methods can be applied to either positive
number of cut fibers can be drawn. See 23.3.1 for an example pressure filtration or negative pressure filtration designs.
on how to determine the sensitivity for a specific parameter. However, the different filtration designs do require different
sampling configurations to eliminate interferences.
23.3.1 Insure the instrumentation is installed at the appro-
priate sample point that addresses the issues associated with 24.3 These technologies can be applied to varying configu-
providing a representative sample to the instrument sensor with rations and sizes of membrane systems. Particle counting and
all interferences being minimized. laser turbidimeter technologies are best applied to small
23.3.2 Allow for the conditioning of the instrument sensors, systems that typically contain no more than three modules. For
which is exhibited by stable measurement baselines. membrane systems with rack designs that contain a larger
number of modules, the application of sensor or liquid multi-
23.3.3 The baseline data generated from stable measure-
plexing can enhance detection sensitivity to a breach.
ment baselines can be used to establish upper control limits
24.3.1 The highest sensitivity will occur when each module
(UCL) as alarm levels that signal a potential integrity breach.
is monitored individually by a dedicated sensor. The use of a
See equations in 29.2. If the technology is to be used to detect
1:1 sensor-to-module monitoring ratio will provide the highest
a relative change in integrity, the prescribed approach is to
sensitivity and direct traceability to the suspect module.
establish a UCL that is based on either a 95 percent confidence
level, a 99.7 percent confidence level, or a 99.997 percent 24.4 The ability of any of these technologies to detect any
confidence level.11,13 Depending on the significance in which given defect is affected by the size of the membrane module
interferences are reduced, these UCL values can be used to (total surface area), and the size of the membrane rack.
establish alarm limits for investigating a potential membrane 24.5 As the surface area of a module increases the sensitiv-
breach. See Guide D3864 for equations relating to deriving the ity may decrease. In this case, those techniques that display the
standard deviation and on the establishment of UCL values. highest sensitivity should be used.
23.3.3.1 If the monitoring approach is to be used to quantify 24.6 If the continuous monitoring instrumentation is to be
the change in filtration performance, such as . in the confirming used as a quantitative tool, the instrument should be calibrated
of an LRV, the UCL should be established according to those using the appropriate calibration materials for the industry.
integrity results that were obtained during fiber cutting studies Consult the manufacturer and/or regulatory authority for cali-
on the same type of module under the same expected feedwater bration materials.
conditions. Fig. 6 provides an example on how a fiber cutting 24.6.1 If the continuous monitoring instrumentation is to be
study can be used to apply a UCL for a specific set of used to detect a non-quantitative change, calibration to a
monitoring parameters to alarm the operator of a possible traceable standard is not required. However, a comparative
membrane breach. calibration should be performed against the filtrate water
23.3.4 Dilution and flow affect sensitivity of these methods. baseline that is leaving a membrane module(s) that have
Flow changes from both the pilot and the full-scale membrane confirmed integrity.
unit must be known and used to calculate the difference in 24.6.1.1 Baseline levels and baseline fluctuation may vary
sensitivity of the between the method that was developed on a with different feedwater types.
pilot system and the full-scale system.
24.7 Particle counting can be used to determine LRV. This
23.4 Instruments that have capabilities of averaging con- requires the simultaneous monitoring of the feedwater and the
secutive measurements to reduce the chance of false occur- filtrate sample across the membrane module or rack. Both
rences are available and can help to increase reliability in instruments should be calibrated using the same materials and
measurement. have the same bin settings.
24.7.1 Particle counters with 1 µm sizing capability will
23.5 The UCL values should be updated at designated
have better sensitivity to integrity losses than sensors that have
intervals to accommodate any long-term drift in the instrumen-
higher size thresholds.
tation and for changes in the feedwater. Consult the instrument
manufacturers for guidance for the establishment and updating
of the measurement UCL value. The UCL calculation and 25. Interferences
alarm settings can be defined using the analytical instrumen- 25.1 This detection method is based upon optical light
tation and/or through the use of many available data manage- scattering techniques. Bubbles that result from outgassing in
ment software platforms that are available in the industry. sample lines that lead to the sensor can result in false positive
spikes and excessive baseline noise. The sample must be
adequately degassed using conditioning techniques such as
13
bubble rejection devices (that is, bubble traps) and backpres-
Hargesheimer E. E., and Lewis, C. M., “A Practical Guide to On-Line Particle
Counting,” American Water Works Association Research Foundation, ISBN sure on the flow cell compartment within a sensor. See Guide
0-89867-785-8, Denver, CO, 1995. D3864.

14
 D6908 − 06 (2010)
25.1.1 Consult with the instrument manufacturer and with loss in membrane integrity. These components are shared
the membrane manufacturer for appropriate sampling points across a large number of sensors that are dedicated to unique
and any specialized techniques. See Test Method D6698 for sample points. The entire sensor multiplexed monitoring sys-
additional sampling information that relates to bubble removal. tem consists of three parts: The multiplexor or centralized
25.2 A significant amount of bubbles will result when the instrument, the fiber optic transmission cables, and the sensors.
rack is in a cleaning process. These events, which include The multiplexor typically contains such components as the
reverse flow, reverse flow air scour, and clean-in-place proce- optics, electronics, and user interface, that are “shared” across
dures can introduce significant residual bubble interference all sensors. The centralized instrument contains the location
into sample lines for some time after the procedures have been and origin of incident light signals and the termination site for
completed. all scattered light signals that would be transmitted back from
25.2.1 The integration of a signal from the membrane rack the sensors. The sensors are typically continuous sampling
that indicates the stoppage of flow can be fed to the data units through which sample passes. They contain the inlet and
logging program for the continuous monitoring instrumenta- outlet ports for sample and connections to receive incident light
tion to stop making measurements and logging data when the from the centralized instrument. The sensors also provide the
rack is not in forward filtration flow. A time lag should also be components that will allow the transmission of scattered light
included to allow for residual air to vacate sample lines after back to the centralized instrument. The fiber optic transmission
forward filtration flow filtration has resumed. cables are the light transmission conduits between the sensors
and the multiplexor. In this example, a dedicated pair of fiber
26. Apparatus optic cables link each sensor to the multiplexor. Analysis or
26.1 Particle Counter—Practice F658 contains the criteria measurement of sample that flows through the linked sensors
appropriate designs that can be applied in this practice. Optical can be performed sequentially or in parallel, depending upon
light obscuration particle counters of in-situ or volumetric the design of the system. See Fig. 7 for a block diagram on how
design that are capable of sizing particles in the 1 to 100 µm a sensor-based multiplexed optical system can be integrated
size range should be used. The particle counter should be able into a membrane rack.
to determine the total particles per mL that are greater than a 26.2.2 Liquid Multiplexed Applications—An approach to a
set size threshold. The particle counter should be able to liquid multiplexing system involves the transport of separate
receive sample continuously at a set flow rate in the range of 50 sample streams from their sources to an inlet location on a
and 250 mL/min.13 manifold valve apparatus. The valve apparatus outlet is
26.1.1 Practices E20 and F658 provide guidance on particle connected, via tubing into the particle detection instrument (for
counter designs and calibration information. example, laser turbidimeter or particle counter). The manifold
valve apparatus sequentially allows the passage of each sample
26.2 Laser Turbidity—The laser turbidimeters should con- stream into and through the particle detection instrument. The
sist of a laser or laser diode light source as the incident light control of the manifold valve can be performed automatically
beam. The detector is typically centered at 90 degrees but other or manually. Time must be allotted accordingly to provide for
detection angles may be desired for increased sensitivity. The adequate purging of the old sample from the manifold valve,
instrument should be designed so that little stray light reaches the outlet tubing, and the particle detection instrument so that
the detector in the absence of particles in the sample. representative sampling is achieved. When using a of fluid
Specifically, the light source shall be at 660 6 30 nm. There multiplexing approach, the length and volume of each sample
shall be no divergence from parallelism at the incident radia- from its source to the manifold valve, and the flow rate must be
tion and any convergence shall not exceed 1.5 degrees. The known. This will help to insure representative sampling and a
distance traversed by the incident light and scattered light shall quality data stream is achieved.
not exceed 10-cm. The detector should be a photomultiplier
tube design with a spectral response that overlaps the incident 27. Reagents
light beam. The wavelength of the incident light beam should 27.1 Turbidity Standards—Test Method D6698 provides
provide overlap to at least 75 % of the total response peak for appropriate calibration standards for the calibration of turbidi-
the detector. Transmission of the incident light and or the meters.
scattered light beams may be performed through the use of
fiber optic couplings.14 27.2 Particle Counters—Practice F658 provides guidance
26.2.1 Sensor-Based Multiplexed Applications—The sensor for calibration materials. Use polystyrene latex calibration
multiplexing design, in which a common set of analytical spheres that have a defined size. See the manufacturer’s
components are used across multiple sampling points, can be instructions for the types and sizes of materials to be used.
applied to either laser turbidimeter or particle counting tech- 27.3 Dilution Water—This shall be prepared by filtration of
nologies. An example of a sensor-based multiplexing system is Type III water or better through a 0.22-µm or smaller filter
presented here. The approach uses a single laser light source within 1 hour of use. Reverse Osmosis water is acceptable and
and a high sensitivity detector that are capable of detecting a preferred.
28. Procedure
14
Sadar, M. J., “Introduction to Laser Nephelometry, An Alternative to Conven-
tional Particulate Analysis Methods, Appendix A,” Hach Company Technical 28.1 System Installation—Refer to manufacturer instruc-
Information Bulletin 7044, Loveland, CO, 2003. tions for the exact installation and integration of the monitoring

15
 D6908 − 06 (2010)

FIG. 7 Block Diagram of the Integration of a Sensor-Based Multiplexed Monitoring System into a Membrane Rack

system into the membrane module or rack. Refer to Test sensors and sample lines. Once readings are become stable, the
Method D6698 for guidance on installation and sampling of monitoring system is considered conditioned.
low turbidity water. Refer to the membrane manufacturer for
the best sampling point(s) from the membrane rack. If LRV is 28.3 Perform a measurement of each sensor (sample point)
to be determined install the appropriate sensor on the feedwa- at a frequency that is no longer than 15 minutes during forward
ter. filtration flow.
28.1.1 The determination of LRV is traceable only to the 28.4 Each time a measurement is performed, compare it
specific technology used. LRV values from one technology against the respective UCL. An alarm should trigger for the
(particle counting or turbidity) cannot be used interchangeably specific sample point when the measurement value exceeds its
with, for example, a pressure decay rate. established UCL value.
28.2 Provide flow to all monitoring sensors. If necessary
adjust pressure and flow. Allow sample to flow for at least 24 28.5 If monitoring for LRV, determine the log difference
hours to allow for conditioning and stabilization of sample between the feedwater particle counts and the filtrate particle
lines and internal surfaces with the sensors before establishing counts for the specific membrane module or rack that is being
upper control limits UCL. See Section 23. monitored over the same time interval.
28.2.1 The 24-h conditioning period provides time to prop- 28.6 Continue to monitor the process for excursions that
erly wet all surfaces that the sample will flow across before and would trigger a UCL alarm. In the event an alarm has been
during analysis. This conditioning period will typically exhibit exceeded, further investigation as to its cause is warranted to
erratic readings as bubbles are removed from these surfaces
determine if an integrity loss has occurred.
during the wetting process. This time may vary depending on
the type of materials that are used in the construction of the 28.7 See Appendix X2 for additional guidance on sampling.

16
 D6908 − 06 (2010)
29. Calculations the black trace represents the variability of the measurement,
29.1 To calculate LRV using particle counting: which is quantified and identified as the % RSD. The horizon-
tal gray-dotted and black-dotted lines represent the established
LRV 5 log S D
Ctsƒ
Ctsp
(14) upper control limits (UCL) for the turbidity and % RSD
parameters respectively. These UCL values were established
where: according to a 99.5 %. confidence limit as is provided by the
Ctsƒ = feedwater particle counts, and equation in 29.2. The fiber cutting tests involved cutting all
Ctsp = filtrate or permeate particle counts. fibers initially and then performing repairs in sequential steps.
Each fiber repair step was separated by the horizontal black-
29.1.1 This same technique will be used to calculate the
dotted lines. A description of the state of the membrane module
LRV for the other measurement technologies.
with respect to the number of cut fibers is provided in the text
29.2 Calculation of UCL values: boxes.
UCL 5 M p 12σ p for a 95 percent confidence upper control limit 30.3.1 The graph illustrates the response of both parameters
(15) relative to significance of broken fibers. The detection of an
integrity breach was confirmed if the response exceeded the
UCL 5 M p 13σ p for a 99.5 percent confidence upper control limit UCL for the respective parameter. Both parameters detect the
UCL 5 M p 14σ p for a 99.997 percent confidence upper control limit
integrity change for all tests involving one or more broken
fibers. It was only when pin-hole integrity losses were tested
where: that the turbidity response was questionable, but the baseline
Mp = the averaged value for a set of at least 7 consecutive variability parameter does detect the changes, which was
measurements on the filtrate or permeate water for a evidence by exceeding the UCL alarm threshold.
specific monitoring parameter, and 30.3.2 The feedwater supplied to this pilot system was a
σp = the standard deviation for the same set of measure- blend from a water treatment plant sedimentation water and
ments used to calculate Mp. raw water. The turbidity was maintained between 1 and 3 NTU
29.3 Some these newer instruments have the capability to throughout the duration of this testing. Filtrate was sampled
perform these types of calculations. In the absence of such immediately as it exited the module. Internal baffling within
calculations, SCADA and PLC programs can perform these the module provided a homogeneous sample. Flow through the
functions. sensors was 50 mL/min and backpressure was 5 psi. The feed
water flow was 30 gpm during this experiment. The membrane
30. Interpretation of Results system was an ultra-filtration, with a positive pressure inside-
30.1 A sudden decrease in the calculated LRV value that out filtration configuration. Flux was 30 to 35 g/ft2/day, and the
exceeds a pre-established control limit is indicative of an module contained two membrane elements, each containing
integrity breach. approximately 15 000 fibers. The configuration of the module
30.2 A measurement that exceeds any pre-established UCL and elements were representative of that which is used in
indicates the possibility of an integrity breach and warrants full-scale racks. However, under conditions of a single or
further investigation. Excessive noise or variance in the mea- pinholes, a portion of the data exceeds the established alarm
surement baselines also indicates a possible integrity breach. condition, and a portion of the data is below the condition. This
For example, if an established baseline that was shown to have would warrant further investigation as to the cause and, in this
no values exceeding the UCL suddenly changes to a baseline in case, the decreases could be due to partial plugging of the
which 25 % of all measurements are exceeding the UCL, this breaches, which re-open when a reverse flow operation takes
is indicative of a potential integrity breach. place. In this example, the data is in alarm at least 25 % of the
30.2.1 At the point where measurements are exceeding the time for any of the breach conditions. It would be prudent to
established upper control limit, the operator should investigate take the membrane system off line and to further investigate the
further to determine if a breach does exist. The running of a integrity breach such as through running a direct integrity test.
pressure-based or vacuum integrity test can be used to confirm
or dismiss the alarm. 31. Keywords
30.3 In Fig. 6, two monitoring parameters were used to 31.1 continuous monitoring; integrity; membrane; multi-
monitor membrane filtrate through series of fiber cutting tests. plexing; pressure decay; soluble dye test; TOC test; vacuum
The gray trace represents the laser turbidity of the filtrate and decay

17
 D6908 − 06 (2010)

FIG. X1.1 Membrane System Under Test and Filtration Conditions

APPENDIXES

(Nonmandatory Information)

X1. DERIVATION

X1.1 The mass flow of particles in the filtrate leaving the be the same as the concentration of particles in suspension and
system is made up of particles passing through the membrane challenging the membrane, which we will call Cfeed. In direct
(Cmembrane · Qmembrane) and particles bypassing the membrane flow (dead-end) filtration systems Cfeed can be assumed to be
through defects or leaks (Cbypass · Qbypass). The mass flow of the same as the concentration of particles in the raw feed to the
particles challenging the membrane is Cfeed · Qfilt. The log membrane system, Craw. However, in the case of recirculation
reduction of particles across the membrane is defined by mass systems or configurations that increase the concentration of
balance as shown in Eq X1.1. suspended solids on the feed side of the membrane (such as
feed and bleed modes) Cfeed can be many times greater than
LRV 5 log10 S~ C raw·Q filt
C bypass·Q bypass1C membrane·Q membrane! D (X1.1) Craw. In such cases Cfeed is calculated as follows:
C feed 5 CF·C raw
where:
Where CF is a concentration factor that typically ranges
Craw = concentration of particles entering the system, from 1 to more than 20. Substituting into Eq X1.1 gives:
Cfilt = concentration of particles in the filtrate leaving

Cbypass
the system,
= concentration of particles in the flow bypassing
LRV 5 log10 S Q filt
Q bypass·CF D (X1.2)

the membrane through defects or leaks, X1.3 As the integrity test is conducted under known condi-
Cmembrane = concentration of particles passing through the
tions of temperature and pressure, it is possible to mathemati-
membrane,
cally estimate the equivalent flow of liquid that would pass
Qfeed = flowrate of feed to the membrane (= Qfilt),
Qfilt = flowrate of filtrate leaving the membrane (= through these same defects under filtration conditions (the
Qbypass + Qmembrane), bypass flow Qbypass). Using Eq X1.2, an estimate of the LRV
Qbypass = flowrate bypassing the membrane through de- for the system can then be determined.
fects or leaks, X1.4 The following derivations are made with reference to
Qmembrane = flowrate passing through the integral portion of Fig. X1.1, assuming laminar flow for both the air and liquid
the membrane, and through cylindrical defects. The PDR or VDR is first expressed
LRV = log reduction value of particles across the mem- as QG,atm, the volumetric air flow rate through defects at
brane system. atmospheric pressure. If we define Vsystem as the volume of the
air cavity pressurized during the PDT, or under vacuum during
X1.2 For particles with a diameter greater than the mini-
the VDT, then at the beginning of the test:
mum defect size (as set by the test pressure chosen, see 9.3) an
intact membrane will give complete rejection. Thus for these P i V system 5 P atm V i
particles Cmembrane will be zero, and Cbypass can be assumed to At the end of the air test:

18
 D6908 − 06 (2010)
P f V system 5 P atm V f assume that the defect geometry remains the same in both test
Where Vi and Vf represent the equivalent volumes of air at and filtration conditions we can introduce a proportionality
atmospheric pressure (Patm) at the beginning and end of the test constant, k, to represent the geometry term πd4 / (128L) in Eq
respectively. Pi and Pf are the initial and final pressures during X1.7.
the test. Applying this to the test conditions for air:
By subtraction:
~ P u,test 2 P d,test!
Q G,test 5 k· (X1.8)
~ P i 2 P f ! V system 5 ~ V i 2 V f ! P atm µ air
V system And for filtration conditions:
V i 2 V f 5 ~ P i 2 P f!
P atm
TMP
Dividing both sides by t, the duration of the test: Q bypass 5 k· (X1.9)
µ water
V i 2 V f ~ P i 2 P f ! V system Divide Eq X1.9 by Eq X1.8 and rearranging:
5 (X1.3)
t t P atm
By definition for PDT: Q bypass 5 Q G,test S Dµ air
µ water
TMP
P u,test 2 P d,test
(X1.10)
Vi 2 Vf Pi 2 Pf Substitute Eq X1.6 into Eq X1.10:
5 Q G,atm and 5 PDR
t t
2P atm TMP µ air
Substituting into Eq X1.3 gives the following for PDT: Q bypass 5 Q G,atm (X1.11)
~ P u,test1P d,test! ~ P u,test 2 P d,test! µ water
V system
Q G,atm 5 PDR (X1.4) µ air 2P atm TMP
P atm
Q bypass 5 Q G,atm
and for VDT by definition: µ water ~ P u,test2 2 P d,test2 !
Define:
Vf 2 Vi Pf 2 Pi
5 Q G,atm and 5 VDR µ water P u,test2 2 P d,test2
t t
ƒ1 5 and ƒ 2 5
Substituting into Eq X1.3 gives the following for VDT: µ air 2P atm TMP
Substituting into Eq X1.11 gives:
V system
Q G,atm 5 VDR (X1.5) Q G,atm
P atm
Q bypass 5 (X1.12)
QG,atm is converted to QG,test, by correcting for the difference ƒ 1, ƒ 2
in pressures using the average pressure through the defect: Where ƒ1 and ƒ2 can be considered to represent viscosity and
pressure corrections respectively.
Q G,test P atm
5 (X1.6) Substituting Eq X1.12 into Eq X1.2 gives:
Q G,atm
S P u,test1P d,test
2 D LRVe 5 log10 S Q filt
ƒ ƒ D (X1.13)
CF·Q G,atm 1 2
2P atm Substituting Eq X1.4 and Eq X1.5 into Eq X1.13 gives:
Q G,test 5 Q G,atm
~ P u,test1P d,test! For PDT:
X1.5 The next step is to convert the air flow through the
defect under test conditions, QG,test, to an equivalent liquid
LRVe log10 S Q filt P atm
ƒ ƒ
CF·PDT·V system 1 2 D (X1.14)

flow under filtration conditions, Qbypass. To do this it is For VDT:

assumed that both the air and liquid flows follow the Hagen-
Poiseuille law for laminar flow in circular pipes, which is: LRVe log10 S Q filt P atm
ƒ ƒ
CF·VDT·V system 1 2 D (X1.15)

πd 4 ∆p X1.6 The assumption of laminar flow for both air and liquid
Q5 (X1.7)
128Lµ is not valid for all configurations, particularly large diameter
Where Q is the flowrate, ∆P is the pressure drop, and µ is the membranes (> 300 to 400 µm lumen diameter) which can lead
fluid viscosity. The parameters d and L refer to the diameter to error in the calculated LRV. As with all methods to correlate
and length of the pipe. In our application we have no integrity test results to LRV the relationship should be verified
information about the nature or number of defects that would by field tests for the particular membrane and configuration
allow the appropriate choice of values for d and L. If we used.

19
 D6908 − 06 (2010)

X2. GUIDANCE TO OBTAINING RELIABLE AND QUALITY MEASUREMENT DATA FROM CONTINUOUS
MONITORING SYSTEMS

X2.1 Throughout Practice D, there contains several pieces X2.1.4.1 For UF, NF and RO systems with no air backwash
of key sampling information related to instrument setup, require the least amount of sampling conditioning because
sampling and data handling that aid in achieving optimum bubble generation during the cleaning processes in minimal.
reliability and sensitivity for these technologies. A summary of Typically, reverse-flow cleanings do not cause spikes in the
this information is provided below: measured values. Standard sampling protocols that are pro-
X2.1.1 Common turbimeters such as those that comply with vided by the instrument manufacturers are adequate (see 23.3).
USEPA method 180.1 are not suggested for use due to having X2.1.4.2 MF and UF systems that are outside infiltration
lower sensitivity. Laser turbidimeters with a greater sensitivity processes often use air scour at the within filtration cycle. For
are suggested (see 22.3). these systems, bubble generation and subsequent bubble inter-
X2.1.2 Insure instrumentation is installed at appropriate ference can occur. For these systems bubble interferences are
sample points that insure a homogeneous and representative minimized through the use of: (1) sample chambers that can be
sample. Sample points that insure filtrate or permeate water pressurized to prevent further outgassing (see 25.1); (2) addi-
changes direction (such as following a 90-degree elbow) are tional bubble traps such as those that contain a larger volume
suggested (see 23.3.1). may be added to dampen bubble interference (see 25.1); (3) the
X2.1.3 Instrument bodies (sensor bodies) must be condi- integration of a signal from the rack that indicates the stoppage
tioned prior to the establishment of any baseline that is used for of forward filtration flow can be fed to either the plant
an upper control limit (UCL) or alarm condition. A stable (SCADA) system or data logging program. Under this condi-
continuous monitoring baseline is represented by a level tion data is either ignored or not logged (see 25.2); and (4) a
measurement values that are void of random and unpredicted time lag should be incorporated into the data logging program
spikes (see 25.1). to ignore data for the first 2 to 6 minutes (user changeable) after
X2.1.4 Continuous monitoring methods can be applied to the completion of an air-scour cleaning procedure. This is to all
either positive or negative filtration, but both types require for air to vacate sample lines prior to resuming the logging of
different sampling to minimize bubble interference. The fol- measurement data (see 25.2). Many instruments contain mea-
lowing information can be used as guidance for specific types surement algorithms that aid in the reduction or rejection of
of membranes: bubble interference (see 24.1).

ASTM International takes no position respecting the validity of any patent rights asserted in connection with any item mentioned
in this standard. Users of this standard are expressly advised that determination of the validity of any such patent rights, and the risk
of infringement of such rights, are entirely their own responsibility.

This standard is subject to revision at any time by the responsible technical committee and must be reviewed every five years and
if not revised, either reapproved or withdrawn. Your comments are invited either for revision of this standard or for additional standards
and should be addressed to ASTM International Headquarters. Your comments will receive careful consideration at a meeting of the
responsible technical committee, which you may attend. If you feel that your comments have not received a fair hearing you should
make your views known to the ASTM Committee on Standards, at the address shown below.

This standard is copyrighted by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959,
United States. Individual reprints (single or multiple copies) of this standard may be obtained by contacting ASTM at the above
address or at 610-832-9585 (phone), 610-832-9555 (fax), or [email protected] (e-mail); or through the ASTM website
(www.astm.org). Permission rights to photocopy the standard may also be secured from the ASTM website (www.astm.org/
COPYRIGHT/).

20