Secondary statistical cutpoint methodology for gas-liquid diffusion integrity testing

Abstract

The present invention relates to a method of reducing background noise during integrity testing of membrane filters to improve sensitivity and more effectively dispose of non-conforming membrane filters.

Description

[0001] Related applications

[0002] This application claims the benefit of U.S. Provisional Patent Application No. 63 / 131,850, filed on December 30, 2020, the entire contents of which are incorporated herein by reference. Technical Field

[0003] The embodiments described herein relate to methods for gas-liquid integrity testing of membrane filters. More specifically, some embodiments of this technology relate to methodologies for analyzing gas-liquid diffusion integrity test data, which uses normalization of a subset of the population, or adjustment of the integrity test data for operating conditions such as temperature and pressure, for the purpose of reducing background noise, thereby improving the signal-to-noise ratio for defect detection. Background Technology

[0004] High-purity filtration of aqueous media used in the biotechnology, chemical, electronics, pharmaceutical, and food and beverage industries is achieved through the use of sophisticated membrane filter modules capable of highly precise separation. These membrane filters also prevent contamination of the environment, the filtered medium, and the resulting filtrate, preventing unwanted and often harmful organisms such as bacteria or viruses, as well as environmental contaminants such as dust and dirt, from entering the process fluids and final products.

[0005] To ensure the sterility and / or retention capacity of membrane filters are not compromised, integrity testing is a requirement in critical process filtration applications. Manufacturers of membrane filters for critical applications also frequently apply integrity testing to finished filter elements as a lot release or 100% testing standard. Integrity testing detects the presence of excessively large pores or defects that can impair the retention capacity of porous materials. Methods for integrity testing include particle challenge testing, liquid-liquid porometry testing, bubble point testing, gas-liquid diffusion testing, and diffusion testing of tracer components. Some of these tests, such as particle challenge testing, are destructive. Therefore, these tests cannot be used as pre-use tests. Liquid-liquid porometry and bubble point testing are suitable for ensuring the installation of membranes with the correct nominal pore size, but these methods lack sensitivity for identifying small, numerous defects.

[0006] Gas-liquid diffusion testing is widely used to evaluate the overall performance of filters. A common gas-liquid pairing for overall testing is air-water due to its safety, low cost, and environmental friendliness. Diffusion testing measures the rate at which gas transfers through the filter. Under a gas pressure differential below the bubble point, gas molecules migrate through the water-filled pores of a wetted membrane according to Fick's law of diffusion:

[0007]

[0008] Where Q is the permeation velocity, A is the membrane area, ε is the membrane porosity, D is the gas diffusivity in the liquid, S is the gas solubility coefficient, and P... f and P p Here, τ represents the feed and permeation side pressure, τ represents the tortuosity factor, and L represents the thickness of the liquid in the membrane.

[0009] For monolithic membranes, the following measured gas velocity is a signal of a defect: it exceeds the velocity predicted by Fick's Law or is higher than the empirically established velocity. The sensitivity of this test is limited by the minimum detectable excess flow. Significant variations in gas diffusion velocity can occur between filters in monolithic membrane filter units due to differences in membrane area, membrane thickness, membrane porosity, and pore tortuosity factor.

[0010] Variations in test system hardware and instrumentation, as well as operating conditions affecting the diffusivity and solubility of gases in liquids, such as changes in pressure and temperature, also contribute to background noise. Background noise can compete with or interfere with defect signals, which are caused by excessive gas flow velocity resulting from convection through the defect. Faced with increasingly significant noise variations, traditional air diffusion integrity testing will have to expand its acceptable specification window (i.e., reduce defect detection capability) or accept more false nonconformities.

[0011] The primary contributing factors to this noise variation are membrane characteristics that vary by batch (referred to herein as the "castdate") and by the test location on the equipment performing the overall test cycle (referred to herein as the "bowl"). The sensitivity of gas-liquid diffusion tests used to detect defects is also directly limited by background noise. High background noise can also cause monolithic filters to erroneously fail tests, leading to increased cost waste. When establishing test acceptance criteria, including upper and lower limits for diffusion specifications, a trade-off must be struck between test sensitivity (product and end-user risk mitigation) and the ability to robustly supply products (volume, cost, and safety to the market).

[0012] Therefore, there is a need for a holistic testing methodology that maintains the convenience of gas-liquid holistic testing while minimizing the impact of inherent background noise variables, which can both impair the ability to detect defects and increase the likelihood of false nonconformities in holistic filters. Summary of the Invention

[0013] The embodiments described herein overcome the shortcomings of the prior art, including some embodiments of methods for reducing background noise and improving the signal-to-noise ratio in holistic testing to dispose of membrane filters, the methods comprising: performing gas-liquid diffusion analysis on the membrane filter; identifying, during the gas-liquid diffusion analysis, whether the membrane filter is an outlier compared to an initial fixed cutoff point; if at least one membrane filter is not an outlier with respect to the initial fixed cutoff point, normalizing the values ​​of at least one characteristic of the at least one membrane filter, the values ​​affecting the background noise in the gas-liquid diffusion analysis; identifying, based on the values ​​of the gas-liquid diffusion analysis from multiple membrane filters, whether the membrane filter is an outlier compared to a secondary statistical cutoff point; and disposing of the membrane filter as holistic or non-compliant.

[0014] In some embodiments, the gas-liquid diffusion method is air-water. In some embodiments, the at least one characteristic is selected from batch, equipment test location, water temperature, air temperature, and pressure. In some embodiments, the values ​​of at least two characteristics are normalized. Some embodiments of the method further include normalizing the pressure of more than one membrane filter. Some embodiments of the method further include detecting at least one defect in the more than one membrane filter. In some embodiments, the at least one defect is selected from about 4 micrometers to about 9 micrometers. In some embodiments, the defect is greater than 4 micrometers. In some embodiments, the defect is selected from about 4 micrometers, about 5 micrometers, about 6 micrometers, about 7 micrometers, about 8 micrometers, and about 9 micrometers. In some embodiments, the initial upper limit cutoff point maintains a safety margin of at least 5% from the diffusion specification standard of the end user. In some embodiments, the initial upper limit cutoff point maintains a safety margin of at least 5% from the diffusion standard issued by quality control (QC).

[0015] In some embodiments, the initial upper limit cutoff point maintains a safety factor selected from approximately 5% to approximately 15% of the end-user diffusion specification standard or the quality control (QC) batch release diffusion standard. In some embodiments, the initial upper limit cutoff point maintains a safety factor selected from approximately 5%, approximately 6%, approximately 7%, approximately 8%, approximately 9%, approximately 10%, approximately 11%, approximately 12%, approximately 13%, approximately 14%, and approximately 15%. In some embodiments, the membrane filter is an outlier for the initial cutoff point and is treated as non-compliant. In some embodiments, the membrane filter is an outlier for the secondary cutoff point and is treated as non-compliant. In some embodiments, the membrane filter is not an outlier for either the initial or secondary cutoff point and is treated as an integral component.

[0016] In some embodiments, the plurality of membrane filters is a statistically significant number of membrane filters. In some embodiments, the membrane filters and the plurality of membrane filters are the same type of filter device. In some embodiments, the filter device is selected from sterile filters, virus filters, clarifiers, and ultrafiltration filters. In some embodiments, the secondary statistical cutoff is established by setting the upper statistical limit to between 2 and 5 standard deviations above the median of the normalized values. In some embodiments, the secondary statistical cutoff is established by setting the upper statistical limit to approximately 2, approximately 3, approximately 4, and approximately 5 standard deviations above the median of the normalized values. In some embodiments, the secondary statistical cutoff is established by setting the lower statistical limit to between 4 and 6 standard deviations below the median of the normalized values. In some embodiments, the secondary statistical cutoff is established by setting the upper statistical limit to approximately 2, approximately 3, approximately 4, and approximately 5 standard deviations above the median of the normalized values. Attached Figure Description

[0017] Figure 1A and Figure 1B A box plot is provided that compares gas diffusion data from a conventional bulk test with gas diffusion data from an embodiment of a bulk test method with a secondary cutoff point described herein.

[0018] Figure 2A and Figure 2B Provides a box plot with a dataset generated from a conventional gas diffusion holistic test. Figure 2A Display the complete dataset. Figure 2B Display the dataset after removing outliers.

[0019] Figure 3A and Figure 3B Provide a box plot, which is obtained by testing a bowl-shaped object ( Figure 3A ) and membrane casting date ( Figure 3B The datasets in Figures 1 and 2 are compared using subgroups of the datasets.

[0020] Figure 4 Provides box plots that normalize both the test location and the membrane casting date from the same dataset in the aforementioned examples.

[0021] The accompanying drawings illustrate some embodiments of the disclosure herein and should therefore not be considered as limiting the scope, as the invention permits other equally effective embodiments. It should be understood that, unless otherwise stated, elements and features of any embodiment can be found in other embodiments, and where possible, the same reference numerals have been used to denote comparable elements common to the drawings. Detailed Implementation

[0022] The disclosure in this document describes some implementation schemes of the holistic testing method.

[0023] Some implementations of the holistic testing method described herein employ normalization of air diffusion data via a bowl and by the casting date to improve the signal-to-noise ratio and increase defect detection capability. Furthermore, compared to traditional holistic testing, this method is more resistant to variations in membranes and equipment, providing more stringent and consistent product handling without sacrificing yield or reducing sensitivity.

[0024] Some implementations of the holistic testing method described herein normalize air diffusion data by the device test location (bowl-shaped structure) and by the casting date (material and upstream process variations). The resulting variations can be interpreted using normalization, and by understanding these variations, subgroups are added to the method in a predictable manner. Without affecting defect signals, the normalized dataset of the holistic testing method described herein exhibits significantly lower background noise levels, leading to an improved signal-to-noise ratio and a stronger ability to detect outliers (defects) within the normal population.

[0025] For 10-inch filter devices, some implementations of the methods described herein can detect defects as small as 5 micrometers (μm), compared to 10-15 micrometers that can be detected in conventional air diffusion tests.

[0026] Some implementations of the integrity testing method described herein include the following procedures for setting upper and lower cutoff points for a normalized dataset. Some implementations of this method include a dual-cutoff system comprising: an initial fixed cutoff point that identifies critically nonconforming items at the time of testing; and a secondary statistical cutoff point applied to the normalized dataset at the end of the batch.

[0027] Some implementation schemes for holistic testing include the following steps to set initial fixed upper and lower limits for specifications:

[0028] Historical air diffusion data were obtained from a large selection of trailing dates, and the extent of diffusion variation was determined.

[0029] Determine the pressure adjustment value, which can be directly compared with the overall specifications and QC batch release diffusion standards before and after end-user use;

[0030] Determine an upper limit cutoff point that has minimal impact on production and maintains a safety margin of at least 15% from end-user diffusion specifications and QC batch release diffusion standards.

[0031] A lower cutoff point is determined that has minimal impact on output but provides an appropriate limit that, if exceeded, will trigger a re-evaluation of the method.

[0032] I. Set cutoff point

[0033] In some implementations, the initial cutoff point is located far enough from the current air diffusion values ​​to tolerate significant variations in casting dates without impacting production. In other implementations, the initial cutoff point maintains a safety margin from end-user pre- and post-use integrated testing and quality control (QC) batch release test specifications. In some implementations, QC batch release test specifications are adjusted for stress. Integrated testing implementations are more sensitive than both end-user specifications and QC batch release test specifications in some cases.

[0034] In some implementations, the diffusion specification is set using a reference test pressure (i.e., less than 30.0 sccm at 40 psi). In some implementations, if the overall test pressure is not equivalent to an existing reference pressure for the end-user specification or QC specification, a correction factor is determined to compare values ​​from different test pressures.

[0035] In some implementations, secondary cutoff points are applied to batch data that has already been normalized according to the bowl and / or the date of the run-off, to significantly reduce the two largest sources of variation in the data and allow for easier layout of cutoff points for optimal separation of outliers from the normal population.

[0036] Some implementation schemes for holistic testing methods include the following steps to set upper and lower limits for secondary statistical specifications:

[0037] Select a holistic test dataset that represents a statistically significant number of membrane filter devices, casting dates, and any other test or filter production variables that may affect the background noise of the air diffusion test;

[0038] Test the filters using the air diffusion method and remove any seriously substandard filters;

[0039] Run the Ryan-joiner (RJ) normality test;

[0040] Remove outliers until the RJ score is an acceptable value;

[0041] Normalization was performed on the air diffusion data according to the methodology described in this paper;

[0042] The recommended upper specification limit is calculated at 3 to 4 standard deviations above the normalized median, and the recommended lower specification limit is calculated at 4 to 6 standard deviations below the normalized median.

[0043] Evaluate bacterial retention performance of units below, at, and above the available upper cutoff point;

[0044] Consider the defect pattern, defect size, and log reduction value (LRV) of the retained nonconforming items at or below the cutoff point, as well as the sample size;

[0045] Review the retained data and determine the final upper limit cutoff point;

[0046] Assess the proximity of diffusion test results to the lower cutoff point; and

[0047] Review the results and determine the final lower limit of the specifications.

[0048] In some embodiments, the holistic testing method described herein enables the identification of incorrect raw materials used during the manufacturing process, which prior art cannot identify during in-process testing. Some embodiments of the holistic testing method have identified inappropriate processed materials (e.g., incorrectly oriented films), which prior art cannot identify during in-process testing.

[0049] Some implementations of the normalization method adjust for variations in test location (bowl) and batch (casting date). Some implementations include additional sources of variation such as water and air temperature, the casting date of the second film layer, the support material rollers, and the actual prestress or test pressure achieved. Because of the reduced sample size of the subgroups, a trade-off in statistical confidence must be acknowledged when incorporating additional factors into the evaluation.

[0050] Some implementations of the overall testing method can be used in any filter device employing gas-liquid diffusion testing, including but not limited to sterile filters, virus filters, clarifiers, and ultrafiltration filters. Some implementations of the overall testing method described herein maintain sufficient throughput and improve defect detection.

[0051] II. equipment

[0052] Some implementations of the holistic testing approach are performed on a constructed control system to operate 24 hours a day and 365 days a year. In some implementations, a combination of equipment and software infrastructure enables the automated collection of both construction and holistic testing material and process data. In some implementations, the data acquisition system simultaneously collects processing and material information.

[0053] In some implementations, the overall performance testing of the membrane filter is performed in a single-piece-flow manner. Alternatively, batch testing can occur.

[0054] In some implementations of the overall testing method, automated systems use initial (coarse) diffusion specification limits to dispose of non-conforming products in real time. In some implementations, a batch typically contains approximately 1,000 to approximately 1,800 10-inch membrane filters.

[0055] Compared to formulation settings or large batch populations, some implementations of holistic testing methods include a step of marking any test bowls with abnormal results to examine the potential impact of factors arising from results exceeding the expected median and / or standard deviation. Similarly, some implementations of holistic testing methods include a step of marking any film rolls (casting dates) with abnormal results to examine the potential impact of factors arising from results exceeding the expected median and / or standard deviation.

[0056] III. limited

[0057] Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0058] Unless the context clearly specifies otherwise, the singular forms “a,” “an,” and “the” used herein include the plural.

[0059] As used herein, "CF1" refers to the correction coefficient used for normalizing a dataset based on the first characteristic. In some implementations, CF1 is the correction coefficient used for normalizing a dataset by bowl shape to around the median value of 0.

[0060] As used in this paper, "CF1MF" refers to the correction factor used for normalizing mass flow rate.

[0061] As used herein, "CF2" refers to a correction factor used for normalizing the dataset based on the second characteristic. In some implementations, CF2 is a correction factor used to normalize the dataset by master roll (casting date) to around the median value of 0.

[0062] As used in this paper, "CF2MF" refers to the correction factor used for normalizing mass flow rate.

[0063] As used herein, "Initial Fixed Cutoff Point" refers to nonconforming items at the time of testing, similar to a traditional overall test cutoff point. The Initial Fixed Cutoff Point is the Upper Size Limit (USL) and Lower Size Limit (LSL) applied at the time of testing to identify moderate to severe defects. During the testing process, units that fail to meet the Initial Cutoff Point are rejected and removed from the batch.

[0064] As used in this article, “integrated” refers to a defect-free membrane filter.

[0065] As used herein, “non-conforming” refers to a defective membrane filter. In some implementations, a non-conforming membrane filter may also be referred to as “non-integral.”

[0066] As used herein, a "secondary statistical cutoff" refers to the use of applied normalized test data at the end of a batch to identify additional nonconforming items that are outliers in the normal population. The secondary statistical cutoff is a fixed USL and LSL applied to the dataset at the end of the batch after the dataset has been normalized by the bowl and master roller. During the accountability step, cells that do not meet the secondary statistical cutoff are rejected and removed from the batch. The secondary statistical cutoff is established by setting the USL in the range of 2 to 4 standard deviations above the normalized median. In some embodiments, the USL is at least or equal to 2 standard deviations above the normalized median. In some embodiments, the USL is approximately 2 standard deviations above the normalized median. In some embodiments, the USL is at least or equal to 3 standard deviations above the normalized median. In some embodiments, the USL is approximately 3 standard deviations above the normalized median. In some implementations, the USL is at least or equal to four standard deviations above the normalized median. In some implementations, the USL is approximately four standard deviations above the normalized median.

[0067] In some implementations, the secondary statistical cutoff is established by setting the LSL within a range of 4 to 6 standard deviations below the normalized median. In some implementations, the LSL is at least or equal to 4 standard deviations below the normalized median. In some implementations, the LSL is approximately 4 standard deviations below the normalized median. In some implementations, the LSL is at least or equal to 5 standard deviations below the normalized median. In some implementations, the LSL is approximately 5 standard deviations below the normalized median. In some implementations, the LSL is at least or equal to 6 standard deviations below the normalized median. In some implementations, the LSL is approximately 6 standard deviations below the normalized median. The standard deviation can be calculated from characteristics of the membrane filter used in development, such as date-varying batches or bowl-shaped substrates.

[0068] As used in this article, "MRMinQty" refers to the minimum sample size per master roll required for standard processing. If the minimum quantity is not met, an alternative processing rule applies. This is a parameter defined by the formulation.

[0069] As used in this article, "BowlMinQty" refers to the minimum sample size per bowl per batch required for standard processing. If the minimum quantity is not met, an alternative processing rule applies. This is a parameter defined by the formulation.

[0070] As used herein, “MaxBowlVar” refers to the limit of permissible variation between the median value of bowls in a standard treatment and the median value of the average batch bowls. This is a parameter defined by the formulation.

[0071] As used herein, "MaxMRVar" refers to the limit of permissible variation between the standard-treated master roll and the average batch master roll median. This is a parameter defined by the formulation.

[0072] As used herein, "MaxMRStDev" refers to the limit of the permissible standard deviation of the normalized data for each master roll in the standard treatment. The sample standard deviation is used in this calculation. This is a parameter defined by the formulation.

[0073] Example

[0074] Example 1. Experimental Procedure:

[0075] In an additional experimental example, the air diffusion data from the online overall test were processed using an initial fixed cutoff point and a secondary statistical cutoff point procedure as follows. Calculations can be performed manually and / or using a database.

[0076] The variables for mass flow rate, USL, LSL, and CF2MF were rounded to tenths of a decimal place. The variables for CF1, CF1MF, CF2, and any standard deviation (StdDev) were rounded to hundredths of a decimal place. Normalization was used to correct the median values ​​of the subgroups of the dataset. This process normalized the dataset by bowl shape to around the median value of 0. The modified dataset was then normalized by the master roll (cast date) to around the median value of 0.

[0077] Air diffusion data were collected at the end of the batch. Data from membrane filters that did not pass the first cutoff point were removed, including aerosol nonconformities, wet defects, and wet retests. One dataset was used for each membrane filter. If a membrane filter had more than one air diffusion test value, the final diffusion test value was used. The sample size N and the median mass flow rate (BOWL_MEDIAN) for each bowl were calculated from the remaining dataset. Then, a correction factor CF1 for normalization was calculated for each bowl B.

[0078] CF1 B =0-BOWL_MEDIAN B

[0079] Example: CF1 343 =0-BOWL_MEDIAN 343

[0080] The aforementioned steps were completed for all the bowl-shaped objects.

[0081] If N for any bowl (B) is less than BowlMinQty, then the following formula is used instead to calculate CF1 for one or more bowls with low sample sizes. BOWL_MEDIAN AVG It is the average bowl median value of all bowls with a sample size of BowlMinQty or greater.

[0082] CF1 B =0-BOWL_MEDIAN AVG

[0083] Example: Bowl 346 has N=20. This is less than BowlMinQty of 24.

[0084] CF1 346 =0-BOWL_MEDIAN AVG

[0085] Example: Four bowl-shaped objects were used for the test, and MaxBowlVar = 0.9 sccm (standard cubic centimeters).

[0086] BOWL_MEDIAN 343 =12.8sccm

[0087] BOWL_MEDIAN 344 =14.4sccm

[0088] BOWL_MEDIAN 345 =13.8sccm

[0089] BOWL_MEDIAN 346 =14.2sccm

[0090] Calculated BOWL_MEDIAN AVG = (12.8 + 14.4 + 13.8 + 14.2) / 4 = 13.8

[0091] For each membrane filter x in the dataset, the dataset has a mass flow rate (MF) value tested on the bowl (B). X The correction factor (CF1MF) used for normalizing the mass flow rate values ​​was calculated using the following formula:

[0092] CF1MF X =MF X +CF1 B

[0093] Example: Sequence 1001 was tested on a bowl-shaped 345, and this sequence 1001 recorded a mass flow rate of 14.3 sccm. CF1 345 The calculated value is -13.8 sccm.

[0094] CF1MF 1001 =14.3 + (-13.8 sccm) = 0.5 sccm

[0095] The CF1MF dataset was reviewed, and the sample size N and median CF1MF were calculated for each master roll (MR_MEDIAN) used in this batch. Then, a correction factor CF2 for normalization was calculated for each master roll (R) using the following formula: CF2 R =0-MR_MEDIAN R .

[0096] Example:

[0097] CF2 3135UE =0-MR_MEDIAN 3135UE

[0098] CF2 3004UD =0-MR_MEDIAN 3004UD

[0099] The aforementioned steps were completed for all the main rollers.

[0100] Limit validation was performed on the sample size and median value of the master rolls. If N of any master roll (R) is less than MRMinQty, the following formula was instead used to calculate CF2 for one or more master rolls with low sample sizes. R =0-MR_MEDIAN AVG MR_MEDIAN AVG It is the average median value of the remaining master rolls with a sample size of MRMinQty or higher.

[0101] Example: Main roll 3162UE has N=15. This is less than MRMinQty of 24.

[0102] CF2 3162UE =0–MR_MEDIAN AVG

[0103] Example: Four master rollers were used in one batch, and MaxMRVar = 1.5 sccm.

[0104] The data is as follows:

[0105] MR_MEDIAN 5240UE =0.0sccm

[0106] MR_MEDIAN 5241UE =-1.0sccm

[0107] MR_MEDIAN 5242UE =1.7sccm

[0108] MR_MEDIAN 5243UE = -0.7sccm

[0109] Calculated MR_MEDIAN AVG = (0.0 ± 1.0 + 1.7 ± 0.7) / 4 = 0.0

[0110] For each group of membrane filters (x) in the dataset, the dataset has the CF1MF value (CF1MF) of the membrane used for the autonomous roller (R). X ), calculate the CF2MF value using the following formula:

[0111] CF2MF X =CF1MF X +CF2 R

[0112] Example: Group 2001 contains a membrane from main roll 3452UE and has a calculated CF1MF value of 1.1 sccm. CF23452UE is calculated to be -0.8 sccm.

[0113] CF2MF 2001 =1.1 + (-0.8sccm) = 0.3sccm

[0114] Limit verification was performed on the standard deviation of the normalized mass flow rate values ​​for the master rolls. Outliers were removed from the CF2MF dataset, and then the standard deviation of CF2MF was calculated for each master roll. The CF2MF value for each unit was compared with the secondary statistical cutoff points USL and LSL. Any unit above USL or below LSL was disposed of as a non-conforming unit. Defective units from the secondary statistical cutoff points were discarded from the production batch.

[0115] Example 2. Comparing standard data with normalized data

[0116] Figure 1A (CGEP = 10-inch SHF filter, CSTVARLT1 = batch name) is a box plot of air diffusion values ​​from the following batch, which consists of monolithic filters (0.2-micron PES (polyethersulfone) sterile 10-inch cartridge membrane filters; diffusion specification: less than 30.0 sccm at 40 psi) from 11 casting dates. Standard diffusion data show a population with a median of 13.4 sccm and a standard deviation of 0.45 sccm. Potential test specification limits, typically present in the range of 3-4 standard deviations, are located at 14.8-15.2 sccm. The upper limit cutoff has been established at 15.0 sccm, which is 1.6 sccm (3.5 standard deviations) from the population median. Devices with a nominal defect signal of 1.7 sccm or greater are disposed of as non-conforming. At a test pressure of 35 psi, this excess flow rate corresponds to a single cylindrical defect size between 7-8 microns calculated for orifice-type defects and has been experimentally confirmed.

[0117] Figure 1B (CF2MF = Normalized Mass Flow Rate) is a box plot of the same data from the same membrane filter, which was normalized by test bowl and membrane casting date using the holistic testing method described herein. The data was normalized to a median of 0 and a population standard deviation of 0.16 sccm. In the normalized dataset, the upper limit cutoff at 3 standard deviations is located 0.5 sccm from the population median. Devices with a nominal defect signal of 0.6 sccm or greater were disqualified. At a test pressure of 35 psi, this excess flow rate corresponds to a defect size between 4 and 5 urn.

[0118] Observations suggest that normalization of the dataset significantly reduces the overall variation in the diffusion population of the holistic filter. Therefore, the holistic test method presented in this paper offers a higher degree of sensitivity and applies a consistent cutoff point to each device because its final disposal is determined relative to the subpopulation median rather than the absolute diffusion value.

[0119] Example 3. Disposal of sterilization-grade filters

[0120] Observations indicate that the air diffusion integrity test with a secondary cutoff point correctly disqualifies the following sterilization-grade filters from passing the bacterial retention test but passing the conventional air diffusion integrity test. Figure 2A The batches shown have air diffusion data plotted against conventional upper and lower limits for the product, which are 9.5 sccm and 5 sccm, respectively. It was observed that the integral filters formed a normally distributed population between the upper and lower limit cutoffs, and filters with elevated diffusivity (i.e., outliers in the normal population) were disposed of as non-compliant units. Figure 2B The qualified filters from this batch are displayed, in which all non-qualified units have been removed.

[0121] Example 4. Normalization of the test bowl and film casting dates

[0122] When tested by a bowl-shaped object ( Figure 3A ) and membrane casting date ( Figure 3B Additional outliers were observed when the subgroup of CVGL viewed the dataset from the aforementioned examples. Batch C0BB87266 represents 0.2-micron PVDF (polyvinylidene fluoride) sterile 10-inch... Tubular membrane filter.

[0123] Example 5. Disposal of membrane filters that failed the retention test

[0124] exist Figure 4 The box plots in the diagram show the same dataset from the aforementioned embodiments, which was normalized for both test location and membrane casting date. Numerous additional outliers were observed for the overall population, all of which passed the conventional air diffusion overall test. None of these additional outliers also failed the retention test, confirming the benefit of applying this overall testing approach to product quality (product and end-user risk).

[0125] equivalent

[0126] All ranges described herein for formulations include the ranges in between and may include or exclude endpoints. Optional included ranges are integer values ​​(or include a single endpoint) within the range on the stated order of magnitude or the next smaller order of magnitude. For example, if the lower limit of the range is 0.2, optional included endpoints could be 0.3, 0.4, ..., 1.1, 1.2, etc., and 1, 2, 3, etc.; if the upper limit of the range is 8, optional included endpoints could be 7, 6, etc., and 7.9, 7.8, etc. Unilateral limits, such as 3 or more, similarly include a consistent limit (or range) starting from an integer value on the stated order of magnitude or the next smaller order of magnitude. For example, 3 or more includes 4, or 3.1 or more.

[0127] Throughout this specification, the terms "an embodiment," "a particular embodiment," "one or more embodiments," "some embodiments," or "an embodiment" indicate that the described features, structures, materials, or characteristics are included in some embodiments of the invention. Therefore, phrases such as "in one or more embodiments," "in a particular embodiment," "in one embodiment," "some embodiments," or "in an embodiment" appearing throughout this specification do not necessarily refer to the same embodiment.

[0128] All patent applications and patent publications, as well as other non-patent references cited in this specification, are incorporated herein by reference in their entirety as if specifically and individually indicated that each individual publication or reference is also incorporated herein by reference as if fully elucidated. Any patent application for which priority is claimed in this application is also incorporated herein by reference in its entirety in the same manner as described above for publications and references.

Claims

1. A method for reducing background noise and improving the signal-to-noise ratio to handle at least one membrane filter in an overall test, comprising: Gas-liquid diffusion analysis was performed on the membrane filter to obtain the values ​​of the membrane filter; During the gas-liquid diffusion analysis, it is identified whether the value of the membrane filter is an outlier compared to an initial fixed cutoff point, wherein the initial fixed cutoff point includes an upper specification limit and a lower specification limit and is used to identify defective products. If at least one membrane filter is not an outlier for the initial fixed cutoff point, then the values ​​of at least one characteristic or operating condition of a subpopulation of membrane filters are normalized, which result in variations in the gas-liquid diffusion analysis. Identify whether the normalized values ​​of the membrane filter are outliers compared to the secondary statistical cutoff points, wherein the secondary statistical cutoff points include upper and lower limits of specifications based on values ​​from previous gas-liquid diffusion analyses of multiple membrane filters. and The membrane filter is disposed of as either integral or non-conforming, wherein a membrane filter that is an outlier for the initial fixed cutoff point or the secondary statistical cutoff point is disposed of as non-conforming, and a membrane filter that is not an outlier for the initial fixed cutoff point or the secondary statistical cutoff point is disposed of as integral.

2. The method according to claim 1, wherein the gas-liquid diffusion method is air-water.

3. The method according to claim 1 or 2, wherein the at least one characteristic or operating condition is selected from batch, equipment test location, water temperature, air temperature, and pressure.

4. The method according to claim 1 or 2, wherein the values ​​of at least two characteristics and / or operating conditions are normalized.

5. The method according to claim 1 or 2, wherein the method further comprises normalizing the pressure of more than one membrane filter.

6. The method according to claim 1 or 2, wherein the method further comprises detecting at least one defect in more than one membrane filter.

7. The method according to claim 1 or 2, wherein the membrane filter is a 10-inch sterilization filter.

8. The method of claim 6, wherein the at least one defect is selected from between 4 micrometers and 9 micrometers.

9. The method of claim 8, wherein the defect is greater than 4 micrometers.

10. The method of claim 6, wherein the defect is selected from 4 micrometers, 5 micrometers, 6 micrometers, 7 micrometers, 8 micrometers, and 9 micrometers.

11. The method of claim 1, wherein the initial upper limit cutoff point maintains a safety factor of at least 5% from the diffusion specification standard of the end user.

12. The method of claim 1, wherein the initial upper limit cutoff point is maintained at a safety factor of at least 5% from the quality control (QC) batch release diffusion standard.

13. The method according to any one of claims 11 and 12, wherein the initial upper limit cutoff point is maintained with a safety factor selected from a range of 5% to 15% of the end-user diffusion specification standard or quality control (QC) batch release diffusion standard.

14. The method of claim 13, wherein the initial upper limit cutoff point is maintained with a safety factor selected from 5%, 6%, 7%, 8%, 9%, 10%, 11%, 12%, 13%, 14%, and 15%.

15. The method according to claim 1 or 2, wherein the plurality of membrane filters is a statistically significant number of membrane filters.

16. The method according to claim 1 or 2, wherein the membrane filter and the plurality of membrane filters are filter devices of the same type.

17. The method of claim 16, wherein the filter device is selected from sterile filters, virus filters, clarifiers, and ultrafiltration filters.

18. The method of claim 1 or 2, wherein the secondary statistical cutoff point is established by setting the statistical upper limit to a standard deviation between 2 and 5 above the median of the normalized values.

19. The method of claim 18, wherein the secondary statistical cutoff is established by setting the statistical upper limit to 2, 3, 4, and 5 standard deviations above the median of the normalized values.

20. The method of claim 1 or 2, wherein the secondary statistical cutoff point is established by setting the statistical lower limit to a standard deviation between four and six below the median of the normalized values.

21. The method of claim 20, wherein the secondary statistical cutoff point is established by setting the statistical upper limit to 2, 3, 4, and 5 standard deviations above the median of the normalized values.

Images