JP2015184019A - Inspection method and inspection device of hollow fiber module - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an inspection method and inspection device of a hollow fiber module capable of accurately determining the presence/absence of a defect (interruption fiber) of the hollow fiber module and its occurrence degree even when a flaw occurs in an end surface.SOLUTION: In the method of inspecting a hollow fiber module on the basis of a picked-up image of an end surface of the hollow fiber module, regarding the picked-up image, by integrating a blocked hollow fiber identified by a particle analysis method as first means and a blocked hollow fiber identified by a pattern matching method as second means (calculating the logical OR, logical AND, or exclusive OR), the blocked hollow fiber is identified without missing out.

Description

  The present invention relates to a hollow fiber module inspection method and inspection apparatus that can be used in an artificial dialyzer, a sewage treatment apparatus, a water purifier, and the like.

  As is well known, for example, a hollow fiber module accommodates a bundle of hollow fibers (hollow semipermeable membrane) whose ends are opened in a container whose both ends or one end is opened, and ends of the yarn bundle. The gap between the container and the hollow fiber is sealed with a resin (hereinafter also referred to as “potting material”). This sealing (hereinafter sometimes referred to as “potting”) is performed by pouring resin into a space between the container and the yarn bundle or between the hollow fibers and hardening the resin, and fixing the container and the yarn bundle with the resin. However, at this time, the resin may enter the hollow portion of the hollow fiber and the hollow fiber may be blocked. If this blocked hollow fiber (hereinafter sometimes referred to as “non-threaded yarn”) is present in the hollow fiber module, for example, in the case of artificial dialysis, blood that has flowed into the hollow fiber remains in the hollow fiber. The dialysis performance is lowered, and if it is remarkable, sufficient artificial dialysis cannot be performed. Therefore, a hollow fiber module in which a certain number or more of thread breakage exists needs to be removed from the production line as a defective product.

  Conventionally, a hollow fiber module pass / fail inspection based on the number of non-threaded yarns is obtained by irradiating light to both ends or one end face of the hollow fiber module and imaging reflected light from the end face. The hollow fiber (normal yarn) in which the hollow portion is hollow is detected by correlating the captured image with a predetermined model image, the normal yarn is masked, and the hollow fiber portion and the resin portion are captured in the captured image. The defect candidate is detected by performing the binarization process with the above, and the hollow fiber in which the hollow portion is blocked is identified from the feature amount of the defect candidate (hereinafter, sometimes referred to as “defect, non-threaded yarn”), A method is known in which the hollow fiber module is determined to be defective when the number of non-threaded yarns exceeds a predetermined threshold (see, for example, Patent Document 1).

  Here, the feature amount refers to an area, luminance value, width or height of a circumscribed rectangle (not limited to this) as image information in a defect candidate, and an arbitrary threshold value is set for these feature amounts ( The defect candidate can be labeled as good / bad by combining a plurality of combinations.

  The mask will be described in detail in an embodiment for carrying out the invention.

JP 2008-32601A

  However, in the above prior art, when the end face is scratched when the end face is cut due to cutter spillage or hardness variation when the potting material is hardened, a threshold value is set for each feature amount of the defect candidate to prevent the defect. Even if the yarn is to be detected, there are various kinds of scratches, so that it may be impossible to accurately detect the defect due to the influence of the feature amount of the defect candidate, for example, the area and the luminance unevenness. In this case, there is a problem that it is difficult to detect the threshold value for defect determination with respect to how to enter any scratches that may occur in the non-threaded portion, and it is difficult to detect all the non-threaded yarns. (Problems of missing defects).

  An object of the present invention is to provide a hollow fiber module inspection method and inspection apparatus capable of accurately determining the presence or absence of a defect (non-threading) in a hollow fiber module and the degree of occurrence of the defect even when the end surface is flawed. Is to provide.

  In order to achieve the above object, the present invention provides a hollow fiber module inspection method based on a captured image of an end face of a hollow fiber module, and a hollow fiber model in which a hollow portion is hollow with respect to the captured image. After detecting the hollow fiber in which the hollow portion is hollow by correlating with the image and masking the hollow fiber in which the hollow portion is hollow, the hollow fiber portion and the resin portion ( Binarization processing with a potting material portion), a defect (hollow fiber in which the hollow portion is blocked) candidate is detected from the binarization processing image, and the defect candidate is compared with a predetermined feature amount Identify the defect (first inspection) and correlate the captured image with the model image of the hollow fiber in which the hollow part is closed to identify the hollow fiber in which the hollow part is blocked as a defect. (No. Inspection), by integrating the detection results of the second test and the first test, to provide an inspection method of a hollow fiber module for detecting defects.

  The unthreaded yarn is identified by the first inspection (hereinafter, mainly described by a particle analysis method) by the following procedure. (1) The degree of correlation R with a predetermined normal yarn model is obtained by normalized correlation for each pixel group centered on each pixel with respect to all pixels of the captured image. (2) Only pixels with R ≧ Rt are selected for the normal yarn correlation degree threshold value Rt. (3) A mask is applied to the selected pixel group. (4) Binary processing is performed to detect defect candidates. (5) The thread breakage is specified with a predetermined threshold from the feature amount of the defect candidate.

  The non-threading is specified by the following procedure by the second inspection (hereinafter, mainly described by using a pattern matching method). (1) For each pixel group centered on each pixel with respect to all pixels of the captured image, a correlation degree R ′ with a predetermined thread-removal model is obtained by normalized correlation. (2) Only pixels with R ′ ≧ Rt ′ are selected for the non-thread-correlation correlation threshold Rt ′. (3) Identify a thread breakage.

  Moreover, in the inspection method of the hollow fiber module of the present invention, it is preferable to set the mask obtained by the process of the particle analysis method to the inspection execution range by the pattern matching method.

  The method for inspecting a hollow fiber module of the present invention is suitable for inspecting an artificial kidney module or a water treatment module.

  In order to achieve the above object, the present invention provides means for irradiating light to an end face of a hollow fiber module, means for imaging reflected light from the end face, a captured image obtained by the imaging means, After detecting the hollow fiber in which the hollow portion is hollow by correlating with the model image of the hollow fiber in which the hollow portion to be defined is hollow, and masking the hollow fiber in which the hollow portion is hollow Means for performing binarization processing of the hollow fiber portion and the resin portion (potting material portion) on the captured image to detect a defect (hollow fiber in which the hollow portion is blocked), and the defect candidate Non-thread passing specifying means A for determining a feature amount with a predetermined threshold and specifying a defect, a captured image obtained by the imaging means, and a model image of a hollow fiber in which a predetermined hollow portion is in a closed state Correlate A non-threading specifying means B for specifying a hollow fiber in which the empty portion is closed as a defect, a means for calculating the number of defects by integrating the inspection results of the non-threading specifying means A and the non-threading specifying means B; A hollow fiber module inspection apparatus comprising:

  The inspection apparatus for a hollow fiber module of the present invention preferably includes means for determining that the hollow fiber module is a defective product when the number of defects (the number of unthreaded yarns) exceeds a predetermined threshold.

According to the hollow fiber module inspection method and hollow fiber module inspection apparatus of the present invention, the following effect of preventing missed yarn and improvement in productivity can be obtained.
(1) It is possible to detect the thread-missing means missed by the thread-missing specifying means A due to the various ways of entering flaws with the thread-missing specifying means B having a different detection method.
(2) The normal thread mask created by the non-thread passing means A in the middle of the process determines the range to which the non-thread passing means B is applied, so that high-accuracy non-thread check can be performed at a higher speed.

It is a figure which shows the whole structure of the inspection apparatus of the hollow fiber module in one Embodiment of this invention. It is a captured image of a hollow fiber membrane module. Fig. 3 is an enlarged view of a part of the image shown in Fig. 2. It is a figure which shows the flow of the process in this invention. It is an image of a normal yarn model. FIG. 4 is an image of a normal yarn mask created from the normal yarn detected in FIG. 3. FIG. It is the image which masked the normal thread of the image shown in FIG. FIG. 3 shows the result of the first inspection. It is an image of a non-threading model. FIG. 3 shows the result of the second inspection. It is an image which shows the result of having integrated the result of the 1st inspection and the 2nd inspection.

  Preferred embodiments of the present invention will be described below with reference to the drawings.

  FIG. 1 shows the overall configuration of a hollow fiber module inspection apparatus according to an embodiment of the present invention. The hollow fiber modules M are inspected one by one. However, a plurality of hollow fiber modules M can be inspected simultaneously.

Hereinafter, the configuration of the inspection device for the hollow fiber module will be described together with its operation.
<Module transfer device>
The module transfer device 1 is for mounting the hollow fiber module M transferred from the previous process on a module support device 2 described later.

<Module support device>
The module support device 2 supports the hollow fiber module M placed by the module transfer device 1. In addition, in order to keep the distance between the end face of the hollow fiber module M and the imaging device 4 to be described later, the position of the end face may be adjusted by a stopper jig. Further, the position of the end face may be measured by a length measuring device, and the imaging device 4 may be moved to keep the focus properly. In this embodiment, since a line sensor camera is used for the imaging device 4 described later, the hollow fiber module M placed on the module support mechanism 2a is moved in the horizontal direction at a predetermined speed by the module scanning mechanism 2b. An image is taken in between.

<Light irradiation device>
The light irradiation device 3 includes a light source 3a and a power source 3b, and is not particularly limited in the present invention, but the light source 3a is preferably a planar light source. In this embodiment, a planar light source that emits white light from a halogen lamp is used as the light source 3a. However, the present invention is not limited to a planar light source, as long as it can irradiate light in a planar shape to such an extent that the reflected light of the entire end surface of the hollow fiber module M to be measured can be imaged. Since the light-emitting surface is applicable to implementation, a diffusing means is installed in a linear line light source having a one-dimensional extension such as a point light source such as an LED or a light bulb or a rod-shaped fluorescent light, and a planar light source As long as the same effect can be obtained, there is no problem in application.

  The power supply 3b has a function of turning on and off the light source 3a and adjusting the amount of light according to a command from an operation commanding device 7 described later.

<Imaging device>
The imaging device 4 is for capturing an image of the end face of the hollow fiber module M on the module support device 2 to obtain an image. In this embodiment, a line sensor camera having 4096 pixels is used. Further, in this embodiment, the end surface of the hollow fiber module M shown in FIG. 3 is such that the flat resin portion 10 has high brightness, the hollow fiber portions 11 and 11 ′, and the hollow fiber interior 12 of the normal yarn 13 having a cavity. Is imaged with low brightness. Since the hollow yarn interior 12 ′ is filled with resin, the non-threaded yarn 14 is imaged with high brightness in the same manner as the resin portion 10.

<Hollow fiber determination device>
The hollow fiber determining device 5 includes a non-threading specifying means A (here, apparatus A section) for specifying non-threading by a particle analysis method, and a non-threading specifying means B (here, a device) for specifying non-threading by a pattern matching method. B part).

  In the apparatus A section, first, a normal thread 13 in which the hollow fiber interior 12 is hollow is detected by correlating the captured image obtained by the imaging apparatus 4 with a predetermined normal thread model image (FIG. 5). Then, the normal yarn 13 is excluded from the defect inspection target by performing a mask process on the region where the normal yarn 13 exists. Next, binarization processing is performed in addition to the mask region of the captured image, and the hollow fiber portion 11 ′ of the non-threaded yarn in which the hollow fiber interior 12 ′ is blocked is the light region (or dark region) and the resin portion 10 is the dark region (or Classified as a bright region) and recognized as the defect candidate. Furthermore, each feature amount of the detected defect candidate is determined with a predetermined threshold value and specified as a defect.

  In the apparatus B part, a hollow fiber in which the hollow fiber interior 12 ′ is blocked is detected as a non-thread 14 by correlating a captured image obtained by the imaging device 4 with a predetermined non-thread model and an image. The defect is identified as it is.

  The processing flow of the hollow fiber determination device 5 is as shown in FIG. The left side is the flow of processing by the particle analysis method, and the right side is the flow of processing by the pattern matching method.

  The flow of processing by the particle analysis method will be described.

  FIG. 5 shows an image of a normal thread model. A method of correlating the captured image (FIG. 2) obtained by the imaging device 4 with a predetermined normal thread model image (FIG. 5) is based on the center pixel of the normal thread model image (FIG. 5) as a reference. By applying the center pixel of the normal yarn model image (FIG. 5) in order to each pixel of the target captured image (FIG. 2), calculation is performed using Equation 1 for all pixels in the captured image (FIG. 2). When the calculated correlation degree R exceeds a predetermined normal thread correlation threshold value Rt, it is determined that the pixel of the captured image (FIG. 2) is included in the normal thread. Is.

  Here, the normalized correlation is obtained by normalizing the degree of correlation with the magnitude of the luminance value of the input image. When the pixel group of the normal yarn model image is f and the pixel group of the captured image is g, The degree of correlation R is obtained by the above formula 1. By using Expression 1, the normal thread model and the linear change (fixed gain and offset) of the captured image are not affected.

  The correlation method includes a correlation method, a vector correlation method, and the like, which can absorb a change in luminance of the captured image due to a change in the amount of light emitted from the light source 3a, and is a normalized correlation highly resistant to noise. Although the method was selected in particular this time, it is preferable to use them properly depending on the state of the system.

  Mask processing is to cover a part of an image with a uniform brightness value (in many cases, the minimum brightness or the maximum brightness in the brightness gradation of the image), and then inspect the area that has been filled and hidden. It means to exclude from the range. An example of a mask created based on FIG. 3 is shown in FIG. In FIG. 3, a predetermined range 15 centered on pixels determined to be included in the normal yarn by the normalized correlation (in the example of this embodiment, a circle having a design external value of the hollow fiber is adopted as the range 15). FIG. 6 was obtained by painting with a uniform luminance value (minimum luminance). In FIG. 6, only the lowest luminance portion 15 shown in black has a function as a mask, and mask processing is applied to FIG. 3 (the information of FIG. 6 is applied to FIG. 3), as shown in FIG. Get an image. That is, by performing the mask process, the normal yarn 13 portion in FIG. 3 is excluded from the inspection range.

  By performing binarization processing on FIG. 7, the hollow fiber portion 11 ′ of the unthreaded yarn 14 is classified and recognized as a bright region (or dark region) and the resin portion 10 is classified as a dark region (or bright region). Further, the former including the hollow fiber interior 12 ′ is detected as a defect candidate. Further, each feature amount (area, particle diameter, average luminance, luminance variation, etc.) of the portion recognized as a defect candidate is determined based on a predetermined threshold value and specified as a thread breakage.

However, there are cases where the feature amount cannot be detected correctly, and in this case, the determination cannot be made accurately. For example, if a scratch exists on the end face and a defect candidate overlaps, the defect candidate causes luminance unevenness, etc., and the feature quantity such as average luminance or luminance variation is not matched with the defect candidate. There may be a difference between them, and defect candidates that overlap with scratches may not be correctly recognized as thread breakage.
Alternatively, even if the binarization process is performed with a uniform threshold value, the hollow fiber portion 11 ′ and the hollow portion 12 ′ of the non-thread 14 may not be correctly classified. That is, the hollow portion 12 ′ that is originally recognized as one defect candidate for one thread breakage is recognized as a defect candidate in a state where it is divided into two or more, so that the area and particle shape And the like may be different from the defect candidates that do not overlap with the scratches, and may not be correctly recognized as thread breakage.
As a result, as shown in FIG.

  Next, the flow of processing by the pattern matching method will be described.

  A method of correlating a captured image (FIG. 2) obtained by the imaging device 4 with a predetermined non-threaded model image (FIG. 9) is based on the center pixel of the non-threaded model image (FIG. 9). By applying the center pixel of the non-threaded model image (FIG. 9) in order to each pixel of the target captured image (FIG. 2), the correlation by normalized correlation is applied to all the pixels in the captured image (FIG. 2). The degree R ′ is calculated, and when the calculated correlation degree R ′ exceeds a predetermined normal yarn correlation degree threshold value Rt ′, it is determined that the pixel of the captured image is included in the non-threaded yarn. In the pattern matching method, it is determined that the thread breakage is specified at this stage.

  Even if the thread is not properly recognized by the particle analysis method due to the scratch, the pattern matching method can be correctly detected as a thread failure as shown in FIG. 10 by setting the correlation threshold Rt ′ to an appropriate value.

  The processing by the particle analysis method and the processing by the pattern matching method are performed independently for each captured image. Therefore, it is more preferable to perform integration by performing logical operations such as logical sum, logical product, exclusive logical sum, etc. for the non-threads detected by the particle analysis method and the pattern matching method.

  FIG. 11 shows an image obtained as a result of logical sum of the non-threading detected by the particle analysis method and the non-threading detected by the pattern matching method.

  In addition, the particle analysis method and the pattern matching method can be separately applied to the entire captured image as the target region. However, when the pattern matching method is considered as an auxiliary means of the particle analysis method, the method described later is used. By limiting the application area, the time required for the entire inspection can be shortened.

  For example, it is conceivable that a mask set in the course of performing the above-described particle analysis method is set in an execution range of the pattern matching method, that is, only an area where the mask is not set is set as an application area. Or it is good also considering the surrounding NxN pixel as an application area | region centering on each pixel contained in the area | region recognized as a defect candidate by the particle | grain analysis method.

  If the application area of the pattern matching method is reduced by the processing as described above, there is an advantage that the time spent for inspection is shortened.

By the particle analysis method and the pattern matching method, the number N of integrated unthreaded yarns is calculated from the result of logical operation such as logical sum, logical product, exclusive logical sum, etc. for the unthreaded yarns detected by each of them.
<Module determination device>
The module judgment device 6 compares the number N of thread breakage obtained by the hollow fiber judgment device 5 with a predetermined number of thread breakage threshold Nt. If the number N of thread breakage exceeds the threshold Nt, the hollow fiber module M Is determined to be defective.
<Operation command device>
The operation command device 7 corresponds to the difference in the outer diameter of the hollow fiber module M to be measured, the outer diameter and inner diameter of the hollow fiber, the assembly method of the module, etc., and is set in advance in the hollow fiber determination device 5 and the module determination device 6. Sending various inspection conditions (normal yarn model image suitable for the empty yarn module M, normal yarn correlation threshold Rt, mask size Dm, non-threading model image, non-threading correlation threshold Rt ′, etc.) is there. Further, a command is sent to the power source 3b of the light irradiation device to turn on and off the light source 3a and adjust the light quantity. Further, a command is sent to the module support device 2 so that the moving speed of the hollow fiber module M becomes a speed suitable for imaging by the imaging device 4.
<Display device>
The display device 8 displays the inspection result.
<Module ejector>
The module discharge device 9 pays out the non-defective hollow fiber module to the next process and the defective hollow fiber module to the defective product discharge line according to the determination by the module determination device 6.

  The manufacturing method of the hollow fiber module in the present invention includes a step of inspecting the end face using the above-described inspection apparatus and method, and is excellent in production efficiency by automatically eliminating defective products by the above-described inspection. In addition, the hollow fiber module according to the present invention is characterized by being particularly excellent in quality because it is manufactured by such a manufacturing method.

  The present invention will be specifically described with reference to the following examples. However, the present invention is not limited to this.

FIG. 2 shows a light irradiation device 3 including a light source (here, a fiber optic light guide MPP90-1500S-2 manufactured by Moritex Corporation) and a power source (here, a halogen lamp light source MHAB-150W-D-100V manufactured by Moritex Corporation). A hollow fiber module in which reflected light (including scattered light) at the end face of the light irradiated to the end face of the hollow fiber module M is photographed as an 8-bit monochrome image by the imaging device 4 (here, a line sensor camera NSUF4010 manufactured by NED). In the image of the end face of M, the number of hollow fibers is approximately 10,000.
Example 1
FIG. 8 shows the result (indicated by white circles) of specifying the non-threaded thread in this image by the method (first inspection) according to Patent Document 1. As described above, in the first inspection method, the hollow thread module having a flaw on the end face could not be recognized as a defect at a portion where the fissure and the flaw overlap.

  Next, with respect to the image shown in FIG. 3 in which a part of the image shown in FIG. 2 is enlarged, the method of taking a correlation with the model image of the non-threading shown in FIG. 9 (second inspection) was performed. The results are shown in FIG. The thread that overlapped with the scratch that could not be identified as a defect in the first inspection could be recognized as a defect (while the thread that was identified as a defect in the first inspection could be recognized as a defect in the second inspection. In some cases).

  Thus, it was found that by applying two or more types of inspections having different detection characteristics to the same object, what cannot be detected on the one hand can be detected on the other hand.

Finally, a logical sum is obtained with respect to the results of the first inspection and the second inspection, thereby obtaining the number N of unthreaded yarns (79 in total in this example) included in the hollow fiber module shown in FIG. As a result of comparing the hollow fiber module threshold Nt (50 in this example) with the hollow fiber module, the detected number N of hollow fibers exceeded the threshold Nt, and the hollow fiber module was excluded from the manufacturing process as a defective product. It was.
(Example 2)
In the captured image, the application area of the second inspection method is defined using the normal thread mask by utilizing the fact that no thread can exist where normal thread is present. The inspection process, which took 520 seconds, has been shortened to 105 seconds. Of course, compared with the first embodiment, there is no difference in the specific number of non-threaded yarns and the quality determination of the module.

  The present invention can be used for the inspection of hollow fiber modules that can be used for artificial kidneys (artificial dialyzers), water treatment (sewage treatment devices, water purifiers, etc.) and the like.

C Non-thread M Hollow fiber module 1 Module transfer device 2 Module support device 2a Module support mechanism 2b Module scanning mechanism 3 Light irradiation device 3a Light source 3b Power source 4 Imaging device 5 Hollow fiber determination device 6 Module determination device 7 Operation command device 8 Display Device 9 Module discharge device 10 Resin part 11 Hollow fiber part (normal thread)
11 'hollow fiber part (non-threaded)
12 Inside hollow fiber (normal thread)
12 'normal thread inside (non-threaded)
13 Normal thread 14 Non-thread 15 Minimum brightness area

Claims (6)

In the inspection method of the hollow fiber module performed based on the captured image of the end surface of the hollow fiber module,
A hollow fiber in which the hollow portion is hollow is detected by correlating the captured image with a model image of a hollow fiber in which the hollow portion is in a hollow state, and the hollow portion is hollow. After the mask processing is performed, binarization processing of the hollow fiber portion and the resin portion is performed on the masked captured image, and a defect (hollow fiber in which the hollow portion is blocked) candidate from the binarization processing image And detecting the defect by comparing the defect candidate with a predetermined feature amount (first inspection),
Correlating the captured image with a model image of a hollow fiber in which the hollow portion is in a closed state, and detecting the hollow fiber in which the hollow portion is closed as a defect (second inspection),
A method for inspecting a hollow fiber module that detects a defect by integrating the detection results of the first inspection and the second inspection.   The method for inspecting a hollow fiber module according to claim 1, wherein, in the captured image, a mask created in the first inspection is set in an implementation range of the second inspection.   The hollow fiber module inspection method according to claim 1 or 2, wherein the hollow fiber module is an artificial kidney module or a water treatment module.   The inspection method for a hollow fiber module according to any one of claims 1 to 3, wherein the hollow fiber module is determined to be defective when the number of the closed hollow fibers exceeds a predetermined threshold value. Means for irradiating light to the end face of the hollow fiber module; means for imaging reflected light from the end face;
Means for detecting a hollow fiber in which the hollow portion is hollow by correlating a captured image obtained by the imaging means with a model image of a hollow fiber in which the predetermined hollow portion is in a hollow state; and After the hollow fiber that is hollow is masked, the imaged image that has been masked is binarized into a hollow fiber part and a resin part to obtain a defect (hollow fiber in which the hollow part is blocked). ) A means for specifying a candidate, and a non-threading specifying means A for detecting a defect by comparing the defect candidate with a predetermined feature amount;
A non-thread-passage specifying means B for detecting a hollow fiber in which the hollow portion is closed as a defect by correlating a captured image obtained by the imaging means with a model image of a hollow fiber in which the predetermined hollow portion is in a closed state; ,
An inspection device for a hollow fiber module, comprising: means for calculating the number of defects by integrating detection results obtained by the non-thread passing means A and the non-thread specifying means B. The hollow fiber module inspection device according to claim 5, further comprising means for determining that the hollow fiber module is defective when the integrated number of defects exceeds a predetermined threshold.