WO2022080416A1

WO2022080416A1 - Performance evaluation system, performance evaluation method, program, and learned model

Info

Publication number: WO2022080416A1
Application number: PCT/JP2021/037909
Authority: WO
Inventors: 敦島崎
Original assignee: 三菱ケミカルアクア・ソリューションズ株式会社
Priority date: 2020-10-13
Filing date: 2021-10-13
Publication date: 2022-04-21
Also published as: KR20230086659A; JPWO2022080416A1

Abstract

This performance evaluation system comprises an evaluation unit which, using a learned model machine-learned on the basis of a learning image of an ion-exchange resin captured for learning and an evaluation result of the appearance of the ion-exchange resin, evaluates the performance of an ion-exchange resin to be inspected from an inspection image in which the ion-exchange resin to be inspected is captured.

Description

Performance evaluation system, performance evaluation method, program, and trained model

The present invention relates to a performance evaluation system, a performance evaluation method, a program, and a trained model.
This application claims priority under Japanese Patent Application No. 2020-172521 filed in Japan on October 13, 2020 and Japanese Patent Application No. 2021-159660 filed in Japan on September 29, 2021. Incorporate the content here.

Ultrapure water is used in various applications such as washing water used in the manufacturing process of semiconductors, liquid crystals, wafers, precision parts, desalted water manufactured by condensing desalination equipment at power plants, and water for manufacturing pharmaceuticals. .. Ultrapure water includes MF (microfiltration) membrane, UF (ultrafiltration) membrane, RO (reverse osmosis filtration) membrane, ion exchange resin, EDI (continuous electric regeneration type pure water system), ultraviolet water sterilizer, and degassing. A wide variety of devices, such as devices, are used and manufactured in combination. Ion exchange resins are used as the main member among these. Such ultrapure water production processes include a system in which an ion exchange resin tower is mainly used in a single bed, a system in which an ion exchange resin tower is used in a mixed bed, and a system in which these are combined. (See, for example, Patent Document 1).

Japanese Unexamined Patent Publication No. 2014-104413

For example, a vendor who manufactures an ion exchange resin and provides it to a customer regularly receives a sample of the ion exchange resin used in the ion exchange resin tower from the customer and evaluates the performance. The appearance index is one of the performance evaluations of ion exchange resins. A new ion exchange resin is a perfect sphere, but when it deteriorates, it cracks or cracks. The appearance index is calculated by observing 300 or more ion exchange resins one by one with a magnifying glass for one sample and dividing them into appearance types (categories) such as complete spheres, cracked spheres, and crushed spheres. .. The appearance index is an index of the degree of deterioration of the ion exchange resin, and serves as a guide for replacing the ion exchange resin used in the customer's ion exchange resin tower.

However, in the above-mentioned performance evaluation of the ion exchange resin, the analyst visually inspects the ion exchange resin one by one using a magnifying glass to calculate the appearance index, and even a skilled person spends a lot of time. I needed it. As the magnifying glass, a microscope, a camera with a magnifying function, or the like can be used.

The present invention has been made in view of the above points, and an object of the present invention is to provide a performance evaluation system, a performance evaluation method, a program, and a trained model capable of accurately evaluating the performance of an ion exchange resin in a short time. ..

The present invention has been made to solve the above problems, and one aspect of the present invention is based on a learning image of an ion exchange resin taken for learning and an evaluation result of the appearance of the ion exchange resin. It is a performance evaluation system including an evaluation unit for evaluating the performance of the ion exchange resin for inspection from an inspection image in which the ion exchange resin for inspection is photographed by using the trained model trained by the machine.

Further, in the performance evaluation system, as the evaluation result of the appearance, classification according to the type of appearance of the ion exchange resin is preset, and the evaluation unit has a plurality of ions photographed for each type of appearance. The performance of the ion exchange resin for inspection may be evaluated using a trained model machine-learned based on the learning image of the exchange resin.

Further, in the performance evaluation system, the type of appearance is classified based on at least one of the presence / absence of cracks and the presence / absence of crushing in the appearance of the ion exchange resin, and the evaluation unit is the ion exchange resin for inspection. With the performance as an evaluation, an appearance index indicating the appearance state of the ion exchange resin may be calculated.

Further, in the performance evaluation system, the evaluation unit inputs the inspection image in which a plurality of ion exchange resins for inspection are taken into the trained model, and the evaluation unit uses the trained model for the inspection. It is analyzed which type of the appearance is classified into each of the ion exchange resins of the above, and based on the number of the ion exchange resins for each type of the classified appearance, the inspection image included in the inspection image is included. The appearance index indicating the appearance state of the ion exchange resin may be calculated.

Further, in the performance evaluation system, the inspection image is an image taken by magnifying a plurality of ion exchange resins for inspection contained in a container with a magnifying mirror camera, and is taken by the magnifying mirror camera. The image may be taken by selecting a portion where there is little overlap between the ion exchange resins to be photographed when the image is taken.

Further, in the performance evaluation system, the inspection image is stirred by adding a surfactant to a container containing a plurality of ion exchange resins for inspection before being photographed by a magnifying glass camera. It may be an image taken by a magnifying glass camera.

Further, the performance evaluation system may include a learning unit for machine learning based on the learning image of the ion exchange resin taken for learning and the evaluation result of the appearance of the ion exchange resin.

Further, in the performance evaluation system, the learning image is an image taken a plurality of times with the appearance of the ion exchange resin for learning changed at least one of size, angle and color tone, or an ion exchange resin. The image obtained by capturing the appearance may include a plurality of images generated by changing at least one of size, angle, and color tone.

Further, one aspect of the present invention is a method for evaluating the performance of an ion exchange resin, in which a step of acquiring an inspection image in which the ion exchange resin for inspection is photographed and a learning of the ion exchange resin photographed for learning are performed. Performance having a step of evaluating the performance of the ion exchange resin for inspection from the inspection image using a trained model machine-learned based on the image and the evaluation result of the appearance of the ion exchange resin. It is an evaluation method.

Further, one aspect of the present invention includes a step of acquiring an inspection image in which an ion exchange resin for inspection is photographed on a computer, a learning image of the ion exchange resin photographed for learning, and the ion exchange resin. It is a program for executing a step of evaluating the performance of the ion exchange resin for inspection from the inspection image using a trained model machine-learned based on the evaluation result of appearance.

Further, one aspect of the present invention is a trained model for evaluating the performance of the inspection ion exchange resin from the inspection image in which the inspection ion exchange resin is photographed, and is photographed for learning. Machine learning is performed based on the learning image of the ion exchange resin and the evaluation result of the appearance of the ion exchange resin, and the computer is made to function so as to evaluate the performance of the ion exchange resin for inspection from the inspection image. It is a trained model.

Further, one aspect of the present invention is a performance evaluation system including a water treatment facility and a performance evaluation device, wherein the water treatment facility comprises an ion exchange resin tower and an ion exchange resin in the ion exchange resin tower. The performance evaluation device includes a communication unit for taking an image and a communication unit for transmitting an image taken by the image pickup unit as an inspection image, and the performance evaluation device is photographed for learning with the communication unit for receiving the inspection image. The performance of the ion exchange resin for inspection is evaluated from the inspection image using a trained model machine-learned based on the learning image of the ion exchange resin and the evaluation result of the appearance of the ion exchange resin. It is a performance evaluation system equipped with an evaluation unit.

Further, in one aspect of the present invention, in the performance evaluation system, the performance evaluation device may transmit information based on the evaluation result by the evaluation unit to the water treatment equipment.

Further, in one aspect of the present invention, the water treatment equipment generates order information of an ion exchange resin for replacement when the ion exchange resin needs to be replaced based on the information acquired from the performance evaluation device. May be good.

According to the present invention, the performance of the ion exchange resin can be evaluated accurately in a short time.

The figure which shows an example of the type of appearance of the ion exchange resin which concerns on embodiment. The figure which shows the outline about the performance evaluation method of the ion exchange resin which concerns on embodiment. The system diagram which shows an example of the structure of the performance evaluation system 1 which concerns on embodiment. The figure which shows an example of the GUI screen which concerns on embodiment. The block diagram which shows an example of the structure of the camera 10b which concerns on embodiment. The block diagram which shows an example of the structure of the performance evaluation apparatus 30 which concerns on embodiment. The figure which shows an example of the judgment criteria of the performance of the ion exchange resin which concerns on embodiment. The flowchart which shows an example of the ion exchange resin evaluation process which concerns on embodiment. The block diagram which shows an example of the structure of the machine learning apparatus 50 which concerns on embodiment. An explanatory diagram illustrating an execution procedure in machine learning according to an embodiment. An explanatory diagram illustrating a learning procedure in machine learning according to an embodiment. The figure which shows the analysis result by AI which concerns on embodiment and the analysis result by visual observation. The figure which compared the analysis result by AI which concerns on embodiment and the analysis result by visual observation. The figure which shows the example which is misjudged as a crack ball before the addition of the anionic surfactant which concerns on embodiment. The figure which shows the example which is erroneously judged as a crushed ball before the addition of the anionic surfactant which concerns on embodiment. The figure which shows the example which cannot determine the part where the resin overlaps before the addition of the anionic surfactant which concerns on embodiment. The figure which shows an example of the analysis result of the inspection image after the addition of the anionic surfactant which concerns on embodiment. The system diagram which shows another example of the configuration of the performance evaluation system which concerns on embodiment.

Hereinafter, an embodiment of the present invention will be described with reference to the drawings.
The ion exchange resin adsorbs cations such as sodium, calcium, and magnesium contained in water and anions such as chlorine and carbon dioxide, and releases the ions that it originally has to exchange ions, resulting in impurities from the water. Get rid of. For example, ultrapure water used for various purposes such as washing water used in the manufacturing process of semiconductors, liquid crystals, wafers, precision parts, desalted water manufactured by a power plant condensate desalination device, and water for manufacturing pharmaceuticals. Used in manufacturing. The ultrapure water production process includes a system in which the ion exchange resin tower is mainly used in a single bed, a system in which the ion exchange resin tower is used in a mixed bed, and a system in which these are combined.

Ion exchange resins are roughly classified into "gel type" and "porous type" according to their structure, and there are cation exchange resin and anion exchange resin, respectively. Since the gel type resin has a larger ion exchange capacity per volume than the porous type, it is considered to be advantageous for producing ultrapure water, but on the other hand, it has a drawback that the cycle strength is lower than that of the porous type. In addition, since gel-type resins generally have a smaller specific surface area than porous-type resins, there is no problem in adsorbing ordinary inorganic ions (chloride ions, etc.), but in order to adsorb high-molecular-weight substances. It is disadvantageous.

The new ion exchange resin is a perfect sphere, but when it deteriorates, it cracks or cracks. When the ion exchange resin deteriorates, the ion exchange performance deteriorates, so it is necessary to replace it. As an index of the degree of deterioration of the ion exchange resin, there is an appearance index indicating the state of appearance. The appearance index is a value calculated according to the appearance state of the ion exchange resin.

For example, based on the appearance of cracks and cracks caused by deterioration of the ion exchange resin, cracked spheres with cracks, crushed spheres with crushing, complete spheres without cracks or crushing, etc. Classify by the type (category) of the appearance of. FIG. 1 is a diagram showing an example of the type of appearance of an ion exchange resin. The illustrated example shows an example of each appearance photograph when the difference in appearance due to deterioration of the "gel type" ion exchange resin is classified into three types of appearance: perfect sphere, cracked sphere, and crushed sphere. ..

For example, conventionally, for one sample, an analyst visually observes 300 ion exchange resins one by one using a magnifying glass (microscope, etc.) and classifies them into complete spheres, cracked spheres, and crushed spheres. The appearance index, which is an index of the deterioration of the ion exchange resin, was calculated by Equation 1 and the performance was evaluated.

Appearance index (%) = (300-total number of crushed balls) / 300 × 100 ... (Equation 1)

However, it is often difficult for an analyst to visually observe 300 ion exchange resins one by one using a magnifying glass and count and calculate the appearance by classifying them into appearance types such as perfect spheres, cracked spheres, and crushed spheres. Since it takes time, a method for accurately evaluating the performance of the ion exchange resin in a shorter time is desired. Therefore, in the performance evaluation system of the ion exchange resin according to the present embodiment, the appearance index can be automatically calculated from the photographed image of the appearance of the ion exchange resin by using AI (Artificial Integrity). Hereinafter, the performance evaluation system for the ion exchange resin according to this embodiment will be described in detail.

[Overview of performance evaluation system]
FIG. 2 is a diagram showing an outline of a performance evaluation method for an ion exchange resin according to the present embodiment. In the present embodiment, a large number of sets (learning data sets) of a photographed image of the appearance of the ion exchange resin and the appearance evaluation result of the ion exchange resin are prepared as learning data, and a large number of ion exchanges for learning are prepared. Machine learning is performed based on the photographed image of the resin and the appearance evaluation result. The appearance evaluation result is a result of classifying the appearance type such as a perfect sphere, a cracked sphere, and a crushed sphere by visually observing the ion exchange resin with a microscope in the past by an analyst. For example, in an example comparing AI and an analyst (visually), which will be described later, the photographed data of about 2000 complete spheres, about 200 cracked spheres, and about 200 crushed spheres photographed in the past are used for learning. It was prepared as the data of the above, and machine learning was performed.

By inputting a photographed image of the appearance of the ion exchange resin for inspection into this machine-learned trained model (AI: Artificial Intelligence), the performance of the ion exchange resin reflected in the photographed image can be evaluated. can. Specifically, each of the ion exchange resins shown in the photographed image of the ion exchange resin for inspection is classified into one of a complete sphere, a cracked sphere, and a crushed sphere, and an appearance index is calculated.

[Configuration of performance evaluation system 1]
FIG. 3 is a system diagram showing an example of the configuration of the performance evaluation system 1 according to the present embodiment. The performance evaluation system 1 includes a microscope camera 10 and a performance evaluation device 30.

In the microscope camera 10, the camera 10b is attached to the microscope 10a. The camera 10b is, for example, a digital camera. The camera 10b captures an optical image magnified by the microscope 10a, converts it into electronic data, and transmits the converted electronic data (photographed image) to the performance evaluation device 30. For example, the microscope camera 10 and the performance evaluation device 30 are connected by USB (Universal Serial Bus). The microscope camera 10 and the performance evaluation device 30 are not limited to USB, and may be connected by other wired or wireless connection methods. The microscope camera 10 may have a structure in which the microscope 10a and the camera 10b are integrated (a structure that cannot be removed), or may be an electron microscope.

The microscope camera 10 photographs a sample of the ion exchange resin for inspection contained in the petri dish 20, and transmits the captured image to the performance evaluation device 30. At this time, the operator photographs different parts of the sample while shifting the dish 20 in order to photograph a large number (for example, 300 or more) of ion exchange resins with respect to one sample contained in the dish 20. Take 10 images as in. If the ion exchange resins next to each other overlap with each other, the operator may determine that they are scratches (for example, crack spheres), so the operator selects and shoots a portion with less overlap. In the following, the photographed image of the ion exchange resin for inspection will be referred to as an “inspection image”.

The performance evaluation device 30 is a so-called desktop computer used by connecting a monitor 30a and a keyboard 30b as an external device (peripheral device). One or both of the monitor 30a and the keyboard 30b may be built in the performance evaluation device 30. For example, the performance evaluation device 30 is not limited to a desktop computer, but may be a tablet computer, a notebook computer, or the like.

The performance evaluation device 30 acquires and stores an inspection image transmitted from the microscope camera 10. Then, the performance evaluation device 30 inspects from the inspection image using the trained model machine-learned based on the photographed image of the ion exchange resin photographed for learning and the evaluation result of the appearance of the ion exchange resin. Evaluate the performance of ion exchange resins for use. In the following, the photographed image of the ion exchange resin for learning will be referred to as a “learning image”. This trained model is machine-learned based on learning images of multiple ion exchange resins taken for each type of appearance of the ion exchange resin (eg, complete spheres, cracked spheres, crushed spheres). .. For example, the performance evaluation device 30 evaluates the performance of the ion exchange resin for inspection, and calculates the appearance index from the inspection image using the trained model.

The operator analyzes and evaluates the performance of the ion exchange resin shown in the inspection image by operating the keyboard 30b while looking at the GUI screen displayed on the monitor 30a of the performance evaluation device 30.

FIG. 4 is a diagram showing an example of a GUI screen according to the present embodiment. The illustrated GUI screen G10 is an image displayed when an inspection image captured in the performance evaluation device 30 is analyzed using AI. The screen area 101 displays a sample option of the ion exchange resin for inspection. When the sample selection button 102 is operated, the inspection image (original image) of the selected sample is displayed in the screen area 103. There are 10 inspection images taken for one sample. By performing the operation of selecting each of No. 1 to No. 10 in the screen area 103, the inspection image displayed in the screen area 103 can be switched to the selected inspection image.

When the analysis start button 104 is operated, the inspection image displayed in the screen area 103 is analyzed using AI, and the analysis image is displayed in the screen area 105. This analysis image is an image showing the results of classifying each of the ion exchange resins shown in the inspection image into either a complete sphere, a cracked sphere, or a crushed sphere using a trained model. In the illustrated example, the sample 1 No. The inspection image of 1 is displayed in the screen area 103, and the analysis image obtained by analyzing the inspection image by AI is displayed in the screen area 105. Here, a solid sphere is shown with a solid line frame, a cracked sphere is shown with a broken line frame, and a crushed sphere is shown with a double line frame. In the analysis image, each type of appearance of the ion exchange resin may be displayed in a distinguishable manner, and the display mode thereof can be arbitrarily determined. For example, each type of appearance of the ion exchange resin may be distinguished by changing the color of the frame, or by changing the shape of the frame.

Further, in the screen area 106, as an analysis result, the number (43 pieces, 1 piece, 2 pieces) and the ratio (93.5%, 2.2%, 4. 3%), the total number of each (46), the value of the appearance index (95.7%), etc. are displayed. The value of the appearance index is a value calculated by the following equation 2. The number of evaluations (total number of pieces) is not particularly limited.

Appearance index (%) = (total-total number of crushed balls) / total x 100 ... (Equation 2)
Here, the "total" is the total number of perfect spheres, cracked spheres, and crushed spheres. Further, this formula 2 is just a generalization of "300" to "total" with respect to the above-mentioned formula 1, and the basic calculation method is the same.

[Camera 10b configuration]
Next, the configuration of the camera 10b according to the present embodiment will be described in detail.
FIG. 5 is a block diagram showing an example of the configuration of the camera 10b according to the present embodiment. In this figure, the same reference numerals are given to the configurations corresponding to the respective configurations of FIG. The camera 10b can be attached to the microscope 10a via an adapter. The camera 10b includes, for example, a communication unit 11, an image pickup unit 12, a storage unit 13, and a control unit 15.

The communication unit 11 is configured to include a digital input / output port such as USB (Universal Serial Bus). For example, the communication unit 11 communicates with the performance evaluation device 30 and transmits the captured image captured by the camera 10b to the performance evaluation device 30. The communication unit 11 may transmit the captured image to the performance evaluation device 30 by a method other than USB. For example, the communication unit 11 is configured to include or in addition to USB, a video output terminal such as HDMI (registered trademark), and a communication device corresponding to a communication standard such as wireless LAN, wired LAN, and Bluetooth (registered trademark). May be done.

The image pickup unit 12 includes an image pickup element and an optical lens provided in front of the image pickup surface of the image pickup element. The image pickup unit 12 captures an optical image magnified by the microscope 10a obtained through the optical lens under the control of the control unit 15. Further, the image pickup unit 12 performs image processing on the captured image under the control of the control unit 15 and stores it in the storage unit 13 as a captured image. Here, the captured image is, for example, an inspection image.

The storage unit 13 includes, for example, an HDD (Hard Disk Drive), an SSD (Solid State Drive), an EEPROM (Electrically Erasable Programle Memory), a ROM (Read-Only Memory), a ROM (Read-Only Memory), a ROM (Read-One Memory), a ROM (Read-One Memory), a ROM (Read-One Memory), a ROM (Read-One Memory), and a ROM (Read-One Memory). The camera 10b stores various information, images, programs, and the like used for processing.

The operation unit 14 receives the user's operation on the camera 10b. For example, the operation unit 14 is an operation button such as a shutter button. The operation unit 14 may include an operation button, an operation dial, an operation switch, and the like for receiving an operation for setting a shooting mode and shooting conditions. Further, by accepting a part or all of the user's operation on the camera 10b by the performance evaluation device 30 communicated with the camera 10b, the performance evaluation device 30 controls the shooting of the camera 10b or the like according to the user's operation. You may go.

The control unit 15 is configured to include a CPU (Central Processing Unit) and the like, executes various programs stored in the storage unit 13, and controls each unit of the camera 10b. For example, the control unit 15 gives an image pickup instruction to the image pickup unit 12 based on the user's operation on the operation unit 14. Further, the control unit 15 stores the captured image based on the image captured by the image pickup unit 12 in the storage unit 13 according to the image pickup instruction. Further, the control unit 15 transmits a captured image based on the image captured by the image pickup unit 12 to the performance evaluation device 30 via the communication unit 11. Here, the captured image is, for example, an inspection image as described above. Further, the captured image is stored (saved) in the storage unit 13 as an image file associated with a file name and shooting date / time information (time stamp) determined according to a preset rule, and is transmitted to the performance evaluation device 30. To.

[Configuration of performance evaluation device 30]
Next, the configuration of the performance evaluation device 30 according to the present embodiment will be described in detail.
FIG. 6 is a block diagram showing an example of the configuration of the performance evaluation device 30 according to the present embodiment. In this figure, the same reference numerals are given to the configurations corresponding to the respective configurations of FIG. The illustrated performance evaluation device 30 includes a communication unit 31, a video output unit 32,

USB connectors

33 and 34, a storage unit 35, and a control unit 36.

The communication unit 31 is configured to include, for example, a plurality of Ethernet (registered trademark) ports and wireless communication ports such as Wi-Fi (registered trademark) and mobile phone lines, and communicates based on the control by the control unit 36. Communicate (send or receive) with an external device via a network (Internet, etc.).

The video output unit 32 includes an external monitor output terminal for outputting a video signal to an external display device (monitor, projector, etc.). The external monitor output terminal is an HDMI (registered trademark) terminal, a DVI terminal, a D-SUB terminal, a DisplayPort terminal, or the like. For example, the video output unit 32 connects to the monitor 30a and outputs a video signal to the monitor 30a.

The monitor 30a is a display device having a display for displaying information such as images and texts. For example, the monitor 30a includes a liquid crystal display panel, an organic EL (Electroluminescence) display panel, and the like.

The

USB connectors

33 and 34 are connection terminals for connecting to an external device compatible with USB. For example, the USB connector 33 is connected to the keyboard 30b and acquires an output signal in response to an operation on the keyboard 30b. Further, the USB connector 34 is connected to the camera 10b of the microscope camera 10 and acquires an inspection image (image file) transmitted from the camera 10b.

The storage unit 35 includes, for example, an HDD, SSD, EEPROM, ROM, RAM, etc., and stores various information, images, programs, etc. used for processing by the performance evaluation device 30. The storage unit 35 is not limited to the one built in the performance evaluation device 30, and may be an external storage device connected by a digital input / output port such as USB. For example, the storage unit 35 includes an inspection image storage unit 351, a learning model storage unit 352, and an evaluation data storage unit 353.

The inspection image storage unit 351 stores an inspection image (image file) transmitted from the camera 10b. For example, in the inspection image storage unit 351, a sample number (for example,

samples

1, 2, ...) And an imaging number (for example, No. 1 to No. 1) are stored in the image file of the inspection image transmitted from the camera 10b. .10) etc. are associated and stored. The sample number and the photographing number may be associated in the performance evaluation device 30 based on the user's operation, or may be automatically associated in the performance evaluation device 30 according to a predetermined rule.

The learning model storage unit 352 stores a trained model for evaluating the ion exchange resin shown in the inspection image. This trained model is machine-learned based on multiple ion exchange resin training images taken for each type of appearance of the ion exchange resin (eg, complete sphere, cracked sphere, crushed sphere) as described above. It is a thing. By inputting an inspection image into the trained model, each of the ion exchange resins shown in the inspection image is classified into a complete sphere, a cracked sphere, or a crushed sphere. The details of the configuration and processing for generating a trained model by machine learning will be described later.

The evaluation data storage unit 353 stores the evaluation result of evaluating the inspection image using the above-mentioned trained model. The evaluation results include the number and ratio of each of the complete spheres, cracked spheres, and crushed spheres contained in the ion exchange resin contained in the inspection image, the total number of each number, and the calculated appearance index. This evaluation result is stored in association with the file name of the inspection image.

The control unit 36 is configured to include a CPU and the like, executes various programs stored in the storage unit 35, and controls each unit of the performance evaluation device 30. For example, the control unit 36 includes an inspection image acquisition unit 361, an evaluation unit 362, and an output control unit 363 as a functional configuration realized by executing various programs stored in the storage unit 35.

The inspection image acquisition unit 361 acquires the inspection image (image file) transmitted from the camera 10b via the communication unit 31 and stores it in the inspection image storage unit 351. For example, the inspection image acquisition unit 361 inspects the image file of the inspection image in association with the sample number (for example,

samples

1, 2, ...) And the photographing number (for example, No. 1 to No. 10). It is stored in the image storage unit 351. As described above, the inspection image acquisition unit 361 may associate the sample number and the photographing number based on the user's operation (for example, the operation on the keyboard 30b), or may automatically associate the sample number and the photographing number according to a predetermined rule. good.

The evaluation unit 362 uses a trained model machine-learned based on a learning image of the ion exchange resin taken for learning and an evaluation result of the appearance of the ion exchange resin to obtain an ion exchange resin for inspection. The performance of the ion exchange resin for inspection is evaluated from the photographed inspection image. The appearance evaluation result is a result of evaluating whether the ion exchange resin is a complete sphere, a cracked sphere, or a crushed sphere classified based on the presence or absence of cracks and the presence or absence of crushing. For example, the trained model is machine-learned based on learning images of a plurality of ion exchange resins taken for each type of appearance of the ion exchange resin (for example, complete sphere, cracked sphere, crushed sphere). Is.

The evaluation unit 362 inputs an inspection image in which a plurality of ion exchange resins for inspection are taken into the trained model, and each of the ion exchange resins for inspection using the trained model has a type of appearance. Analyze whether it is classified as (complete sphere, crack sphere, crushed sphere). Then, the evaluation unit 362 calculates the appearance index based on the number of ion exchange resins for each classified type of appearance. Specifically, the evaluation unit 362 calculates the appearance index using the above-mentioned equation 2.

The output control unit 363 controls the display of the GUI screen G10 displayed on the monitor 30a when analyzing the performance of the ion exchange resin. For example, the output control unit 363 displays the GUI screen G10 on the monitor 30a, an inspection image selected according to the user's operation, an evaluation result by the evaluation unit 362 (an analysis image obtained by analyzing the inspection image, and an analysis). The result, etc.) is displayed on the monitor 30a. Further, the output control unit 363 stores the evaluation result (analysis result) by the evaluation unit 362 in the evaluation data storage unit 353.

Note that the performance (deterioration degree) of the ion exchange resin may be determined using the evaluation category shown in FIG. 7 based on the appearance index calculated by the evaluation unit 362. FIG. 7 is a diagram showing an example of a criterion for determining the performance (deterioration degree) of the ion exchange resin according to the present embodiment. In the example of the judgment criteria shown in the figure, the performance (deterioration degree) of the ion exchange resin is classified into five evaluation categories. When the appearance index is 95 to 100%, it is classified into "evaluation category 1" which is judged as "almost no crushed balls". When the appearance index is 80 to 94%, it is classified into "evaluation category 2" which is judged as "there is a small amount of crushed balls". When the appearance index is 60 to 79%, it is classified into "evaluation category 3" which is judged to have "crushed balls". When the appearance index is 40 to 59%, it is classified into "evaluation category 4" which is judged to have "many crushed balls". If the appearance index is less than 40%, it is classified into "evaluation category 5" which is judged to be "quite many crushed balls".

For example, the ion exchange resin may be exchanged when the evaluation category is 4 or 5, or the ion exchange resin may be exchanged when the evaluation category is 3 to 5. It can be arbitrarily determined in which evaluation category the ion exchange resin is exchanged. Further, the criteria for determining the performance (deterioration degree) of the ion exchange resin shown in FIG. 7 is an example, and the range of the appearance index of each evaluation category can be arbitrarily determined. Further, the number of evaluation categories is not limited to five, and may be classified into any number of evaluation categories.

[Operation of evaluation process of ion exchange resin]
Next, the operation of the ion exchange resin evaluation process in which the control unit 36 of the performance evaluation device 30 evaluates the ion exchange resin from the inspection image will be described.
FIG. 8 is a flowchart showing an example of the ion exchange resin evaluation process according to the present embodiment.

(Step S101) The operator uses the microscope camera 10 to take an image (inspection image) of the ion exchange resin for inspection. Specifically, the operator takes 10 images for one sample so that different parts of the sample are taken while shifting the petri dish 20. The microscope camera 10 transmits the captured inspection image to the performance evaluation device 30. Then, the process proceeds to step S103.

(Step S103) When the control unit 36 acquires an inspection image (image file) from the microscope camera 10, the control unit 36 saves the acquired inspection image (image file) in the inspection image storage unit 351. Then, the process proceeds to step S105.

(Step S105) The control unit 36 analyzes the type of appearance of the ion exchange resin shown in the inspection image by using the trained model stored in the learning model storage unit 352. Specifically, the control unit 36 reads the inspection image from the inspection image storage unit 351 in response to the operator's operation on the GUI screen G10 (see FIG. 4), and the learning is stored in the learning model storage unit 352. Enter in the finished model. The control unit 36 classifies the ion exchange resin for inspection according to the type of appearance (complete sphere, cracked sphere, crushed sphere) based on the analysis result output from the trained model, and outputs the number of each type of appearance. .. Then, the process proceeds to step S107.

(Step S107) The control unit 36 calculates the appearance index using the above-mentioned equation 2 based on the analysis result (the number of each appearance type) of the ion exchange resin for inspection. Then, the process proceeds to step S109.

(Step S109) The control unit 36 outputs an analysis result (evaluation result). For example, the control unit 36 uses the analysis image obtained by analyzing the inspection image and the analysis result (the number of perfect spheres, cracked spheres, and crushed spheres, calculated appearance index, etc.) as the analysis result (evaluation result) of the monitor 30a. (Display on GUI screen G10 (see FIG. 4)). Further, the control unit 36 stores the analysis result (evaluation result) in the evaluation data storage unit 353.

As a result, the performance evaluation system 1 can accurately evaluate the performance of the ion exchange resin in a short time by photographing the ion exchange resin for inspection and inputting it to the performance evaluation device 30.

[Configuration of machine learning device]
Next, the configuration of the machine learning device 50 that generates the trained model used for the ion exchange resin evaluation process in the performance evaluation system 1 will be described.
FIG. 9 is a block diagram showing an example of the configuration of the machine learning device 50 according to the present embodiment. The machine learning device 50 includes a communication unit 510, a learning data setting unit 520, a learning data storage unit 530, a learning unit 540, and an output unit 550. The configuration of the machine learning device 50 may be included in the performance evaluation device 30.

The communication unit 510 includes, for example, a plurality of Ethernet (registered trademark) ports, a plurality of digital input / output ports such as USB, and a wireless communication port such as Wi-Fi (registered trademark) and a mobile phone line, and communicates with the communication unit 510. Communicate with other devices and terminals via the network.

The training data setting unit 520 acquires the information necessary for generating the trained model. For example, the learning data setting unit 520 may acquire a learning image for each ion exchange resin for each type of appearance taken in the past from the camera 10b, a performance evaluation device 30, another device, or the like. Alternatively, it may be acquired via a storage medium such as an optical disk or a memory card. The learning data setting unit 520 associates the acquired learning image with the evaluation result (type of appearance) of the appearance of the ion exchange resin reflected in the learning image as a learning data set and stores it in the learning data storage unit 530. .. Hereinafter, the evaluation result (type of appearance) of the appearance is referred to as an “evaluation value”.

The learning unit 540 performs machine learning using a learning data set in which a learning image of an ion exchange resin and an evaluation value are associated with each other. Specifically, the learning unit 540 reads the learning data set from the learning data storage unit 530. The learning unit 540 performs machine learning using the read learning data set, and generates a trained model. The output unit 550 transmits the trained model to the performance evaluation device 30 via the communication unit 510. As a result, the trained model is stored and can be used in the learning model storage unit 352 of the performance evaluation device 30. The trained model may be input to the performance evaluation device 30 via a storage medium such as an optical disk or a memory card instead of the communication unit 510. Further, the trained model stored in the performance evaluation device 30 may be updated at any time as a new learning data set is input by the machine learning device 50 and machine learning progresses.

[Details of ion exchange resin evaluation process]
Hereinafter, the ion exchange resin evaluation process using the trained model will be described in detail.
The learning unit 540 uses the pixel value of the learning image as an input variable for inputting the pixel value of the learning image to the CNN (convolutional neural network) for learning, and the evaluation value of the learning image is an output variable output from the output layer. Set as. The learning unit 540 performs machine learning using a learning image and a learning data set of evaluation values.
Further, the evaluation unit 362 of the performance evaluation device 30 inputs the pixel value of the image of the ion exchange resin for inspection (inspection image) to the input layer for the trained CNN, and acquires the evaluation value from the output layer. do.

(About CNN)
FIG. 10 is an explanatory diagram illustrating an execution procedure in machine learning according to the present embodiment.
In this figure, the CNN is composed of I + 1 layers L0 to LI. The layer L0 is also called an input layer, the layers L1 to L (I-1) are also called an intermediate layer or a hidden layer, and the layer LI is also called an output layer. I is determined by the structure of the CNN, for example I = 3 or 4.

For CNN, an input image is input to the input layer L0. The input image is represented by a pixel matrix D11 whose matrix positions are the vertical position and the horizontal position of the input image. For each element of the pixel matrix D11, an R (red) sub-pixel value, a G (green) sub-pixel value, and a B (blue) sub-pixel value are input as pixel sub-pixel values corresponding to the positions of the matrix. There is.
The first intermediate layer L1 is a layer on which a convolution process (also referred to as a filter process) and a pooling process are performed.

(Convolution processing)
An example of the convolution process of the intermediate layer L1 will be described. The convolution process is a process of filtering the original image and outputting a feature map.
Specifically, the input pixel value is divided into a sub-pixel matrix D121 of R, a sub-pixel matrix D122 of B, and a sub-pixel matrix D123 of G, respectively. Each sub-pixel matrix D121, D122, D123 (each is also referred to as "sub-pixel matrix D12") is a convolution matrix of each element of the submatrix and s-row-t-column for each sub-matrix of s-row-t-column. The first pixel value is calculated by multiplying and adding the elements of CM1 (also called the kernel). The first pixel value calculated by each sub-pixel matrix D12 is multiplied by a weighting coefficient and added to calculate the second pixel value. The second pixel value is set in each element of the convoluted image matrix D131 as a matrix element corresponding to the position of the submatrix. By shifting the position of the submatrix for each element (sub-pixel) in each sub-pixel matrix D12, the second pixel value at each position is calculated, and all the matrix elements of the convoluted image matrix D131 are calculated.

For example, FIG. 10 is an example of the case of the convolution matrix CM1 having 3 rows and 3 columns, and the convolution pixel value D1311 is the 2nd to 4th rows and the 2nd to 4th columns of each sub pixel matrix D12. The first pixel value is calculated for the submatrix of 3 rows and 3 columns up to the eye. By calculating and adding a weighting coefficient to the first pixel value of each sub-pixel matrix D121, D122, D123, the second pixel value is used as the matrix element of the second row and second column of the convoluted image matrix D131. It is calculated. Similarly, the second pixel value of the matrix element in the third row and the second column of the convoluted image matrix D131 is calculated from the submatrix in the third to fifth rows and the second to fourth columns.
Similarly, the convolution image matrix D132, ... Is calculated using another weighting factor or another convolution matrix.

(Pooling process)
An example of the pooling process of the intermediate layer L1 will be described. The pooling process is a process of reducing an image while retaining the characteristics of the image.
Specifically, in the convoluted image matrix D131, the representative value of the matrix element in the region is calculated for each region PM of u rows and v columns. The representative value is, for example, the maximum value. The representative value is set in each element of the CNN image matrix D141 as a matrix element corresponding to the position of the region. By shifting the region in the convolutional image matrix D131 for each region PM, the representative value at each position is calculated, and all the matrix elements of the convolutional image matrix D131 are calculated.

For example, FIG. 10 is an example of the case of the area PM of 2 rows and 2 columns, and is the area of 2 rows and 2 columns from the 3rd row to the 4th row and the 3rd column to the 4th column of the convolution image matrix D131. The maximum value of the second pixel values in the region is calculated as a representative value. This representative value is set in the matrix element of the second row and the second column of the CNN image matrix D141. Similarly, from the submatrix of the 5th to 6th rows and the 2nd to 4th columns, the representative value of the matrix element of the 3rd row and the 2nd column of the CNN image matrix D141 is calculated. Similarly, the CNN image matrix D142, ... Is calculated from the convolutional image matrix D132, ....

Vector x is generated by arranging each matrix element (N) of the CNN image matrix D141, D142, ... In a predetermined order. In FIG. 10, the element xn (n = 1, 2, 3, ... N) of the vector x is represented by N nodes.

Intermediate layer Li represents the intermediate layer of the i-th intermediate layer (i = 2 to I-1). From the node of the third intermediate layer, the vector z (i) is output as the value of the vector u (i) input to the function f (u (i)). The vector u (i) is obtained by multiplying the vector z (i-1) output from the node of the intermediate layer of the i-1st th layer by the weight matrix W (i) from the left and adding the vector b (i). It is a vector. The function f (u (i)) is an activation function, and the vector b (i) is a bias. Further, the vector u (0) is a vector x.

The node of the output layer L4 is z (I-1), and its output is M ym (m = 1, 2, ... M). That is, a vector y (= (y1, y2, y3, ... yM)) having ym as an element is output from the output layer LI of the CNN.
As described above, the CNN outputs the vector y as an output variable when the pixel value of the input image is input as the input variable. The vector y represents an evaluation value.

FIG. 11 is an explanatory diagram illustrating a learning procedure in machine learning according to the present embodiment.
This figure is an explanatory diagram when the CNN of FIG. 10 performs machine learning.
Let the vector x output from the first intermediate layer be the vector X with respect to the pixel value of the image of the training data set. Let the vector representing the definite class of the training data set be the vector Y.

Initial values are set in the weight matrix W (i). When the input image is input to the input layer and, as a result, the vector X is input to the second intermediate layer, the vector y (X) corresponding to the vector X is output from the output layer. The error E between the vector y (X) and the vector Y is calculated using the loss function. The gradient ΔEi of the i-th layer is calculated using the output zi from each layer and the error signal δi. The error signal δi is calculated using the error signal δi-1. In addition, transmitting the error signal from the output layer side to the input layer side in this way is also called back propagation.
The weight matrix W (i) is updated based on the gradient ΔEi. Similarly, in the first intermediate layer, the convolution matrix CM or the weighting coefficient is updated.

(Trained model settings)
The learning unit 540 determines the number of layers, the number of nodes in each layer, the connection method of the nodes in each layer, the activation function, the error function, and the gradient descent algorithm, the pooling area, the kernel, the weight coefficient, and the weight matrix for the CNN. Set.
The learning unit 540 sets, for example, three layers (I = 3) as the number of layers. The learning unit 540 sets the number of nodes in each layer (also referred to as “the number of nodes”) to 800 for the number of elements (number of nodes N) of the vector x and 500 for the number of nodes in the second intermediate layer (i = 2). Set 10 in the output layer (i = 3). However, the present invention is not limited to this, and the total number may be four or more layers, or another value may be set for the number of nodes.

The learning unit 540 sets 20 convolution matrix CMs having 5 rows and 5 columns and a region PM having 2 rows and 2 columns. However, the present invention is not limited to this, and a different number of matrices or a different number of convolution matrix CMs may be set. Further, a region PM having a different number of matrices may be set.
The learning unit 540 may perform more convolution processing or pooling processing.

The learning unit 540 sets a full connection as a connection of each layer of the neural network. However, the present invention is not limited to this, and the bonding of some or all layers may be set to non-total bonding. The learning unit 540 sets a sigmoid function in the activation function of all layers as the activation function. However, the present invention is not limited to this, and the activation function of each layer is a step function, a linear connection, a soft sign, a soft plus, a ramp function, a truncated power function, a polymorphic value, an absolute value, a radial basis function, a wavelet, a maxout, etc. , May be another activation function. Also, the activation function of one layer may be of a different type from that of other layers.

The learning unit 540 sets a square loss (mean square error) as an error function. However, the present invention is not limited to this, and the error function may be cross entropy, τ-division loss, Huber loss, and ε sensitivity loss (ε tolerance function). Further, the learning unit 540 sets SGD (stochastic gradient descent) as an algorithm for calculating the gradient (gradient descent algorithm). However, the present invention is not limited to this, and Momentum (inertia term) SDG, AdaGrad, RMSprop, AdaDelta, Adam (Adaptive momentation) and the like may be used for the gradient descent algorithm.

The learning unit 540 is not limited to the convolutional neural network (CNN), and may set other neural networks such as a perceptron neural network, a recurrent neural network (RNN), and a residual network (ResNet). In addition, the learning unit 540 partially or completely includes trained models of supervised learning such as decision tree, regression tree, random forest, gradient boosting tree, linear regression, logistic regression, or SVM (support vector machine). It may be set.

(Varied example of learning)
In the above embodiment, the machine learning may be supervised learning other than the neural network. For example, the learning unit 540 performs machine learning of supervised learning not only by a neural network but also by a decision tree, a regression tree, a random forest, a gradient boosting tree, a linear regression, a logistic regression, an SVM (support vector machine), or the like. You may.

Further, in the above embodiment, the machine learning may be machine learning using unsupervised learning. For example, the learning unit 540 may perform machine learning of unsupervised learning by performing regression or classification by inputting a large number of learning images of an ion exchange resin.

Further, in the above embodiment, the machine learning may be machine learning using reinforcement learning. For example, the learning unit 540 may perform reinforcement learning (Q learning) using the Q value as reinforcement learning, or may perform reinforcement learning using Sarsa or the Monte Carlo method.

[Verification of evaluation results by AI]
Next, the result of verifying the accuracy of the evaluation result by the ion exchange resin evaluation process using AI according to this embodiment will be described. In AI in this verification, Python was used as a programming language, Pytorch was used as a machine learning library, and deep learning was performed by the machine learning device 50. As the image processing library, OpenCV (Open Source Computer Vision Library), which is a library for general computer vision, was used. As the learning images used for machine learning, external photographs of ion exchange resins taken in the past by analysts (complete spheres: about 2000, cracked spheres: about 200, crushed spheres: about 200) were used. ..

In addition, as the inspection image for verification, the analyst intends an individual whose evaluation category (see FIG. 7) is difficult to divide into, a very average individual, and an individual having a different appearance index. Twelve samples of ion exchange resins were prepared, and 10 photographs were taken with a microscope camera 10 so that the total number of ion exchange resins (total count number) was 300 or more for one sample. .. Individuals that are difficult to categorize are, for example, individuals whose appearance is wrinkled, whether or not an individual with few cracks is judged to be a crack, and whether a half-moon sphere looks round but is a complete sphere. It is an individual that is easy for beginners to make a mistake.

The performance evaluation device 30 analyzed the inspection image for verification taken by the microscope camera 10 using the trained model subjected to the above deep learning, and calculated the appearance index using the above equation 2. Then, the analysis result by AI was compared with the analysis result visually by the analyst, and the accuracy of the evaluation result of the ion exchange resin using AI was verified. The analysis result visually by the analyst is not an inspection image for verification, but a complete sphere, a cracked sphere, or a crushed sphere by visually confirming 300 optical images of the ion exchange resin magnified by a microscope. The appearance index was calculated using the above-mentioned formula 1.

FIG. 12 is a diagram showing the analysis result by AI and the analysis result visually by the analyst. In this figure, for each of Samples 1 to 12, the analysis results by the analyst's visual inspection and the analysis results by AI are shown side by side. As the analysis result, the total number of ion exchange resins (number of total resins), the appearance index, and the ratio of cracked spheres are described. The appearance indexes of Samples 2 to 6 and 8 to 10 have the same values as the analysis results by the analyst's visual inspection and the analysis results by AI. Further, the appearance indexes of

Samples

1, 7, 11 and 12 are judged to be the same evaluation category in the evaluation categories shown in FIG. 7, although there is a difference in the values between the analysis result by the analyst's visual inspection and the analysis result by AI. Since it is within the range that is satisfied, it can be judged that the difference is at a level that does not cause any problem. That is, it can be said that there is no problem in the accuracy of the evaluation result by the ion exchange resin evaluation process using AI.

In addition, samples 7 and 12 have more crack spheres than other samples. Rhagades are not included in the calculation of the appearance index because even if there are cracks, they do not cause any defects in the performance and equipment of the resin (problems of pressure loss, etc.). However, the cracked sphere is at all stages of crushing, and it can be said that it is a sign that the appearance index will deteriorate in the future. Therefore, in the illustrated analysis result, when the crack sphere exceeds 3%, it is described as a warning.

FIG. 13 is a diagram summarizing the comparison between the analysis result by AI and the analysis result visually by the analyst. The accuracy of the analysis result visually by the analyst is naturally good, but it is a condition that the analyst is an expert. In addition, since humans make judgments, there are some variations even if they are experts. Further, in the visual analysis, the calculation time of the appearance index per sample took about 13 minutes even for a skilled person and about 25 minutes for a beginner. On the other hand, the accuracy of the analysis result by AI is similarly good regardless of who operates it because it is analyzed by AI, and there is no variation. Further, in the analysis by AI, the calculation time of the appearance index per sample is about 6 minutes, which is only half the time of the visual analysis result. Of these 6 minutes, the analysis time by AI takes only about 1 minute, and the remaining time is almost the time required for shooting (time for shooting 10 images).

As described above, in the performance evaluation system 1 according to the present embodiment, the evaluation unit 362 is based on the learning image of the ion exchange resin taken for learning and the evaluation result of the appearance of the ion exchange resin. Using a machine-learned trained model, the performance of the inspection ion exchange resin is evaluated from the inspection image taken by the inspection ion exchange resin.

As a result, the performance evaluation system 1 can accurately evaluate the performance (deterioration degree) of the ion exchange resin from the photographed image of the ion exchange resin. Further, since the performance evaluation system 1 evaluates the performance of the ion exchange resin by AI, it is possible to obtain good evaluation results without variation regardless of who operates it without the need for an expert in analysis.

For example, as an evaluation result of the appearance, the classification according to the type of the appearance of the ion exchange resin is preset. Then, the evaluation unit 362 evaluates the performance of the ion exchange resin for inspection by using the trained model machine-learned based on the learning images of the plurality of ion exchange resins taken for each type of appearance.

Thereby, the performance evaluation system 1 can accurately evaluate the performance (deterioration degree) of the ion exchange resin based on the appearance of the ion exchange resin in a short time.

Here, the types of appearance are classified based on the presence or absence of cracks and the presence or absence of crushing in the appearance of the ion exchange resin (for example, classified into complete spheres, cracked spheres, and crushed spheres).

Thereby, the performance evaluation system 1 can accurately evaluate the performance (deterioration degree) of the ion exchange resin in a short time based on the appearance state such as the presence / absence of cracks and the presence / absence of crushing in the appearance of the ion exchange resin.

Specifically, the evaluation unit 362 inputs an inspection image in which a plurality of ion exchange resins for inspection are taken into the trained model, so that each of the ion exchange resins for inspection using the trained model is eventually used. Analyze whether it is classified into the type of appearance of. Then, the evaluation unit 362 determines the appearance of the inspection ion exchange resin included in the inspection image based on the number of ion exchange resins for each classified type of appearance (for example, perfect sphere, cracked sphere, crushed sphere). The appearance index indicating the state of is calculated.

As a result, the performance evaluation system 1 calculates the appearance index based on the number of ion exchange resins for each type of appearance classified by AI (for example, perfect sphere, cracked sphere, crushed sphere), so that the analyst can use a microscope. The performance (deterioration degree) of the ion exchange resin can be judged based on the same judgment criteria as when visually observing and analyzing the ion exchange resin.

The inspection image is an image taken by magnifying the plurality of inspection ion exchange resins placed in a petri dish (an example of a container) with a microscope camera 10. When this inspection image is photographed by the microscope camera 10, a portion where there is little overlap between the ion exchange resins to be imaged is selected and photographed.

As a result, the performance evaluation system 1 can accurately classify the types of appearance of the ion exchange resin for inspection.

Further, in the performance evaluation system 1 according to the present embodiment, the learning unit 540 performs machine learning based on the learning image of the ion exchange resin taken for learning and the evaluation result of the appearance of the ion exchange resin.

As a result, the performance evaluation system 1 can accurately evaluate the performance (deterioration degree) of the ion exchange resin from the photographed image of the ion exchange resin using AI.

Further, as the learning image, the appearance of the ion exchange resin for learning may be a plurality of images taken a plurality of times by changing at least one of the size, the angle and the color tone, or an image obtained by taking the appearance of the ion exchange resin. , Includes multiple images generated with at least one of different sizes, angles and tones. That is, for example, when the appearance of the ion exchange resin is photographed, the image for learning is the size to be photographed, the angle to be photographed (for example, the angle in the pan, tilt, or roll direction), the color tone correction at the time of photographing, and the like. It was taken multiple times with at least one of the changes. Further, as another example, the learning image is an image obtained by photographing the appearance of the ion exchange resin, and the size or angle (for example, the angle in the rotation direction) of the ion exchange resin shown in the image is changed, or the image is inverted. , A plurality of images are generated by performing at least one image processing such as color tone correction of an image and partial extraction (cutting) of an image. As described above, the learning image may be an image taken a plurality of times by changing the shooting conditions at the time of shooting, or a plurality of shot images are generated by changing the conditions later by image processing. It may be one or a mixture of them.

As a result, the performance evaluation system 1 has improved robustness and can accurately classify the types of appearances of the ion exchange resin for inspection shown in the inspection image.

As described above, if the inspection ion exchange resins contained in the petri dish 20 overlap each other when the inspection image is taken, the overlapped portion is erroneously determined to be a scratch (for example, a crack ball). There is a possibility that it will end up. For example, when the ion exchange resin is suspended in the aqueous layer, the floating ion exchange resin and the settled ion exchange resin may overlap each other and be photographed.

When air, oil, or hydrophobic solid adheres to the ion exchange resin, the ion exchange resin floats in water or receives the repulsion of water and sticks to the adjacent ion exchange resin, resulting in the ion exchange resin. It is thought that they will overlap each other. For example, when an organic substance is trapped in a functional group of an ion exchange resin, adsorption or aggregation of the ion exchange resins occurs, and the resins tend to overlap each other. In particular, the anionic ion exchange resin tends to capture organic substances having a carboxyl group, which are abundant in nature, and its influence becomes large.

Surfactants have the effect of solubilizing water, oil, or hydrophobic solids by adsorbing them on the interface. Therefore, by using a surfactant, the floating ion exchange resin is likely to settle in water, and the repulsion from water is also reduced, so that the overlap between the ion exchange resins can be reduced. As a result, erroneous judgment at the overlapping portion of the ion exchange resins can be reduced, and the accuracy of the performance evaluation of the ion exchange resins can be improved. Among the surfactants, anionic surfactants are preferable from the viewpoint of improving the accuracy of performance evaluation of ion exchange resins.

The experimental results confirming the effect of adding the surfactant will be described below.
In this experiment, an anionic surfactant was used. As the anionic surfactant, polyoxyethylene (3) sodium lauryl ether sulfate and polyoxyethylene (2) sodium lauryl ether sulfate, which are alkyl ether sulfate ester salts (AES), were used for confirmation. It is also effective to use a commercially available synthetic detergent containing an anionic surfactant.

(experimental method)
1. 1. Add an appropriate amount of ion exchange resin to the petri dish 20.
2. 2. The ion exchange resin contained in the petri dish 20 is photographed by the microscope camera 10, and the photographed inspection image is analyzed by the performance evaluation system 1 to evaluate the appearance of the ion exchange resin.
3. 3. Anionic surfactant is added dropwise to the same petri dish 20 containing the ion exchange resin, and the mixture is well stirred.
4. Similar to the above "2.", the ion exchange resin contained in the petri dish 20 is photographed by the microscope camera 10, the photographed inspection image is analyzed by the performance evaluation system 1, and the appearance of the ion exchange resin is evaluated.
5. The evaluation results of the above "2." and the above "4." are compared.

(Evaluation results)
The appearance evaluation results before and after the addition of the anionic surfactant will be described with reference to FIGS. 14 to 17. 14 to 17 are analysis images showing the results of photographing the ion exchange resin contained in the petri dish 20 and classifying them into one of a complete sphere, a cracked sphere, and a crushed sphere, and are displayed in the screen area 105 of FIG. Corresponds to the analyzed image to be performed.

First, the evaluation result of the appearance before adding the anionic surfactant of "2." of the above experimental method will be described. In the evaluation results of the appearance before the addition of the anionic surfactant, erroneous judgments have occurred in some of the places where the ion exchange resins overlap each other.

FIG. 14 is a diagram showing an example in which a crack ball is erroneously determined before the addition of an anionic surfactant. The example shown in this figure shows an example of the analysis result of the inspection image before the addition of the anionic surfactant, and the ion exchange resin is determined to be a crack ball at the three locations indicated by the symbols R1, R2, and R3. There is. The ion exchange resin at the location indicated by the reference numeral R1 is correctly determined to be a crack sphere, but at the two locations indicated by the reference numerals R2 and R3, it is erroneously determined to be a crack sphere due to the overlap of the ion exchange resins.

FIG. 15 is a diagram showing an example in which a crushed ball is erroneously determined before the addition of an anionic surfactant. The example shown in this figure shows an example of the analysis result of the inspection image before the addition of the anionic surfactant, and the ion exchange resin is determined to be a crushed ball at the three locations indicated by the symbols R4, R5, and R6. There is. The ion exchange resin at the location indicated by the reference numeral R4 is correctly determined to be a crushed sphere and a cracked sphere, but at the two locations indicated by the reference numerals R5 and R6, it is erroneously determined to be a crushed sphere due to the overlap of the ion exchange resins.

FIG. 16 is a diagram showing an example in which it is not possible to determine the portion where the resins overlap before the addition of the anionic surfactant. The example shown in this figure shows an example of the analysis result of the inspection image before the addition of the anionic surfactant, and the range indicated by the reference numeral R7 (the range indicated by the solid circle) includes a plurality of ion exchange resins. There are some resins whose appearance cannot be judged due to the overlap.

Next, the evaluation result of the appearance after adding the anionic surfactant of the above "4." will be described.
FIG. 17 is a diagram showing an example of the analysis result of the inspection image after the addition of the anionic surfactant. In the example shown in this figure, the addition of the anionic surfactant reduces the overlapping portion of the ion exchange resins, so that the appearance can be accurately determined. In this example, the ion exchange resin is correctly determined to be a crack sphere at the three locations indicated by the reference numerals R8, R9, and R10. In addition, the ion exchange resin without cracks or crushing is correctly judged as a perfect sphere.

According to these evaluation results, in the analysis result of the appearance of the ion exchange resin before the addition of the anionic surfactant, it is erroneously determined that the portion is a cracked sphere or a crushed sphere because there is an overlapping portion between the resins. Or, in some cases, it was not judged at all. Therefore, the number of cracked spheres or crushed spheres calculated may increase, or the number of ion exchange resins may not be calculated accurately, and the appearance index may not be calculated accurately. On the other hand, after the addition of the anionic surfactant, the overlap between the resins is reduced, so the number of perfect spheres, cracked spheres, and crushed spheres is calculated accurately, so the appearance index is calculated accurately, and the ion exchange resin The accuracy of performance evaluation has improved.

As described above, before the inspection image is taken by the microscope camera 10 (an example of a magnifying mirror camera), a surfactant (for example, an example of a container) is placed in a chalet 20 (an example of a container) containing a plurality of inspection ion exchange resins. , Anionic surfactant) is added and stirred, and then the image is taken by the microscope camera 10, so that the accuracy of the performance evaluation of the ion exchange resin can be improved.

Although an anionic surfactant was used in the above experiment, a surfactant other than the anionic surfactant may be used. However, an anionic surfactant is preferable in terms of the effect of reducing the overlap between the ion exchange resins and the fact that it is easily available.

Although an example of adding an anionic surfactant has been described here, other surfactants may be used. However, it is preferable to use an anionic surfactant in terms of reducing the overlap between the ion exchange resins and in terms of availability.

Although one embodiment of the present invention has been described in detail with reference to the drawings, the specific configuration is not limited to the above, and various design changes and the like are made without departing from the gist of the present invention. It is possible to do.

Further, in the above embodiment, the number of perfect spheres and the number of crushed spheres in the sample are used when calculating the appearance index of the sample, and the ratio of the number of perfect spheres to the total number of perfect spheres and crushed spheres is used as the appearance index. Calculated, but not limited to this. For example, the ratio of the number of perfect spheres to the total number of perfect spheres, crushed spheres, and cracked spheres may be calculated as an appearance index.

In the above embodiment, the ion exchange resin is classified into three types of appearance, a complete sphere, a cracked sphere, and a crushed sphere, based on the appearance state of the ion exchange resin, but the present invention is not limited to the three types. The classification according to the type of appearance may be two types (for example, a complete sphere and a crushed sphere). For example, the types of appearance may be classified based on the presence or absence of crushing in the appearance of the ion exchange resin (for example, classified into complete spheres and crushed spheres). In this case, the ratio of the number of perfect spheres to the total number of perfect spheres and crushed spheres may be calculated as an appearance index. Further, the types of appearance may be classified based on the presence or absence of cracks in the appearance of the ion exchange resin (for example, classified into perfect spheres and cracked spheres). In this case, the ratio of the number of perfect spheres to the total number of perfect spheres and cracked spheres may be calculated as an appearance index. That is, the type of appearance may be classified based on at least one of the presence / absence of cracks and the presence / absence of crushing in the appearance of the ion exchange resin. Further, the classification according to the type of appearance may be four or more types. For example, when the gel-type ion exchange resin deteriorates, the appearance is rarely wrinkled. This wrinkled state may be machine-learned in addition to the type of appearance. Further, an appearance type indicating a state of appearance other than the above may be added.

Further, in the above embodiment, the camera 10b is not limited to a dedicated camera device such as a digital camera, but may be an electronic device having a camera function as a part of the function such as a smartphone. Further, the camera 10b and the performance evaluation device 30 do not have to be connected by communication, and the image (inspection image or the like) taken by the camera 10b is delivered to the performance evaluation device 30 via a storage medium such as an optical disk or a memory card. You may.

Further, the camera 10b and the performance evaluation device 30 may be integrally configured as one device. Further, the microscope camera 10 and the performance evaluation device 30 may be integrally configured as one device.

Further, in the above embodiment, the sample of the ion exchange resin put in the petri dish is photographed or observed with the microscope camera 10, but a container other than the petri dish may be used. However, it is desirable that the container has a shape in which the ion exchange resins are unlikely to overlap (for example, a shape having a wide planar region as much as possible).

Further, in the above embodiment, an example of performing an ion exchange resin evaluation process using a trained model generated as learning data of an ion exchange resin learning image and an appearance evaluation result (evaluation value) has been described. , Not limited to this. For example, the evaluation process of the ion exchange resin may be performed using a data table in which the learning image of the ion exchange resin and the evaluation result (evaluation value) of the appearance are associated with each other, or the image for learning the ion exchange resin may be used. The ion exchange resin may be evaluated using a program that embodies an algorithm regarding the relationship with the appearance evaluation result (evaluation value).

The camera 10b, the performance evaluation device 30, or the machine learning device 50 included in the performance evaluation system 1 described above has a computer system inside. Then, a program for realizing the functions of each configuration included in the camera 10b, the performance evaluation device 30, or the machine learning device 50 described above is recorded on a computer-readable recording medium, and the program recorded on the recording medium is recorded. The processing in each configuration included in the above-mentioned camera 10b, the performance evaluation device 30, or the machine learning device 50 may be performed by loading and executing it in a computer system. Here, "loading and executing a program recorded on a recording medium into a computer system" includes installing the program in the computer system. The term "computer system" as used herein includes hardware such as an OS and peripheral devices. Further, the "computer system" may include a plurality of computer devices connected via a network including a communication line such as the Internet, WAN, LAN, and a dedicated line. Further, the "computer-readable recording medium" refers to a portable medium such as a flexible disk, a magneto-optical disk, a ROM, or a CD-ROM, and a storage device such as a hard disk built in a computer system. As described above, the recording medium in which the program is stored may be a non-transient recording medium such as a CD-ROM.

The recording medium also includes an internal or external recording medium accessible from the distribution server for distributing the program. It should be noted that the program is divided into a plurality of parts, downloaded at different timings, and then combined with each configuration of the camera 10b, the performance evaluation device 30, or the machine learning device 50, or the divided programs are distributed. The servers may be different. Furthermore, a "computer-readable recording medium" is a volatile memory (RAM) inside a computer system that serves as a server or client when a program is transmitted via a network, and holds the program for a certain period of time. It shall include things. Further, the above program may be for realizing a part of the above-mentioned functions. Further, a so-called difference file (difference program) may be used, which can realize the above-mentioned function in combination with a program already recorded in the computer system.

Further, a part or all of the camera 10b, the performance evaluation device 30, or the machine learning device 50 may be realized as an integrated circuit such as an LSI (Large Scale Integration). Further, each component in the camera 10b, the performance evaluation device 30, or the machine learning device 50 of the present embodiment may be individually made into a processor, or a part or all of them may be integrated into a processor. Further, the method of making an integrated circuit is not limited to the LSI, and may be realized by a dedicated circuit or a general-purpose processor. Further, when an integrated circuit technology that replaces an LSI appears due to advances in semiconductor technology, an integrated circuit based on this technology may be used.

<Other Examples>
In the above embodiment, for example, an example in which a sample of an ion exchange resin for inspection sent from a facility equipped with an ion exchange resin tower is placed in a chalet 20 and photographed with a microscope camera 10 to perform the performance of the ion exchange resin has been described. However, the photographed image taken by enlarging the ion exchange resin in the ion exchange resin tower may be transmitted to the performance evaluation device 30. At this time, the performance evaluation device 30 may be in a facility provided with an ion exchange resin tower, or may be in a facility of a vendor who manufactures the ion exchange resin and provides it to the customer. In this embodiment, a system for transmitting an image taken by enlarging the ion exchange resin in the ion exchange resin tower to the performance evaluation device 30 will be described.

FIG. 18 is a system diagram showing an example of the configuration of the performance evaluation system 1A according to the present embodiment. The performance evaluation system 1A includes a performance evaluation device 30 and a water treatment facility 100. The basic configuration of the performance evaluation device 30 is the same as the configuration shown in FIG. The water treatment equipment 100 is used for various purposes such as washing water used in the manufacturing process of semiconductors, liquid crystals, wafers, precision parts, desalted water manufactured by a power plant condensate desalination device, water for manufacturing pharmaceuticals, and the like. This is a facility that manufactures ultrapure water. The water treatment equipment 100 includes an ion exchange resin tower 101, a communication unit 110, an image pickup unit 120, a storage unit 130, and a control unit 140.

The communication unit 110 is configured to include communication devices compatible with communication standards such as wireless LAN and mobile communication, and communicates (transmits or receives) with the performance evaluation device 30 and other devices via the communication network NW. .. The image pickup unit 120 includes an image pickup element and an optical lens provided in front of the image pickup surface of the image pickup element. A magnifying glass for magnifying and photographing the ion exchange resin is used for the optical lens. The imaging unit 120 photographs the ion exchange resin in the ion exchange resin tower 101 as a sample for inspection. For example, a transparent window is provided on a part of the bottom surface or the side surface of the ion exchange resin tower 101, and the image pickup unit 120 uses a large number of ion exchange resins existing in a place corresponding to the window as a sample for inspection. Take a picture. At this time, a plurality of captured images may be acquired while the angle of view of the imaging unit 120 is automatically or manually moved. The storage unit 130 includes, for example, an HDD, SSD, EEPROM, ROM, RAM, etc., and stores various information, images, programs, and the like.

The control unit 140 transmits a photographed image (inspection image) of the ion exchange resin for inspection in the ion exchange resin tower 101 photographed by the image pickup unit 120 to the performance evaluation device 30 via the communication unit 110. The communication unit 31 of the performance evaluation device 30 receives a photographed image (inspection image) of the ion exchange resin for inspection transmitted from the water treatment equipment 100. The evaluation unit 362 of the performance evaluation device 30 evaluates the performance of the ion exchange resin for inspection from the received inspection image. Specifically, as described above, for inspection using a trained model machine-learned based on the learning image of the ion exchange resin taken for learning and the evaluation result of the appearance of the ion exchange resin. Evaluate the performance of the ion exchange resin for inspection from the image.

Further, the performance evaluation device 30 may transmit information indicating the evaluation result and the analysis result of the inspection image evaluated by the evaluation unit 362 to the water treatment equipment 100. As a result, the control unit 140 acquires information indicating the evaluation result and the analysis result of the inspection image evaluated by the evaluation unit 362 of the performance evaluation device 30 from the performance evaluation device 30. For example, the control unit 140 displays information based on the evaluation results and analysis results acquired from the performance evaluation device 30 (for example, information such as the value of the appearance index, the evaluation category, and the necessity of replacement) in the water treatment equipment 100. It may be displayed in a unit (not shown), or may be transmitted to a terminal used by the equipment manager of the water treatment equipment 100 or the like.

Further, when the ion exchange resin needs to be replaced based on the evaluation result of the inspection image acquired from the performance evaluation device 30, the control unit 140 orders the replacement ion exchange resin from the provider of the ion exchange resin. You may go to the vendor. For example, the control unit 140 may generate order information such as the type, quantity, and delivery destination of the replacement ion exchange resin, and transmit it to the service department of the vendor via the communication unit 110. At this time, the control unit 140 may transmit the order information after confirming that the ordering department has been approved.

As described above, the performance evaluation system 1A can determine the performance (deterioration degree) of the ion exchange resin in the ion exchange resin tower 101 based on the inspection image taken by the ion exchange resin tower 101 of the water treatment facility 100. It is convenient because it is not necessary to send a sample for inspection and the time can be shortened.

The machine learning device 50 acquires an inspection image taken by the water treatment facility 100 and an evaluation result in which the performance evaluation device 30 evaluates the inspection image, and the acquired inspection image is used as a learning image. The trained model used by the evaluation unit 362 for evaluation based on the training image and the evaluation result may be further machine-learned.

As a result, the ion exchange resin used in the ion exchange resin tower 101 can be continuously evaluated and the evaluation results and the evaluation results can be updated at any time, so that the ion exchange resin can be evaluated more accurately. become.

1, 1A performance evaluation system, 10 microscope camera, 10a microscope, 10b camera, 11 communication unit, 12 imaging unit, 13 storage unit, 14 operation unit, 15 control unit, 20 chalet, 30 performance evaluation device, 30a monitor, 30b keyboard , 31 communication unit, 32 video output unit, 33, 34 USB connector, 35 storage unit, 36 control unit, 351 inspection image storage unit, 352 learning model storage unit, 353 evaluation data storage unit, 361 inspection image acquisition unit, 362 evaluation unit, 363 output control unit, 100 water treatment equipment, 101 ion exchange resin tower, 110 communication unit, 120 imaging unit, 130 storage unit, 140 control unit

Claims

For inspection, the ion exchange resin for inspection was photographed using a trained model machine-learned based on the learning image of the ion exchange resin taken for learning and the evaluation result of the appearance of the ion exchange resin. Evaluation unit that evaluates the performance of the ion exchange resin for inspection from the image,
Performance evaluation system equipped with.
As the evaluation result of the appearance, the classification according to the type of appearance of the ion exchange resin is preset.
The evaluation unit
The performance of the ion exchange resin for inspection is evaluated using a trained model machine-learned based on the learning images of a plurality of ion exchange resins taken for each type of appearance.
The performance evaluation system according to claim 1.
The types of appearance are classified based on at least one of the presence or absence of cracks and the presence or absence of crushing in the appearance of the ion exchange resin.
The evaluation unit
Using the performance of the ion exchange resin for inspection as an evaluation, an appearance index indicating the appearance state of the ion exchange resin is calculated.
The performance evaluation system according to claim 2.
The evaluation unit
By inputting the inspection image in which a plurality of the ion exchange resins for inspection are taken into the trained model, each of the ion exchange resins for inspection using the trained model has any of the appearance types. Based on the number of ion exchange resins classified for each type of appearance, an appearance index indicating the appearance state of the inspection ion exchange resin contained in the inspection image is obtained. calculate,
The performance evaluation system according to claim 3.
The inspection image is an image taken by magnifying a plurality of ion exchange resins for inspection contained in a container with a magnifying glass camera, and is taken when the image is taken with the magnifying glass camera. This is an image taken by selecting a part where there is little overlap between the target ion exchange resins.
The performance evaluation system according to any one of claims 1 to 4.
The inspection image is an image taken by the magnifying camera after adding a surfactant to a container containing a plurality of the ion exchange resins for inspection and stirring the image before being taken by the magnifying camera. Is,
The performance evaluation system according to any one of claims 1 to 5.
A learning unit that performs machine learning based on the learning image of the ion exchange resin taken for learning and the evaluation result of the appearance of the ion exchange resin.
The performance evaluation system according to any one of claims 1 to 6.
The size of the learning image is an image of the appearance of the learning ion exchange resin taken a plurality of times by changing at least one of the size, angle and color tone, or an image of the appearance of the ion exchange resin. Includes multiple images generated with at least one of different angles and tones.
The performance evaluation system according to any one of claims 1 to 7.
It is a performance evaluation method for ion exchange resins.
The step of acquiring an inspection image in which the ion exchange resin for inspection was taken, and
Using a trained model machine-learned based on the learning image of the ion exchange resin taken for learning and the evaluation result of the appearance of the ion exchange resin, the ion exchange resin for inspection is used from the inspection image. Steps to evaluate the performance of
Performance evaluation method with.
On the computer
The step of acquiring an inspection image in which the ion exchange resin for inspection was taken, and
Using a trained model machine-learned based on the learning image of the ion exchange resin taken for learning and the evaluation result of the appearance of the ion exchange resin, the ion exchange resin for inspection is used from the inspection image. Steps to evaluate the performance of
A program to execute.
It is a trained model for evaluating the performance of the ion exchange resin for inspection from the inspection image taken by the ion exchange resin for inspection.
Machine learning is performed based on the learning image of the ion exchange resin taken for learning and the evaluation result of the appearance of the ion exchange resin, and the performance of the ion exchange resin for inspection is evaluated from the inspection image.
A trained model for making your computer work.
It is a performance evaluation system equipped with water treatment equipment and performance evaluation equipment.
The water treatment facility
Ion exchange resin tower and
An image pickup unit that photographs the ion exchange resin in the ion exchange resin tower,
A transmission unit that transmits an image captured by the image pickup unit as an inspection image, and a transmission unit.
Equipped with
The performance evaluation device is
A receiving unit that receives the inspection image and
Using a trained model machine-learned based on the learning image of the ion exchange resin taken for learning and the evaluation result of the appearance of the ion exchange resin, the ion exchange resin for inspection is used from the inspection image. Evaluation department that evaluates the performance of
Performance evaluation system equipped with.
The performance evaluation device is
Information based on the evaluation result by the evaluation unit is transmitted to the water treatment facility.
The performance evaluation system according to claim 12.
The water treatment facility
When it is necessary to replace the ion exchange resin based on the information acquired from the performance evaluation device, the order information of the replacement ion exchange resin is generated.
The performance evaluation system according to claim 13.