CN112198230A

CN112198230A - Nondestructive inspection system

Info

Publication number: CN112198230A
Application number: CN202010346979.3A
Authority: CN
Inventors: 浅野英久; 二宫悠; 伊藤达志; 大西哲雄
Original assignee: Komatsu Ltd
Current assignee: Komatsu Ltd
Priority date: 2019-06-20
Filing date: 2020-04-27
Publication date: 2021-01-08
Also published as: JP2021001774A; JP7394545B2

Abstract

The present invention relates to a nondestructive inspection system, a method of manufacturing a learned exhaust treatment filter inspection model, and a method of generating data for learning. A nondestructive inspection system for an exhaust gas treatment filter includes: a nondestructive testing device (200) for performing nondestructive testing on the exhaust gas treatment filter; and a computer. The computer has a learned exhaust treatment filter inspection model for determining whether or not the exhaust treatment filter subjected to the non-destructive test can be used. A computer acquires non-destructive test result data (320) representing the results of a non-destructive test of an exhaust treatment filter of a non-destructive testing device (200), and outputs a result (120) of estimating the availability of the exhaust treatment filter from the non-destructive test result data (320) using a learned exhaust treatment filter inspection model.

Description

Nondestructive inspection system

Technical Field

The present disclosure relates to a nondestructive inspection system, a method of manufacturing a learned exhaust treatment filter inspection model, and a method of generating data for learning.

Background

In the past, U.S. Pat. No. 7849747 describes a technique for performing ultrasonic flaw detection on a ceramic monolith used for a diesel particulate filter.

A diesel engine is generally used as a power source of the work machine. Particulate Matter (PM) discharged from a Diesel engine is removed from exhaust gas by a Diesel Particulate Filter (DPF) in the same manner as in a normal automobile. The removed particulate matter is mainly coal generated during combustion, and is thus burned in the DPF.

In PM, since incombustibles are contained in addition to the coal as the main component, incombustibles accumulate in the DPF. For this reason, the DPF is replaced after a given time has elapsed from the start of use. In general, although a DPF is replaced with a new DPF in a vehicle, a DPF using a regenerated product subjected to a washing process after use is widely used in a working machine. The DPF after use is examined for the absence of defects such as cracks by nondestructive examination such as ultrasonic testing before the washing treatment.

The output data of the ultrasonic flaw detection test is represented by a graph of the elapsed time from the emission of the ultrasonic wave to the subject and the intensity of the reflected ultrasonic wave. The tester observes the graph to determine whether or not the filter is defective, and the frequency of erroneous determination is high because the filter is a human. Therefore, the determination criteria for the availability of recycling are made excessively strict to prevent recycling of defective products, and there is a high possibility that products that can be recycled are actually discarded.

Disclosure of Invention

The present disclosure provides a nondestructive inspection system for improving the accuracy of determining whether or not an exhaust treatment filter is usable, a method for manufacturing a learned exhaust treatment filter inspection model, and a method for generating data for learning.

According to an aspect of the present disclosure, a nondestructive inspection system for an exhaust treatment filter is provided. The nondestructive inspection system includes: a nondestructive testing device for performing nondestructive testing on the exhaust gas treatment filter; and a computer. The computer has a learned exhaust treatment filter inspection model for determining whether or not the exhaust treatment filter subjected to the non-destructive test can be used. The computer acquires test result data representing a result of a nondestructive test of the exhaust treatment filter of the nondestructive testing device, and outputs a result of estimating the availability of the exhaust treatment filter based on the test result data using the learned exhaust treatment filter inspection model.

According to an aspect of the present disclosure, a method of manufacturing a learned exhaust treatment filter inspection model is provided. The manufacturing method includes the following processes. In the 1 st process, data for learning including a combination of test result data representing a result of a nondestructive test of the exhaust gas treatment filter and usability determination result data obtained by determining whether or not the exhaust gas treatment filter subjected to the nondestructive test is usable is acquired. The 2 nd process generates an exhaust gas treatment filter inspection model that has test result data as input and outputs a value relating to whether or not the exhaust gas treatment filter is usable, based on the acquired plurality of data for learning.

According to an aspect of the present disclosure, there is provided a method of generating data for learning an exhaust gas treatment filter inspection model for determining whether or not an exhaust gas treatment filter to be inspected can be used. The generation method includes the following processing. In the 1 st process, test result data representing the results of the non-destructive test of the exhaust gas treatment filter is acquired. In the process 2, use availability determination result data obtained by determining whether or not the exhaust treatment filter subjected to the non-destructive test is available for use is acquired.

According to an aspect of the present disclosure, a method of manufacturing a learned exhaust treatment filter inspection model is provided. The manufacturing method includes the following processes. In the 1 st process, test result data representing the results of the non-destructive test of the exhaust gas treatment filter is acquired. In the 2 nd process, test result data is input to the 1 st learned exhaust treatment filter check model to obtain an output of a usability estimation result obtained by estimating usability of the exhaust treatment filter. In the 3 rd process, the 2 nd exhaust gas treatment filter inspection model is learned by learning data including test result data and estimation availability results.

The above and other objects, features, aspects and advantages of the present invention will become apparent from the following detailed description of the present invention, which is to be read in connection with the accompanying drawings.

Drawings

Fig. 1 is a schematic diagram showing a schematic configuration of an exhaust gas aftertreatment device.

Fig. 2 is a block diagram showing an example of the configuration of a computer.

Fig. 3 is a block diagram showing an example of the configuration of the learning computer.

Fig. 4 is a functional block diagram for explaining a functional configuration of the learning computer.

Fig. 5 is a flowchart showing a processing procedure of the learning processing in the learning computer.

Fig. 6 is a schematic diagram of a process for learning the exhaust gas treatment filter inspection model.

Fig. 7 is a flowchart showing processing steps for preprocessing data of test results obtained by an ultrasonic testing machine.

Fig. 8 is a graph showing an example of the point cloud data.

Fig. 9 is a graph showing an example of extracted data.

Fig. 10 is a graph showing an example of extraction data of a defective product.

Fig. 11 is a graph showing an example of extraction data of qualified products under the conditions.

Fig. 12 is a flowchart showing a processing procedure for preprocessing image data acquired by the imaging device.

Fig. 13 is a diagram showing an example of an entry image.

Fig. 14 is a diagram showing an example of an exit image.

Fig. 15 is a flowchart showing a processing procedure for preprocessing X-ray image data acquired by the X-ray imaging apparatus.

Fig. 16 is a diagram showing an example of transmission image data of non-defective products.

Fig. 17 is a diagram showing an example of transmission image data of a qualified product.

Fig. 18 is a diagram showing an example of cross-sectional image data of a defective product.

Fig. 19 is a functional block diagram for explaining a functional configuration of the determination computer.

Fig. 20 is a flowchart showing a processing procedure of an estimation process in the determination computer.

Fig. 21 is a schematic diagram showing a process for estimating whether or not the exhaust treatment filter is usable based on the exhaust treatment filter inspection model.

Fig. 22 is a flowchart showing a process for generating a distillation model.

Detailed Description

The embodiments are described below based on the drawings. In the following description, the same components are denoted by the same reference numerals. Their names and functions are also the same. And thus their detailed description will not be repeated.

[ Structure of exhaust gas aftertreatment device 1 ]

Fig. 1 is a schematic diagram showing a schematic configuration of an exhaust gas aftertreatment device 1. The exhaust gas post-treatment device 1 is a device for purifying exhaust gas by treating residual substances contained in the exhaust gas of a diesel engine (hereinafter, referred to as "engine") 10. The residual substances are, for example, Particulate Matter (PM), nitrogen oxides (NOx), and the like. The exhaust gas post-treatment device 1 performs treatment such as PM collection and NOx reduction to purify the exhaust gas.

The exhaust gas post-treatment device 1 includes a Diesel Particulate Filter (DPF) device 2 and a Selective Catalytic Reduction (SCR) device 3. The DPF device 2 and the SCR device 3 are connected to an exhaust pipe 11 of exhaust gas flowing through the engine 10. The DPF device 2 and the SCR device 3 are arranged in this order in the direction of exhaust gas flow indicated by the arrow in fig. 1. The SCR device 3 is disposed downstream of the DPF device 2 in the flow direction of the exhaust gas.

The DPF device 2 includes a Diesel Oxidation Catalyst (DOC) 21 and a Catalyzed Soot Filter (CSF) 22.

The DOC21 is a catalyst that oxidizes fuel charge (Dosing fuel) supplied to the exhaust gas as needed to generate heat, thereby raising the exhaust gas temperature to a predetermined high temperature range. The exhaust gas having this increased temperature burns PM deposited on CSF22 described later by itself and is incinerated and removed, thereby regenerating CSF 22.

The charge fuel is, for example, the same light oil as the engine fuel in the case where the internal combustion engine is a diesel engine. The fuel injection is supplied to the exhaust gas by, for example, a fuel injection device (not shown) for dosing provided in the exhaust pipe 11, and flows into the DPF device 2 together with the exhaust gas.

The CSF22 is a filter that traps PM in exhaust gas. CSF22 has: allowing exhaust gas to flow into inlet end 22A of CSF 22; and an outlet port 22B from which exhaust flows from CSF 22. The CSF22 corresponds to the exhaust gas treatment filter in the embodiment.

Although not shown in detail, CSF22 has a honeycomb structure with a large number of small pores. The pores of CSF22 extend from inlet end 22A to outlet end 22B. The cross section of the small hole of CSF22 is formed in a polygonal shape (for example, a quadrangular shape, a hexagonal shape, etc.).

The pores of CSF22 are arranged alternately: the inlet 22A is opened with a small hole closed at the outlet 22B, and the inlet 22A is closed with a small hole opened at the outlet 22B. From the small hole opening at the inlet end 22A, the exhaust gas flows into the CSF 22. The exhaust gas flows through the small holes that open to the outlet end 22B through the boundary wall separating the adjacent small holes, and flows out from the CSF22 at the outlet end 22B. The PM is then trapped at the boundary wall.

The material of CSF22 is, for example, ceramics such as cordierite and silicon carbide, and is appropriately determined depending on the application. An oxidation catalyst different from the DOC21 material is coated on the inlet side of the CSF22 by sealing (Wash coat).

The SCR device 3 includes a denitration catalyst 31 and an ammonia oxidation catalyst 32.

The denitration catalyst 31 is a catalyst for reducing and removing nitrogen oxides (NOx) in the exhaust gas by using ammonia obtained by decomposition of urea water injected from the reducing agent supply device into the exhaust pipe 11 as a reducing agent.

The ammonia oxidation catalyst 32 disposed downstream of the denitration catalyst 31 is a catalyst for detoxifying ammonia by oxidizing ammonia that is not used in the reduction reaction in the denitration catalyst 31, and further reduces harmful components in the exhaust gas.

[ Structure of computer 100 ]

Next, the configuration of the computer 100 included in the nondestructive inspection system according to the embodiment for performing nondestructive inspection on the exhaust gas treatment filter (CSF22) will be described. Fig. 2 is a block diagram showing an example of the configuration of the computer 100.

As shown in fig. 2, the computer 100 includes, as main hardware elements, a display 102, a processor 104, a memory 106, a network controller 108, a storage 110, an optical drive 122, and an input device 126. The input device 126 includes a keyboard 127 and a mouse 128. The input device 126 may be provided with a touch panel.

The display 102 displays information necessary for processing in the computer 100. The display 102 is formed of, for example, an lcd (liquid Crystal display), an organic EL (Electroluminescence) display, or the like.

The processor 104 is an arithmetic main body that executes various programs to execute processing necessary for realizing the computer 100. The processor 104 is constituted by, for example, 1 or more CPUs (Central Processing units), GPUs (Graphics Processing units), or the like. A CPU or GPU with multiple cores may also be used.

The memory 106 provides a storage area for temporarily storing program codes, a work memory, and the like when the processor 104 executes programs. As the Memory 106, a volatile Memory device such as a DRAM (Dynamic Random Access Memory) or an SRAM (Static Random Access Memory) can be used.

The network controller 108 transmits and receives data to and from any external device including a nondestructive testing device 200 and a learning computer 300, which will be described later. The Network controller 108 can support any communication system such as the internet (registered trademark), a Local Area Network (LAN), and Bluetooth (registered trademark).

The storage 110 stores an OS (Operating System) 112 executed in the processor 104, an application 114 for realizing a given functional configuration, a learned model 116, and the like. As the storage 110, for example, a nonvolatile memory device such as a hard disk or an SSD (Solid State Drive) can be used.

As a part of libraries, functional modules, which are required when the processor 104 executes the application program 114, libraries or functional modules provided in standards by the OS112 can be used. In this case, the application 114 alone does not include all the program modules necessary for realizing the corresponding functions, but can realize a predetermined functional configuration by being installed in the execution environment of the OS 112. For this reason, even a program not including such a part of the library or the function module can be included in the scope of the technique of the present invention.

The optical drive 122 reads information such as a program stored in an optical disk 124 such as a CD-ROM (Compact Disc Read Only Memory) or a DVD (Digital Versatile Disc). The optical disc 124 is an example of a non-transitory (non-transitory) recording medium, and is distributed in a state where an arbitrary program is stored in a nonvolatile manner. The computer 100 according to the present embodiment can be configured by reading a program from the optical disk 124 by the optical drive 122 and installing the program in the storage 110. Therefore, the subject of the present invention may be the program itself installed in the storage 110 or the like, or a recording medium such as the optical disk 124 storing the program for realizing the functions and processes according to the present embodiment.

In fig. 2, an Optical recording medium such as an Optical disk 124 is shown as an example of a non-temporary recording medium, but the present invention is not limited to this, and a semiconductor recording medium such as a flash memory, a magnetic recording medium such as a hard disk or a storage tape, or an Magneto-Optical recording medium such as an MO (Magneto-Optical disk) may be used.

Alternatively, the program for realizing the computer 100 may be distributed by being downloaded from a server device or the like via the internet or an intranet, as well as being distributed by being stored in any of the above-described recording media.

Fig. 2 shows an example of a configuration in which the application program 114 is executed by a general-purpose computer (processor 104) to realize the computer 100, but all or part of functions required for realizing the computer 100 may be realized by a hard-wired circuit (integrated circuit) or the like. For example, the present invention can be implemented using an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), or the like.

[ Structure of learning computer 300 ]

Next, the configuration of the learning computer 300 for learning the exhaust gas treatment filter test model for determining whether or not the exhaust gas treatment filter (CSF22) to be tested can be regenerated and used will be described. Fig. 3 is a block diagram showing an example of the configuration of the learning computer 300.

As shown in fig. 3, the learning computer 300 includes, as main hardware elements, a display 302, a processor 304, a memory 306, a network controller 308, a storage 310, and an input device 330.

The display 302 displays information necessary for processing in the learning computer 300. The display 302 is constituted by, for example, an LCD, an organic EL display, or the like.

The processor 304 is an arithmetic unit that executes various programs to execute processing necessary for realizing the learning computer 300. The processor 304 is configured by, for example, 1 or more CPUs, GPUs, and the like. A CPU or GPU with multiple cores may be used. In the learning computer 300, it is preferable to use a GPU or the like suitable for a learning process for generating a learned model.

The memory 306 provides a storage area for temporarily storing program codes, a work memory, and the like when the processor 304 executes a program. As the memory 306, a volatile memory device such as a DRAM or an SRAM can be used.

The network controller 308 transmits and receives data to and from any external device including the computer 100 and the nondestructive testing device 200. The network controller 308 can support any communication method such as the internet, wireless LAN, and Bluetooth.

The storage 310 stores an OS312 executed by the processor 304, an application 314 for realizing a predetermined functional configuration, a preprocessing program 316 for generating a learning data set 324 from non-destructive test result data 320 and result data 322 using the possibility determination result, a learning program 318 for generating a learning completed model 326 using the learning data set 324, and the like.

For the sake of convenience of explanation, reference numerals (116, 326) different from each other are given to the learned model stored in the computer 100 and the learned model generated by the learning computer 300. However, since the learned model 116 stored in the computer 100 is a learned model transmitted (distributed) from the learning computer 300, the 2 learned

models

116 and 326 are substantially the same. In detail, learned model 116 and learned model 326 are substantially identical in network structure and learned parameters.

The learning data set 324 is a training data set to which the non-destructive test result data 320 is given a endorsement (or label) using the possibility determination result data 322. The learned model 326 is an estimation model obtained by performing a learning process using the learning data set 324.

As the storage 310, for example, a nonvolatile memory device such as a hard disk or an SSD can be used.

As a part of libraries and functional modules required when the processor 304 executes the application 314, the preprocessing program 316, and the learning program 318, libraries or functional modules provided as standards by the OS312 can be used. In this case, all the program modules necessary for realizing the corresponding functions are not included in the single application program 314, the preprocessing program 316, and the learning program 318, but a predetermined functional configuration can be realized by being installed in the execution environment of the OS 312. For this reason, even such a program not including a part of the library or the function module can be included in the technical scope of the present invention.

The application 314, the preprocessing program 316, and the learning program 318 may be stored in and circulated to a non-transitory recording medium such as an optical disk, a semiconductor recording medium such as a flash memory, a magnetic recording medium such as a hard disk or a storage tape, and an optical magnetic recording medium such as an MO, and installed in the storage 310. Therefore, the subject of the present invention may be the program itself installed in the storage 310 or the like, or a recording medium storing a program for realizing the functions and processes according to the present embodiment.

Alternatively, the program for realizing the learning computer 300 may be distributed by being stored in any of the above-described recording media and distributed, or downloaded from a server device or the like via the internet or an intranet.

The input device 330 accepts various input operations. As the input device 330, for example, a keyboard, a mouse, a touch panel, or the like can be used.

Fig. 3 shows an example of a configuration in which the general-purpose computer (processor 304) executes the application program 314, the preprocessing program 316, and the learning program 318 to realize the learning computer 300, but all or part of the functions required for realizing the learning computer 300 may be realized by a hard-wired circuit such as an integrated circuit. For example, may be implemented using ASICs, FPGAs, etc.

[ learning processing ]

The learning process executed by the learning computer 300 will be described. Specifically, a method of manufacturing the learned model 326 will be described. Fig. 4 is a functional block diagram for explaining a functional configuration of the learning computer 300.

As shown in fig. 4, the learning computer 300 includes an input receiving unit 350, a control unit 360, and a communication IF (Interface) unit 370. The control unit 360 includes a learning unit 362. The learning unit 362 includes a model 366 for learning and a program 368 for learning. The learning model 366 is composed of a network structure 366N and parameters 366P. The network structure 366N is pre-constructed and stored in the learning computer 300.

The input receiving unit 350 receives an input of the learning dataset 324. The learning data set 324 includes non-destructive test result data for learning and use availability determination result data as teacher data. Specifically, the learning data set 324 includes a plurality of sets of data (learning data), and each set (each learning data) includes non-destructive test result data for learning and use permission determination result data as teacher data. In addition, a part of the plurality of data groups of the data set for learning may be used to evaluate the accuracy of the learned model.

The control unit 360 controls the overall operation of the learning computer 300.

The learning unit 362 in the control unit 360 generates the learned model 326. The generation of the learned model 326 will be described below.

The learning unit 362 updates the value of the parameter 366P of the learning model 366 by machine learning using the learning data set 324. Specifically, the learning section 362 updates the value of the parameter 366P by using the learning program 368. The update of the parameter 366P repeats the number of sets of data (except for the set of data for evaluation) used in learning.

After learning is completed, a learned model 326 can be obtained. Learned model 326 has network structure 366N and learned parameters. The updated parameter 366P corresponds to the learned parameter.

The generated learned model 326 is transmitted to the computer 100 via the communication IF unit 370. As described above, the learned model 326 transmitted to the computer 100 is referred to as "learned model 116" for convenience of description.

Fig. 5 is a flowchart showing the processing procedure of the learning processing in the learning computer 300. The steps shown in fig. 5 are typically implemented by the processor 304 of the learning computer 300 executing an OS312, an application 314, a preprocessing program 316, and a learning program 318 (all with reference to fig. 3).

As shown in fig. 5, in step S1, the learning computer 300 acquires non-destructive test result data 320.

Fig. 6 is a schematic diagram showing a process for learning an exhaust gas treatment filter inspection model. As shown in fig. 6, the nondestructive inspection system includes a nondestructive testing device 200. The nondestructive testing apparatus 200 includes an ultrasonic flaw detector 210, an imaging device 220, and an X-ray CT (computed tomography) scanner 230. The non-destructive testing device 200 may replace the X-ray CT scanner 230 or may include an X-ray transmission device 240 based thereon.

The ultrasonic flaw detector 210 has a flaw detector. The flaw detector sends out pulses of ultrasonic waves and receives reflected echoes. When ultrasonic waves are emitted from a flaw detector disposed at the entrance end 22A of the CSF22, the ultrasonic waves that have not hit a flaw are reflected and returned at the exit end 22B, and a waveform of a reflected echo appears over a distance corresponding to the length of the CSF 22. In the case where there is a defect inside the CSF22, the peak of the waveform of the reflected echo reflected at the defect appears at a position having a short distance compared with the length of the CSF 22. As such, the presence of a defect inside the CSF22 and the position of the defect are detected based on the position of the apex of the waveform of the reflected echo.

The imaging device 220 acquires an optical image of an imaging object and detects the external shape of the imaging object. The imaging Device includes an imaging element such as a CCD (Charge-Coupled Device) image sensor or a CMOS (Complementary Metal Oxide Semiconductor) image sensor. The imaging device 220 of the embodiment images the CSF22, more specifically, the inlet end 22A and the outlet end 22B of the CSF 22.

The X-ray CT scanner 230 generates tomographic image data of the inside of the CSF22 by irradiating X-rays to the CSF22 of the imaging target and processing a detection signal of the X-rays transmitted through the CSF 22. The generated cross-sectional image data is used to detect the occurrence of defects such as cracks or fusion in the CSF22, or the occurrence of a situation in which the CSF22 to be examined cannot be directly reused, such as clogging of the small holes of the CSF22 with coal or incombustible.

The X-ray transmission device 240 generates transmission image data of the entire inside of the CSF22 by irradiating X-rays to the CSF22 of the imaging target and processing detection signals of the X-rays transmitted through the CSF 22. The generated transmission image data is used to detect a situation in which the CSF22 to be examined cannot be directly reused, such as the occurrence of a defect such as a crack or a melt in the CSF22, or the blockage of a pinhole by clogging of coal or incombustible in the pinhole of the CSF 22.

The nondestructive test apparatus 200 transmits nondestructive test result data 320 representing the result of the nondestructive test to the learning computer 300, specifically, to the network controller 308. Specifically, the ultrasonic testing machine 210 transmits ultrasonic testing result data 320A to the learning computer 300. The imaging device 220 transmits the entrance image data 320B to the learning computer 300. The X-ray CT scanner 230 transmits the sectional image data 320C to the learning computer 300. The X-ray transmission device 240 transmits transmission image data 320D to the learning computer 300.

The ultrasonic flaw detection test result data 320A includes 10 pieces of data for 1CSF 22. These 10 data sets include 5 data sets obtained by arranging flaw detectors at 5 positions of the inlet end 22A of 1CSF 22, typically 1 position at the center of the inlet end 22A and 4 positions at equally spaced apart peripheries. The 10 data items additionally include 5 data items obtained by arranging flaw detectors at 5 points of the outlet end 22B of the same 1CSF 22, typically 1 point at the center of the outlet end 22B and 4 points at the periphery spaced at equal intervals.

The preprocessing routine 316 shown in fig. 3 executes preprocessing on each of a plurality of data input from the ultrasonic flaw detector 210 to the learning computer 300. Fig. 7 is a flowchart showing processing steps for preprocessing data of test results obtained by the ultrasonic testing machine 210.

As shown in fig. 7, in step S101, the learning computer 300 acquires 1 ultrasonic flaw detection test result data 320A from the ultrasonic flaw detector 210. Next, in step S102, the learning computer 300 generates point group data in which the ultrasonic flaw detection test result data 320A acquired in step S101 is plotted on the orthogonal coordinate plane.

Fig. 8 is a graph showing an example of the point cloud data. The horizontal axis of the graph shown in fig. 8 represents the length of CSF 22. The end portion of CSF22 on the side where the flaw detector is disposed corresponds to the coordinate zero on the horizontal axis. The vertical axis of the graph shown in fig. 8 represents the echo level received by the flaw detector. The learning computer 300 creates point group data, which is a graph in which the ultrasonic flaw detection test result data 320A is plotted on the orthogonal coordinate plane shown in fig. 8. The data of each point depicted in fig. 8 indicates the magnitude of a reflected echo reflected at a certain point in the longitudinal direction of the CSF 22.

Returning to fig. 7, in step S103, the learning computer 300 generates polyline data in which a straight line connecting the respective points of the point group data shown in fig. 8 is drawn. In step S104, the learning computer 300 colors a part of the generated polyline data. In step S105, the learning computer 300 generates extraction data for extracting data of a part in the longitudinal direction of the CSF22 from the polyline data generated in step S103 and partially colored in step S104.

Fig. 9 is a graph showing an example of extracted data. The data in the area a1 shown in fig. 8 is extracted, and a straight line connecting the point groups included in the area a1 is drawn to generate the extracted data shown in fig. 9. Since the non-sensitive band is present near the coordinate zero of the horizontal axis and the accuracy of the data is low, by removing the data of the non-sensitive band, the resolution of a portion having a high value in the determination of the availability of the CSF22 can be increased, thereby improving the accuracy of the determination of the availability.

By generating the broken line data connecting the straight lines for drawing, the peaks of the waveform of the reflected echo can be easily recognized, as compared with the scatter diagram shown in fig. 8. This improves the accuracy of the determination of the usability.

The threshold T of the echo level is shown in fig. 9. The threshold T is defined to be, for example, 50% of the maximum echo level among the echo levels of the respective data included in the ultrasonic flaw detection test result data 320A. The extracted data shown in fig. 9 is colored in a range where the echo level is equal to or higher than the threshold value T. By coloring in this manner, it becomes easy to accurately recognize the peak of the waveform of the reflected echo, and therefore it is possible to more accurately determine whether or not the CSF22 has a defect inside, and therefore the accuracy of the determination of whether or not the CSF22 can be used is improved.

In the extracted data shown in fig. 9, the peak of the waveform of the reflected echo is located at a position corresponding to the length of CSF 22. Ultrasonic waves emitted from the flaw detector at one end of the CSF22 (e.g., the inlet end 22A) are shown propagating to the other end of the CSF22 (e.g., the outlet end 22B) and are reflected back at the other end and received at the flaw detector. In this case, no defect such as a crack or a melt is present in CSF22, and it is judged to be CSF22 of non-defective products.

Fig. 10 is a graph showing an example of extraction data of a defective product. In the extracted data shown in fig. 10, the peak of the waveform of the reflected echo is located at a position halfway in the longitudinal direction of the CSF 22. Ultrasonic waves emitted from the flaw detector at one end (e.g., inlet end 22A) of the CSF22 are shown not propagating to the other end (e.g., outlet end 22B) of the CSF22, are reflected at a midway point and return, and are received at the flaw detector. In this case, propagation of the ultrasonic wave is blocked at a midway position, and a defect such as a crack is present at the midway position, so that CSF22 becomes discontinuous. Therefore, CSF22 is judged to be a defective product having a defect inside.

Fig. 11 is a graph showing an example of extraction data of qualified products under the conditions. In the extracted data shown in fig. 11, there is no clear peak in the waveform of the reflected echo. The incombustible component trapped in the small hole of CSF22 is solidified into a solid shape, and the peak of the waveform of the reflected echo is not obtained by increasing the scattering attenuation of the ultrasonic wave. In this case, since CSF22 can be reused as long as clogging in the pores can be eliminated by washing, CSF22 that is a qualified product is determined to be acceptable.

The washing may be performed using a solution suitable for melting the nonflammable component, or may be performed by blowing the nonflammable component with air. In addition to or instead of washing, heat may be applied to CSF22 to combust the incombustible partially, thereby eliminating clogging in the pores.

The ultrasonic testing result data 320A thus preprocessed is stored in the storage 310, and constitutes nondestructive testing result data 320.

Fig. 12 is a flowchart showing a processing procedure of preprocessing image data acquired by the imaging device 220. As shown in fig. 12, in step S111, the learning computer 300 acquires a captured image from the imaging device 220. Specifically, the learning computer 300 acquires an entrance image obtained by imaging the entrance end 22A of the CSF22 and an exit image obtained by imaging the exit end 22B of the CSF 22.

Fig. 13 is a diagram showing an example of an entry image. Fig. 13 shows a still image obtained by imaging the inlet end 22A of the CSF 22. The CSF22 is held by the stainless steel holding tube 28 and is contained in the holding tube 28. For this reason, in the inlet image shown in fig. 13, the holding tube 28 is caught around the CSF 22. Fig. 14 is a diagram showing an example of an exit image. Fig. 14 shows a still image obtained by imaging the outlet end 22B of the CSF 22.

Returning to fig. 12, next in step S112, trimming of the entrance image and the exit image is performed. In the images of the CSF22 shown in fig. 13 and 14, objects other than the CSF22, such as the holding tube 28 and the photographer's foot, are captured. For this reason, the entrance image and the exit image are trimmed to cut out the outside of the region a2 shown in fig. 13 and 14, leaving only the CSF22 in the region a 2. By removing objects other than the CSF22 from the entrance image and the exit image, the resolution of a portion having a high value in the determination of the availability of the CSF22 can be increased, thereby improving the accuracy of the determination of the availability of use.

The portal image data 320B thus preprocessed is stored in the storage 310, and constitutes nondestructive test result data 320.

Fig. 15 is a flowchart showing a processing procedure for preprocessing X-ray image data acquired by the X-ray imaging apparatus. If the X-ray imaging apparatus is the X-ray CT scanner 230, the X-ray image data is the sectional image data 320C. If the X-ray imaging apparatus is the X-ray transmission apparatus 240, the X-ray image data is the transmission image data 320D. As shown in fig. 15, in step S121, the learning computer 300 acquires X-ray image data from the X-ray imaging apparatus.

Fig. 16 is a diagram showing an example of transmission image data of a non-defective product as an example of X-ray image data. Fig. 16 shows 1 transmission image data of CSF 22. The transmission image data shown in fig. 16 includes an entrance end 22A and an exit end 22B.

Returning to fig. 15, next in step S122, trimming of the X-ray image data is performed. The transmission image data shown in fig. 16 includes CSF22 in addition. For this reason, the transmission image data is trimmed to cut out the part outside the region A3 shown in fig. 16, leaving the part where only the CSF22 in the region A3 is captured. By removing objects other than the CSF22 from the transmission image data, the resolution of a region having a high value in the determination of the availability of use of the CSF22 can be increased, thereby improving the accuracy of the determination of the availability of use.

In the transmission image data shown in fig. 16, no defect such as a crack or a melt is present in the CSF22 from the entrance end 22A to the exit end 22B, and it is determined to be the CSF22 of the non-defective product.

Fig. 17 is a diagram showing an example of transmission image data of a qualified product. The transmission image data shown in fig. 17 shows a state in which incombustibles are blocked in the area a4 on the exit end 22B side. If the CSF22 can be recycled by washing the incombustible, the CSF22 is judged to be a qualified product.

Fig. 18 is a diagram showing an example of cross-sectional image data of a defective product. In the cross-sectional image data shown in fig. 18, a melted portion exists in the middle of CSF22, and a crack exists near the outlet end 22B. Therefore, CSF22 is judged to be a defective product having a defect inside.

The X-ray image data (the cross-sectional image data 320C and the transmission image data 320D) thus preprocessed are stored in the storage 310, and the nondestructive test result data 320 is configured.

Returning to fig. 5 and 6, in step S2, the learning computer 300 acquires the usability determination result data 322. The usability determination result data 322 includes results of determining whether each CSF22 subjected to the nondestructive test is usable for regeneration as it is, usable for regeneration after washing, or unusable and discarded.

The use availability determination result data 322 as the teacher data can be obtained by, for example, checking the cross-sectional image data 320C acquired by the X-ray CT scanner 230. The usability determination result data 322 can be obtained based on the result of a blockage check test in which a needle is inserted into the small hole from the opening of the inlet end 22A (or the outlet end 22B) of the CSF22 and whether the tip of the needle reaches the blocked outlet end 22B (or the inlet end 22A).

Further, by cutting off and visually observing the CSF22 subjected to the nondestructive test for generating the nondestructive test result data 320 for learning, the usability determination result data 322 can be obtained.

Next, in step S3, the learning computer 300 generates a learning data set 324 by associating the non-destructive test result data 320 with the use propriety determination result data 322. In step S4, the learning computer 300 selects a group of 1 data (learning data) from the generated learning data set 324.

Next, at step S5, the learning computer 300 inputs the data for learning selected at step S4 to the model 366 for learning, thereby obtaining an output using the estimation result.

As explained with reference to fig. 4, the learning model 366 has a network structure 366N. The network fabric 366N comprises the neural network shown in fig. 6.

The network structure 366N includes, for example, a deep neural network such as a Convolutional Neural Network (CNN). The neural network includes an input layer 366Na, an intermediate layer (hidden layer) 366Nb, and an output layer 366 Nc. The intermediate layer 366Nb is multilayered. The input layer 366Na, the intermediate layer 366Nb, and the output layer 366Nc have 1 or more cells (neurons). The number of elements of the input layer 366Na, the intermediate layer 366Nb, and the output layer 366Nc can be set as appropriate.

The cells of adjacent layers are combined with each other, and a weight is set for each combination. An offset is set for each cell. A threshold value is set for each cell. The output value of each cell is determined based on whether or not a value obtained by adding an offset to the sum of the products of the input value and the weight to each cell exceeds a threshold value.

The learning model 366 performs learning to determine whether or not the CSF22 subjected to the nondestructive test can be reused based on the nondestructive test result data 320 of the CSF 22. The learning model 366 includes parameters 366P obtained by learning. The parameters 366P include, for example, the number of layers of the neural network, the number of cells in each layer, the connection relationship between the cells, the weight of connection between the cells, the bias associated with each cell, and the threshold of each cell.

The learning computer 300 performs arithmetic processing for forward propagation of the neural network of the network structure 366N using the nondestructive test result data 320 as an input to the input layer 366 Na. Thus, the learning computer 300 obtains the result of estimating the availability of use of the CSF22 as an output value output from the output layer 366Nc of the neural network.

Next, in step S6, the learning computer 300 calculates an error between the usability judgment result data of the selected learning data and the usability estimation result obtained in step S5.

Next, in step S7, the learning computer 300 updates the learning model 366(CSF examination model). The learning computer 300 updates the parameters 366P of the learning model 366, such as the weight of the combination between the cells, the offset of each cell, and the threshold of each cell, based on the error of the result of estimation of the availability of use of the availability determination result data 322 calculated in step S6.

In this manner, the learning computer 300 executes a process of optimizing the parameter 366P of the learning model 366 so that the result of estimation of the availability of use, which is output by inputting the non-destructive test result data 320 to the learning model 366, is close to the availability determination result data 322 to which the endorsement is added to the non-destructive test result data 320. Then, when the same non-destructive test result data 320 is input to the input layer 366Na, an output value closer to the usability determination result data 322 can be output.

Next, in step S8, the learning computer 300 determines whether all of the learning data sets 324 generated in step S3 have been processed. If all of the learning data sets 324 are not processed (no in step S8), the process from step S4 onward is repeated. When all the learning data sets 324 are processed (yes in step S8), the process proceeds to step S9, and the learning computer 300 saves the current parameters 366P. Through the above, the learning process is completed (END).

In addition, the initial values of the parameters 366P of the learned model 366 can be given by templates. Alternatively, the initial value of the parameter 366P may be given manually by human input. When the model 366 for learning is relearned, the learning computer 300 may prepare an initial value of the parameter 366P based on a value stored as the parameter 366P of the model 366 for learning to be a target of relearning.

[ use of model for learning ]

The learned model 326 defined by the current parameters 366P, which has been subjected to the learning process as described above, is transmitted to the computer 100 (determination computer). The use of the learned model 116 in the computer 100 will be described below. Specifically, a process of estimating whether or not the use of the CSF22 is possible, which is executed by the computer 100, will be described.

Fig. 19 is a functional block diagram for explaining a functional configuration of the computer 100 (determination computer). As shown in fig. 19, the computer 100 includes an input receiving unit 150, a control unit 160, and a display unit 170. The control unit 160 includes a usability estimating unit 161 and a display control unit 162. The usability estimation unit 161 includes a learned model 116.

The input receiving unit 150 receives input of the nondestructive test result data 320.

The control unit 160 controls the overall operation of the computer 100.

The availability estimating unit 161 in the control unit 160 includes a learned model 116. The learned model 116 is composed of a network structure 116N and learned parameters 116P. The network structure 116N is substantially the same as the network structure 366N (see fig. 4), and includes an input layer 116Na, an intermediate layer (hidden layer) 116Nb, and an output layer 116Nc (see fig. 21 described later).

The usability estimating unit 161 estimates whether or not the exhaust gas treatment filter (CSF22) subjected to the nondestructive test can be regenerated using the learned model 116 based on the nondestructive test result data 320. The use availability estimating unit 161 sends the use availability estimating result to the display control unit 162.

The display control unit 162 displays the result of the estimation of availability on the display unit 170. The display section 170 corresponds to the display 102 (see fig. 2).

Fig. 20 is a flowchart showing a processing procedure of an estimation process in the computer 100 (determination computer). The steps shown in fig. 20 are typically performed by the processor 104 of the computer 100 executing an OS112 and application programs 114 (both with reference to fig. 2). Fig. 21 is a schematic diagram showing a process for estimating whether or not the exhaust treatment filter is usable based on the exhaust treatment filter inspection model.

As shown in fig. 20, in step S201, the computer 100 acquires nondestructive test result data 320. The nondestructive test apparatus 200 transmits nondestructive test result data 320 representing the result of the nondestructive test to the computer 100, specifically, to the input receiving unit 150. Specifically, the ultrasonic testing machine 210 transmits the ultrasonic testing result data 320A to the computer 100. The imaging device 220 transmits the entrance image data 320B to the computer 100. The X-ray CT scanner 230 transmits the sectional image data 320C to the computer 100. The X-ray transmission device 240 transmits transmission image data 320D to the computer 100.

Next, in step S202, the computer 100 inputs the non-destructive test result data 320 to the learned model 116. The nondestructive test result data 320 acquired in step S201 is used as input data to the input layer 116Na of the network structure 116N of the learned model 116. The availability estimating unit 161 inputs the non-destructive test result data 320 to each cell included in the input layer 116Na of the network structure 116N.

In step S203, the computer 100 outputs the usage availability estimation result 120. The usability estimating unit 161 outputs the usability estimation result 120 obtained by estimating the usability of regeneration of the CSF22 subjected to the nondestructive test from the output layer 116Nc of the network configuration 116N of the learned model 116. More specifically, the availability estimation unit 161 generates an estimation result that estimates, using the learned model 116, whether or not the CSF22 subjected to the nondestructive test can be directly reused, can be reused by washing, or cannot be reused and should be discarded.

Finally, in step S204, the computer 100 displays the result of usability estimation 120 on the display 102. The process then ENDs (END).

As described above, in the nondestructive inspection system according to the embodiment, the computer 100 includes the learned model 116 for determining whether or not the exhaust gas treatment filter (CSF22) subjected to the nondestructive test is usable for regeneration. As shown in fig. 20, 21, computer 100 is programmed to: non-destructive test result data 320 representing the result of the non-destructive test using CSF22 of non-destructive test apparatus 200 is acquired, and using learned model 116, use availability estimation result 120 that estimates the availability of regeneration use of CSF22 from non-destructive test result data 320 is output.

Therefore, the possibility of regeneration of the CSF22 can be estimated using the artificial intelligence learned model 116 suitable for estimation of the possibility of use of the CSF 22. This makes it possible to easily estimate the availability of CSF22 by computer 100 using artificial intelligence, and to improve the accuracy of determining the availability.

As shown in fig. 5 and 6, the computer 100 is programmed to: the learning model 366 is updated based on the result of estimation of the availability of use of the CSF22 from the non-destructive test result data 320 using the learning model 366 and the error between the availability of use determination result data 322 of the CSF22 included in the learning data set 324. In this way, the model 36 for learning can be sufficiently learned in advance, and a model with high accuracy can be created. By transmitting the learning model 366 to the computer 100 to be used as the learned model 116, the availability of the CSF22 can be accurately determined.

As shown in fig. 6 and 7 to 11, the non-destructive test of CSF22 includes an ultrasonic flaw detection test. The nondestructive test result data 320 includes ultrasonic flaw detection test result data 320A representing the result of the ultrasonic flaw detection test. Thus, the result of the ultrasonic flaw detection test can be used for learning the model 366 for learning, and the result of the ultrasonic flaw detection test can be used to determine whether or not the CSF22 is usable.

As shown in fig. 8, a graph is prepared by plotting the results of the ultrasonic flaw detection test with the horizontal axis representing the length of CSF22 and the vertical axis representing the echo level. Using this chart, it is possible to determine whether or not CSF22 was used in the ultrasonic flaw detection test.

As shown in fig. 9 and 10, a part of the graph is colored. By identifying the colored part of the graph, the accuracy of determining whether or not the CSF22 is usable can be improved.

As shown in fig. 9 to 11, extracted data in which data of a part of the CSF22 in the longitudinal direction is extracted is generated. By removing a region having low value as data such as dead bands, the resolution of a region having high value in the determination of availability of CSF22 can be increased, and therefore the accuracy of the determination of availability can be improved.

As shown in fig. 6 and 12 to 14, the nondestructive test of the CSF22 includes image analysis of an entrance image obtained by imaging the entrance end 22A of the CSF22 and an exit image obtained by imaging the exit end 22B. The nondestructive test result data 320 includes entrance image data 320B representing the result of the image analysis. This allows the results of image analysis of the entrance image and the exit image to be used for learning the model 366 for learning, and allows the use of the CSF22 to be determined more accurately by using the results of image analysis.

As shown in fig. 13 and 14, a part of the entrance image is clipped, and a part of the exit image is clipped. Since the resolution of a portion having a high value in the determination of the availability of the CSF22 can be improved by removing an object other than the CSF22 from the entrance image and the exit image, the accuracy of the determination of the availability can be improved.

As shown in fig. 6 and 18, the non-destructive test of CSF22 includes an X-ray computed tomography test. The non-destructive test result data 320 includes sectional image data 320C representing the result of the X-ray computed tomography test. This allows the result of the X-ray computed tomography test to be used for learning the model 366 for learning, and allows the availability of the CSF22 to be determined more accurately by using the result of the X-ray computed tomography test.

As shown in fig. 6, 16 and 17, the nondestructive test of CSF22 may be performed in place of or in addition to the X-ray computed tomography test. In this case, the nondestructive test result data 320 includes transmission image data 320D indicating the result of the X-ray transmission test. Thus, the result of the X-ray transmission test can be used for learning the model 366 for learning, and the use availability of the CSF22 can be determined more accurately by using the result of the X-ray transmission test.

As shown in fig. 16, a part of the transmission image data 320D is clipped. By removing objects other than the CSF22 from the transmission image data 320D, the resolution of a region having a high value in the determination of the availability of use of the CSF22 can be increased, and therefore the accuracy of the determination of the availability of use can be improved.

[ method for producing distillation model ]

The learned model 116 is not limited to a model obtained by the learning computer 300 through machine learning using the nondestructive test result data 320, and may be a model generated using the model obtained through the learning. For example, the learned model 116 may be another model (distillation model) that is learned based on a result obtained by repeating input and output of data to and from the learned model. Fig. 22 is a flowchart showing a process for generating a distillation model.

As shown in fig. 22, first, in step S301, nondestructive test result data is acquired. Similarly to step S1 shown in fig. 5, the learning computer acquires non-destructive test result data 320 representing the result of the non-destructive test of the exhaust gas treatment filter (CSF22) from the non-destructive testing apparatus 200.

Next, in step S302, the learning computer checks the model input non-destructive test result data 320 with respect to the learned 1 st exhaust gas treatment filter (CSF22) to estimate whether or not the CSF22 subjected to the non-destructive test can be regenerated and used. In step S303, the learning computer outputs a result of the estimation of the usability.

The learning computer inputs the nondestructive test result data 320 acquired in step S301 to the input layer of the learned 1 st CSF inspection model. The result of estimating the availability of use, specifically, the result of estimating whether or not the CSF22 subjected to the nondestructive test can be directly reused, can be reused by washing, or is not reusable and should be discarded, is output from the output layer of the learned 1 st CSF inspection model.

Next, in step S304, the learning computer stores the non-destructive test result data 320 acquired in step S301 and the result of estimation of the availability of use output in step S303 as learning data.

Next, in step S305, the learning computer performs learning of the 2 nd exhaust gas treatment filter (CSF22) inspection model. The learning computer inputs the non-destructive test result data 320 saved in step S304 to the input layer of the 2 nd CSF inspection model. The learning computer outputs an output value indicating whether or not the CSF22 subjected to the nondestructive test is usable or not from the output layer of the 2 nd CSF inspection model. An error between the result of availability estimation output from the 2 nd CSF inspection model and the result of availability estimation output from the 1 st CSF inspection model output in step S303 is obtained. Based on the error, the learning computer updates the parameters of the 2 nd CSF inspection model. The learning of the 2 nd CSF examination model is performed in this manner.

Finally, in step S306, the updated parameters of the 2 nd CSF inspection model are saved as the learned parameters. The process then ENDs (END).

As described above, the 2 nd CSF inspection model (distillation model) is learned by using the nondestructive test result data and the result of estimation of the availability of use of CSF22 using the 1 st CSF inspection model as learning data, and the learning computer can estimate the availability of regeneration of CSF subjected to the nondestructive inspection using the 1 st CSF inspection model and the simple 2 nd CSF inspection model. This reduces the burden on the determination computer for estimating the availability of CSF.

In the above embodiment, the learned model 116 and the learning model 366 include a neural network. The learned model 116 and the model 366 for learning may be models that can accurately estimate whether the exhaust treatment filter can be used for regeneration or not from the nondestructive inspection result of the exhaust treatment filter using machine learning, such as a support vector machine and a decision tree.

The embodiments of the present invention have been described, but the embodiments disclosed herein are merely illustrative in all points and should not be construed as being limited thereto. The scope of the present invention is indicated by the appended claims, and all changes that come within the meaning and range of equivalency of the claims are intended to be embraced therein.

Claims

1. A nondestructive inspection system for an exhaust gas treatment filter, the nondestructive inspection system comprising:

a nondestructive testing device for performing a nondestructive test on the exhaust gas treatment filter; and

a computer having a learned exhaust treatment filter inspection model for determining whether or not the use of the exhaust treatment filter subjected to the nondestructive test is acceptable,

the computer acquires test result data representing a result of the nondestructive test of the exhaust treatment filter by the nondestructive testing device, and outputs a result of estimation of availability of use of the exhaust treatment filter based on the test result data using the learned exhaust treatment filter inspection model.

2. The nondestructive inspection system of claim 1,

the learned exhaust gas treatment filter check model performs a learning process using a learning data set so that the usability estimation result is output based on the test result data when the test result data is input.

3. The nondestructive inspection system of claim 1,

the learned exhaust treatment filter check model is generated by a learning process using a data set for learning,

the learning data set includes a plurality of pieces of learning data in which the use availability determination result data obtained by determining the use availability of the exhaust treatment filter subjected to the non-destructive test is signed with the test result data.

4. The nondestructive inspection system according to any one of claims 1 to 3,

the non-destructive testing comprises ultrasonic flaw detection testing.

5. The nondestructive inspection system of claim 4,

the non-destructive test further includes at least any one of an X-ray transmission test and an X-ray computed tomography test.

6. The nondestructive inspection system of claim 4,

the exhaust gas treatment filter includes: flowing exhaust gas into an inlet end of the exhaust treatment filter; and an outlet end for flow of the exhaust gas from the exhaust treatment filter,

the nondestructive test further includes image analysis for analyzing an entrance image obtained by imaging the entrance end and an exit image obtained by imaging the exit end.

7. A manufacturing method of a learned exhaust treatment filter inspection model, comprising the steps of:

acquiring data for learning including a combination of test result data indicating a result of a nondestructive test on an exhaust gas treatment filter and usability determination result data obtained by determining whether or not the exhaust gas treatment filter subjected to the nondestructive test is usable; and

and generating the exhaust gas treatment filter inspection model based on the plurality of acquired data for learning, the exhaust gas treatment filter inspection model having the test result data as an input and outputting a value relating to availability of the exhaust gas treatment filter.

8. The manufacturing method according to claim 7,

the step of generating the exhaust treatment filter inspection model includes:

inputting the test result data into the exhaust treatment filter inspection model to obtain an output of a usability estimation result obtained by estimating usability of the exhaust treatment filter;

obtaining a comparison result of comparing the usability determination result data corresponding to the test result data input to the exhaust gas treatment filter inspection model with the usability estimation result; and

a step of updating the exhaust treatment filter inspection model based on the comparison result.

9. A method for generating data for learning an exhaust gas treatment filter inspection model for determining whether or not an exhaust gas treatment filter to be inspected can be used, the method comprising the steps of:

a step of acquiring test result data representing a result of a non-destructive test of the exhaust treatment filter;

and acquiring use availability determination result data obtained by determining whether or not the exhaust treatment filter subjected to the nondestructive test is available for use.

10. The learning data generation method according to claim 9,

the non-destructive testing comprises an ultrasonic flaw detection test,

the step of obtaining the test result data comprises: and acquiring ultrasonic flaw detection test result data representing a result of the ultrasonic flaw detection test.

11. The learning data generation method according to claim 10,

the step of obtaining the test result data comprises:

and generating a graph in which the ultrasonic testing result data is plotted on the horizontal axis of the length of the exhaust gas treatment filter and on the vertical axis of the echo level.

12. The learning data generation method according to claim 11,

the step of obtaining the test result data comprises:

a step of coloring a portion of the chart.

13. The method for generating learning data according to any one of claims 10 to 12, wherein the data is generated by a computer,

the step of obtaining the test result data comprises:

and generating extraction data for extracting data of a part of the exhaust gas treatment filter in the longitudinal direction from the ultrasonic testing result data.

14. The learning data generation method according to claim 10,

the nondestructive test further includes image analysis for analyzing an entrance image obtained by imaging the entrance end and an exit image obtained by imaging the exit end,

the step of obtaining the test result data comprises:

acquiring the entrance image; and

and acquiring the exit image.

15. The learning data generation method according to claim 14,

the step of obtaining the test result data comprises:

a step of generating a partial portal image in which a part of the portal image is clipped; and

a step of generating a partial exit image in which a portion of the exit image is cropped.

16. The learning data generation method according to claim 10,

the non-destructive testing comprises X-ray computed tomography testing,

the step of obtaining the test result data comprises:

and acquiring cross-sectional image data representing the result of the X-ray computed tomography test.

17. The learning data generation method according to claim 16,

the step of obtaining the test result data comprises:

and acquiring partial tomogram data obtained by trimming a part of the tomogram data.

18. A method for manufacturing a learned exhaust treatment filter inspection model, comprising the steps of:

acquiring test result data representing a result of a non-destructive test of the exhaust gas treatment filter;

inputting the test result data to the learned 1 st exhaust treatment filter inspection model to obtain an output of a usability estimation result obtained by estimating usability of the exhaust treatment filter; and

and learning the 2 nd exhaust gas treatment filter inspection model by learning data including the test result data and the result of availability estimation.