JP7471031B2 - DEFECT ESTIMATION DEVICE, DEFECT ESTIMATION METHOD, AND PROGRAM - Google Patents

Description

Article 30, Paragraph 2 of the Patent Act applies. Published in the Proceedings of the 60th Kanto Student Association Student Graduation Research Presentation Symposium, sponsored by the Japan Society of Mechanical Engineers, on March 10, 2021, entitled "Analysis of Surface Stress and Defects in Carbon Fiber Reinforced Plastics with Delamination Using Machine Learning."

Article 30, paragraph 2 of the Patent Act applies. May 27, 2021, Published in the Proceedings of the 26th Computational Engineering Conference, hosted by the Japan Society for Computational Engineering, Vol. 26 B-07-03, "Inverse estimation of internal defects based on surface stress in carbon fiber reinforced plastics using machine learning"

Application of Article 30, Paragraph 2 of the Patent Act September 7, 2021, https://complas2021. cimne. com/technical_program, published in the COMPLAS2021 Lecture Abstracts "Estimation of Defects in Carbon Fiber Reinforced Plastic Based on Surface Pressure by Machine Learning"

This invention relates to a defect estimation device, a defect estimation method, and a program.

Carbon Fiber Reinforced Plastic (CFRP) is a composite material that uses resin as the base material and carbon fiber as the reinforcing material. CFRP is generally used by laminating prepregs whose strength and rigidity are reinforced in one direction. Hereinafter, CFRP with a laminated structure will be referred to as a "CFRP laminate."

To detect internal defects in CFRP laminates, non-destructive testing such as ultrasonic measurement (see Non-Patent Document 1) or radiographic transmission (see Non-Patent Document 2) has been conventionally performed. In addition, to solve the shortcomings of conventional non-destructive testing, research has been conducted on, for example, stress analysis using infrared thermography (see Non-Patent Document 3) and methods using machine learning models (see Non-Patent Document 4).

Shuichi Mikami, Toshiyuki Oshima, Noboru Sugawara, Tomoyuki Yamazaki, "Quantitative evaluation of flaw detection by ultrasonic inspection using detailed analysis of echo waveforms", Journal of the Japan Society of Civil Engineers, vol. 501, pp. 103-112, 1994. Sultan, M., Worden, K., Pierce, S., Hickey, D., Staszewski,W., Dulieu-Barton, J., and Hodzic, A., "On impact damage detection and quantification for CFRP laminates," Mechanical Systems and Signal Processing, vol. 25, pp. 3135-3152, 2011. Sakagami, T., Mizokami, Y., Shiozawa, D., Fujimoto, T., Izumi, Y., Hanai, T., and Moriyama, A., "Verification of the repair effect for fatigue cracks in members of steel bridges based on thermoelastic stress measurement", Engineering Fracture Mechanics, vol. 183, pp. 1-12, 2017. Begoro, K. Nishi, Z. Huang, and Y. Fujikawa, "Damage Identification of CFRP Laminates Using Neural Networks and Experimental Data," Transactions of the Japan Society of Mechanical Engineers, Series A, vol. 62, pp. 2338-2343, 1996.

However, the techniques described in Non-Patent Documents 3 and 4 have the problem that they are unable to pinpoint the location of internal defects in a CFRP laminate in detail.

In view of the above technical problems, the present invention aims to estimate the three-dimensional position of internal defects in a fiber-reinforced plastic laminate.

In order to solve the above problem, a defect estimation device according to one aspect of the present invention includes a feature amount receiving unit configured to receive input of feature data based on the difference between the temperature distribution before a temperature change occurs on the surface of a fiber-reinforced plastic laminate and the temperature distribution after the temperature change occurs, and a defect position estimation unit configured to estimate the three-dimensional position of an internal defect in the laminate by inputting the feature data received by the feature amount receiving unit into a model that has learned the relationship between the feature data and position information that represents the three-dimensional position of an internal defect. The position information is information that divides each layer of the laminate into a predetermined number of meshes and indicates whether or not an internal defect exists in each mesh.

According to one aspect of the present invention, it is possible to estimate the three-dimensional position of an internal defect in a fiber-reinforced plastic laminate.

FIG. 1 illustrates an example of an overall configuration of a defect estimation system. FIG. 2 illustrates an example of a hardware configuration of a computer. FIG. 2 illustrates an example of a hardware configuration of a temperature measuring device. FIG. 1 is a diagram showing an example of a CFRP laminate. FIG. 2 is a diagram illustrating an example of a functional configuration of a defect estimation system according to the first embodiment. FIG. 4 is a diagram illustrating an example of a processing procedure of a learning method in the first embodiment. FIG. 4 is a diagram illustrating an example of a processing procedure of an estimation method in the first embodiment. FIG. 13 is a diagram illustrating an example of a functional configuration of a defect estimation system in a first modified example. FIG. 11 is a diagram showing an example of a processing procedure of a learning method in Modification 1. FIG. 11 is a diagram illustrating an example of a processing procedure of an estimation method in Modification 1. FIG. 13 is a diagram illustrating an example of a functional configuration of a defect estimation system according to a third embodiment. FIG.

Hereinafter, each embodiment of the present invention will be described with reference to the accompanying drawings. In this specification and drawings, components having substantially the same functional configuration are designated by the same reference numerals, and redundant description will be omitted.

[First embodiment]
Carbon fiber reinforced plastics (CFRP) have high specific elasticity and are widely used in the aerospace and aviation fields. In the future, their use is expected to expand in the automotive field as the electrification of automobiles progresses.

CFRP (CFRP laminates) with a laminated structure may develop internal defects during use or manufacturing. Internal defects that occur during use include, for example, delamination, fiber breakage, and base material cracks. Internal defects that occur during manufacturing include, for example, the inclusion of foreign matter.

To detect internal defects in CFRP laminates, non-destructive testing such as ultrasonic measurement (see Non-Patent Document 1) or radiography (see Non-Patent Document 2) has been used. However, ultrasonic measurement requires contact materials such as water or oil, and the skill of the testing technician has a large impact on the accuracy of the test results. In addition, radiography requires safety management such as setting up controlled areas, and requires development time, resulting in high economic and time costs.

In this context, stress analysis using infrared thermography (see Non-Patent Document 3) has attracted attention. Infrared thermography is a device that uses an infrared sensor to measure the distribution of infrared energy emitted from the surface of an object and converts it into a temperature distribution. Non-Patent Document 3 discloses that defect detection using infrared thermography is useful for structures such as bridges. However, the technology disclosed in Non-Patent Document 3 cannot pinpoint the detailed location of internal defects.

Research is also being conducted on a method using a machine learning model (see Non-Patent Document 4). In Non-Patent Document 4, a neural network is created that uses the first to third natural frequencies as input data to determine the location and amount of damage. However, in Non-Patent Document 4, the CFRP laminate plate is divided into 10 parts in the longitudinal direction, and the location and amount of damage is determined for 8 elements excluding the 2 elements at the tip. Therefore, the technology disclosed in Non-Patent Document 4 has low resolution for the output position information, and cannot specify the detailed location of the internal defect.

The defect estimation device in this embodiment estimates the three-dimensional location of internal defects in a CFRP laminate from an image obtained by measuring the surface temperature of the laminate using infrared thermography. For the estimation, a Convolutional Neural Network (CNN) is used that has learned the relationship between the distribution of changes in the surface temperature of the CFRP laminate and the three-dimensional location of internal defects in the laminate.

The defect estimation device of this embodiment can obtain an estimate of the position information (three-dimensional data) of an internal defect in a CFRP laminate from an image (two-dimensional data) that represents the surface temperature of the laminate. In other words, it is possible to precisely estimate the three-dimensional position of an internal defect in a CFRP laminate.

<Overall configuration of defect estimation system>
First, the overall configuration of a defect estimation system according to this embodiment will be described with reference to Fig. 1. Fig. 1 is a block diagram showing an example of the overall configuration of a defect estimation system according to this embodiment.

As shown in FIG. 1, the defect estimation system 100 in this embodiment includes a temperature measuring device 1 and a defect estimation device 2. The temperature measuring device 1 and the defect estimation device 2 are connected to each other so as to be able to communicate data with each other via a communication network N1 such as a LAN (Local Area Network) or the Internet.

The temperature measuring device 1 is an electronic device that measures the surface temperature of a CFRP laminate that is the learning or estimation target. The temperature measuring device 1 changes the surface temperature of the CFRP laminate, and measures the surface temperature of the CFRP laminate with an infrared camera before and after the surface temperature change. The temperature measuring device 1 also calculates an image that shows the change in surface temperature (hereinafter also referred to as a "temperature change image") based on two images that show the distribution of the surface temperature measured before and after the surface temperature change (hereinafter also referred to as "surface temperature images").

The method for changing the surface temperature of a CFRP laminate varies depending on the shape and application of the CFRP laminate. In this embodiment, the surface temperature is changed by applying a fixed external force to the CFRP laminate. Specifically, both ends of a plate-shaped CFRP laminate are supported and pulled in opposite directions.

The method of changing the surface temperature of a CFRP laminate is not limited to applying a fixed external force. For example, the surface temperature may be changed by applying heat to the CFRP laminate. In this case, infrared rays are irradiated onto the surface of the CFRP laminate to raise the surface temperature, and then the temperature change over time is measured.

The defect estimation device 2 is an information processing device such as a PC (Personal Computer), workstation, or server that estimates the position of an internal defect in a CFRP laminate. The defect estimation device 2 learns an estimation model that takes as input feature data based on the temperature change image and outputs an estimate of the position information of the internal defect, based on a temperature change image that measures the temperature change of the CFRP laminate to be learned and position information of the internal defect input by a user. The defect estimation device 2 also uses the learned estimation model to estimate the position of the internal defect from the temperature change image that measures the temperature change of the CFRP laminate to be learned.

Note that the overall configuration of the defect estimation system 100 shown in FIG. 1 is just one example, and various system configuration examples are possible depending on the application and purpose. For example, the defect estimation device 2 may be realized by multiple computers, or may be realized as a cloud computing service.

<Hardware configuration of defect estimation system>
Next, the hardware configuration of the defect estimation system in this embodiment will be described with reference to FIGS.

<Computer hardware configuration>
The defect estimation apparatus 2 in this embodiment is realized by, for example, a computer. Fig. 2 is a block diagram showing an example of the hardware configuration of a computer 500 in this embodiment.

As shown in FIG. 2, the computer 500 has a CPU (Central Processing Unit) 501, a ROM (Read Only Memory) 502, a RAM (Random Access Memory) 503, a HDD (Hard Disk Drive) 504, an input device 505, a display device 506, a communication I/F (Interface) 507, and an external I/F 508. The CPU 501, the ROM 502, and the RAM 503 form a so-called computer. Each piece of hardware in the computer 500 is connected to each other via a bus line 509. Note that the input device 505 and the display device 506 may be connected to the external I/F 508 for use.

The CPU 501 is a computing device that reads programs and data from a storage device such as the ROM 502 or the HDD 504 onto the RAM 503 and executes processing to realize the overall control and functions of the computer 500.

ROM 502 is an example of a non-volatile semiconductor memory (storage device) that can retain programs and data even when the power is turned off. ROM 502 functions as a main storage device that stores various programs, data, etc. required for CPU 501 to execute various programs installed in HDD 504. Specifically, ROM 502 stores boot programs such as BIOS (Basic Input/Output System) and EFI (Extensible Firmware Interface) that are executed when computer 500 is started, as well as data such as OS (Operating System) settings and network settings.

RAM 503 is an example of a volatile semiconductor memory (storage device) from which programs and data are erased when the power is turned off. RAM 503 is, for example, a dynamic random access memory (DRAM) or a static random access memory (SRAM). RAM 503 provides a working area into which various programs installed in HDD 504 are expanded when executed by CPU 501.

The HDD 504 is an example of a non-volatile storage device that stores programs and data. The programs and data stored in the HDD 504 include an OS, which is basic software that controls the entire computer 500, and applications that provide various functions on the OS. Note that instead of the HDD 504, the computer 500 may use a storage device that uses flash memory as a storage medium (e.g., an SSD: Solid State Drive, etc.).

The input device 505 includes a touch panel that the user uses to input various signals, operation keys or buttons, a keyboard or mouse, a microphone for inputting sound data such as voice, etc.

The display device 506 is composed of a display such as a liquid crystal or organic EL (Electro-Luminescence) display for displaying a screen, a speaker for outputting sound data such as voice, etc.

The communication I/F 507 is an interface that connects to a communication network and enables the computer 500 to perform data communication.

The external I/F 508 is an interface with external devices. Examples of external devices include a drive device 510.

The drive device 510 is a device for setting the recording medium 511. The recording medium 511 here includes media that record information optically, electrically, or magnetically, such as CD-ROMs, flexible disks, and magneto-optical disks. The recording medium 511 may also include semiconductor memory that records information electrically, such as ROM and flash memory. This allows the computer 500 to read and/or write to the recording medium 511 via the external I/F 508.

The various programs to be installed in the HDD 504 are installed, for example, by setting the distributed recording medium 511 in a drive device 510 connected to the external I/F 508 and reading the various programs recorded on the recording medium 511 by the drive device 510. Alternatively, the various programs to be installed in the HDD 504 may be installed by downloading them from a network different from the communication network via the communication I/F 507.

<Hardware configuration of temperature measurement device>
3 is a block diagram showing an example of a hardware configuration of the temperature measuring device 1 according to the present embodiment. As shown in FIG. 3, the temperature measuring device 1 according to the present embodiment includes a control device 101, a pair of support units 102, a drive control unit 103, and an infrared camera 104.

The support part 102 consists of a first support part 102A that supports one end of the CFRP laminate 1000, and a second support part 102B that supports the other end. The drive control part 103 controls the first support part 102A and the second support part 102B to move linearly along a line connecting the first support part 102A and the second support part 102B.

With the first support part 102A and the second support part 102B supporting both ends of the CFRP laminate 1000, the drive control part 103 moves the first support part 102A and the second support part 102B in opposite directions to each other, thereby applying tensile stress to the CFRP laminate 1000. This causes a temperature drop in the CFRP laminate 1000 proportional to the stress fluctuation, and the surface temperature of the CFRP laminate 1000 changes.

The infrared camera 104 receives infrared radiation emitted from the surface of the CFRP laminate 1000 and generates a surface temperature image of the CFRP laminate 1000 through infrared analysis.

The control device 101 is realized by, for example, a computer. The control device 101 has a CPU 501, a ROM 502, a RAM 503, an input device 505, a display device 506, and an external I/F 508. The CPU 501, the ROM 502, and the RAM 503 form a so-called computer. Each piece of hardware of the control device 101 is connected to each other via a bus line 509. Note that the input device 505 and the display device 506 may be connected to the external I/F 508 for use.

The drive control unit 103 and the infrared camera 104 are connected to the external I/F 508 and are controlled by the control device 101. In addition, the surface temperature image acquired by the infrared camera 104 is input to the control device 101 via the external I/F 508.

Here, the structure of the CFRP laminate in this embodiment will be described with reference to Figure 4. Figure 4 is a conceptual diagram showing an example of a CFRP laminate in this embodiment.

As shown in FIG. 4, the CFRP laminate 1000 in this embodiment is constructed by laminating multiple prepregs 1001. Each prepreg 1001 has a resin base material 1002, which is reinforced in one direction by carbon fiber reinforcing fibers 1003. The CFRP laminate 1000 is constructed to exhibit isotropy overall by rotating the fiber orientation at an equal angle in each layer when laminating the prepregs 1001.

Figure 4 shows an example in which each prepreg is rotated by 90°, but it may be rotated by, for example, 30°, 45°, 60°, etc. Also, while Figure 4 shows an example in which 10 layers are laminated, the number of layers is not limited as long as it is two or more. In this embodiment, a CFRP laminate is assumed, but the types of reinforcing fibers and resins may be changed as appropriate.

<Functional configuration of the defect estimation system>
Next, the functional configuration of the defect estimation system in this embodiment will be described with reference to Fig. 5. Fig. 5 is a block diagram showing an example of the functional configuration of the defect estimation system in this embodiment.

<Functional configuration of temperature measuring device>
As shown in FIG. 5, the temperature measuring device 1 in this embodiment includes a temperature changing section 11, a temperature measuring section 12, and an image generating section 13.

The temperature change unit 11 is realized by the support unit 102 and drive control unit 103 shown in FIG. 3. The temperature measurement unit 12 is realized by the infrared camera 104 shown in FIG. 3. The image generation unit 13 is realized by the processing that the CPU 501 executes by the program expanded from the ROM 502 to the RAM 503 shown in FIG. 3.

The temperature change unit 11 changes the surface temperature of the CFRP laminate to be learned and estimated by the drive control unit 103 driving the support unit 102.

The temperature measurement unit 12 uses an infrared camera 104 to measure the surface temperature of the CFRP laminate to be learned and estimated. The temperature measurement unit 12 also generates a surface temperature image of the CFRP laminate. The temperature measurement unit 12 then sends the generated surface temperature image to the image generation unit 13.

The image generation unit 13 generates a temperature change image based on two surface temperature images generated before and after the temperature change unit 11 changes the surface temperature of the CFRP laminate. The image generation unit 13 also transmits the generated temperature change image to the defect estimation device 2.

<Functional configuration of the defect estimation device>
As shown in FIG. 5 , the defect estimation device 2 in this embodiment includes an image receiving unit 21, a position information receiving unit 22, a feature calculation unit 23, a model learning unit 24, a model storage unit 25, a defect position estimation unit 26, and a result output unit 27.

The image receiving unit 21, the position information receiving unit 22, the feature amount calculation unit 23, the model learning unit 24, the defect position estimation unit 26, and the result output unit 27 are realized by processing executed by the CPU 501 of a program loaded onto the RAM 503 from the HDD 504 shown in FIG. 2. The model storage unit 25 is realized by the HDD 504 shown in FIG. 2.

The image receiving unit 21 receives a temperature change image from the temperature measuring device 1. The image receiving unit 21 also sends the temperature change image to the feature calculation unit 23.

The position information receiving unit 22 receives input of position information of internal defects in the CFRP laminate to be learned in response to user operation. The position information receiving unit 22 also sends the received position information of the internal defects to the model learning unit 24.

The feature amount calculation unit 23 calculates feature data based on the temperature change image received from the image reception unit 21. In this embodiment, the feature data is the surface principal stress sum distribution of the CFRP laminate.

When the feature amount calculation unit 23 calculates feature data based on a temperature change image related to the CFRP laminate to be learned, it sends the feature data to the model learning unit 24. In addition, when the feature amount calculation unit 23 calculates feature data based on a temperature change image related to the CFRP laminate to be estimated, it sends the feature data to the defect position estimation unit 26.

The model learning unit 24 generates learning data by associating the position information of the internal defect accepted by the position information accepting unit 22 with the feature data calculated by the feature amount calculating unit 23. In addition, the model learning unit 24 uses the generated learning data to learn an estimation model.

The model storage unit 25 stores the trained estimation model generated by the model training unit 24.

The defect position estimation unit 26 reads out the estimation model from the model storage unit 25 and inputs the feature data received from the feature amount calculation unit 23 into the estimation model to calculate an estimate of the position information of the internal defect in the CFRP laminate to be estimated.

The result output unit 27 outputs the estimation result of the internal defect to the display device 506 or the like. The estimation result includes an estimate of the position information of the internal defect calculated by the defect position estimation unit 26.

<Processing Procedure of Defect Estimation System>
Next, a processing procedure of a defect estimation method executed by the defect estimation system in this embodiment will be described with reference to Fig. 6 and Fig. 7. The defect estimation method in this embodiment includes a learning method (see Fig. 6) for learning an estimation model, and an estimation method (see Fig. 7) for estimating a three-dimensional position of an internal defect using the learned estimation model.

<How to learn>
FIG. 6 is a flowchart showing an example of a processing procedure of a learning method according to this embodiment.

In step S12-1, the temperature measurement unit 12 of the temperature measurement device 1 measures the surface temperature of the CFRP laminate to be learned before changing the surface temperature of the laminate. Next, the temperature measurement unit 12 generates a surface temperature image (hereinafter also referred to as a "pre-temperature change image") that shows the distribution of the measured surface temperature. Next, the temperature measurement unit 12 sends the pre-temperature change image to the image generation unit 13.

In step S11, the temperature change unit 11 of the temperature measurement device 1 changes the temperature of the CFRP laminate to be learned. Specifically, a predetermined fixed external force is applied to the CFRP laminate to be learned.

In step S12-2, the temperature measurement unit 12 included in the temperature measurement device 1 changes the surface temperature of the CFRP laminate to be learned, and then measures the surface temperature of the laminate. Next, the temperature measurement unit 12 generates a surface temperature image (hereinafter also referred to as the "post-temperature change image") that shows the distribution of the measured surface temperature. Next, the temperature measurement unit 12 sends the post-temperature change image to the image generation unit 13.

In step S13, the image generating unit 13 included in the temperature measuring device 1 receives the pre-temperature change image and the post-temperature change image from the temperature measuring unit 12. Next, the image generating unit 13 generates a temperature change image representing the surface temperature change of the CFRP laminate to be learned based on the pre-temperature change image and the post-temperature change image. Next, the image generating unit 13 transmits the temperature change image to the defect estimation device 2.

Specifically, the image generating unit 13 calculates the difference between the image before the temperature change and the image after the temperature change. That is, for each pixel of each image, the surface temperature represented by the image before the temperature change is subtracted from the surface temperature represented by the image after the temperature change.

In step S21, the image receiving unit 21 included in the defect estimation device 2 receives a temperature change image of the CFRP laminate to be learned from the temperature measuring device 1. Next, the image receiving unit 21 sends the temperature change image received from the temperature measuring device 1 to the feature calculation unit 23.

In step S22, the position information receiving unit 22 included in the defect estimation device 2 receives, in response to a user's operation, input of position information representing the three-dimensional position of an internal defect in the CFRP laminate to be learned. Next, the position information receiving unit 22 sends the received position information of the internal defect to the model learning unit 24.

In this embodiment, the position information of the internal defect includes the three-dimensional coordinates of the internal defect in the CFRP laminate and the size of the internal defect. The size of the internal defect is the range in which the internal defect exists in each layer of the CFRP laminate.

Specifically, the position information of internal defects is information in which each layer of the CFRP laminate is divided into a specified number of meshes, and a channel is set for each mesh to indicate whether or not an internal defect exists. For example, the channel is set to 1 if an internal defect exists, and 0 if it does not. In other words, the position of each mesh represents a three-dimensional coordinate obtained by adding the depth of the layer to the two-dimensional coordinate as seen from the surface of the CFRP laminate. Also, the range in each layer where the channel value is 1 represents the size of the internal defect.

In this embodiment, the mesh is divided into equal orthogonal intervals. However, if a location where defects are likely to occur is known, the area near that location may be divided finely and other areas may be divided roughly.

In step S23, the feature amount calculation unit 23 included in the defect estimation device 2 receives the temperature change image from the image reception unit 21. Next, the feature amount calculation unit 23 calculates the surface principal stress sum distribution based on the temperature change image. Then, the feature amount calculation unit 23 sends the calculated surface principal stress sum distribution to the model learning unit 24.

Specifically, the feature calculation unit 23 obtains the surface temperature change distribution by dividing the temperature change image into a predetermined number of meshes and calculating the temperature change amount corresponding to each mesh. The temperature change amount of a mesh may be the temperature change amount of a representative point (such as the center) of the mesh, or may be the average temperature change amount within the mesh. The method of dividing the mesh is the same as that for the position information of the internal defect.

Next, the feature calculation unit 23 calculates the surface principal stress sum distribution by applying Kelvin's theory (see Reference 1) to the surface temperature change distribution. Kelvin's theory gives the relationship between the temperature change ΔT due to the thermoelastic effect and the change Δσ in the principal stress sum as ΔT = -kTΔσ, where k is the thermoelastic coefficient and T is the absolute temperature.

[Reference 1] W. Thomson (Lord Kelvin), "On the Dynamical Theory of Heat", Trans. Roy. Soc., Vol.20, pp.261-283, 1853.

Steps S12-1 to S23 are repeatedly executed for each CFRP laminate to be learned. As a result, the surface principal stress distribution and internal defect position information corresponding to each CFRP laminate to be learned are input to the model learning unit 24.

In step S24, the model learning unit 24 included in the defect estimation device 2 receives the position information of the internal defect from the position information receiving unit 22. The model learning unit 24 also receives the surface principal stress distribution from the feature calculation unit 23.

Next, the model learning unit 24 generates learning data by associating the surface principal stress distribution with the position information of the internal defect. At this time, the model learning unit 24 may standardize the surface principal stress sum distribution included in the learning data with respect to the entire learning data.

The model learning unit 24 then uses the generated learning data to learn an estimation model. The estimation model in this embodiment is a convolutional neural network that receives the surface principal stress sum distribution of the CFRP laminate as input and outputs an estimate of the position information of the internal defects in the laminate.

A convolutional neural network is a special type of neural network that is used to process data with a lattice structure. Examples of data with a lattice structure include time series data, in which samples taken at equal time intervals are arranged in one dimension, or image data, in which pixels are arranged in two dimensions.

The convolutional neural network in this embodiment is configured as follows. The loss function uses binary intersection error. The optimization method uses Adam. In this case, the learning rate is set to 0.00003. The mini-batch size is 16, and the number of training times is 1000. Note that regularization methods such as L2 regularization or batch normalization do not need to be used.

In step S25, the model learning unit 24 of the defect estimation device 2 stores the learned estimation model in the model memory unit 25.

<Estimation method>
FIG. 7 is a flowchart showing an example of a processing procedure of the estimation method according to this embodiment.

In step S12-1, the temperature measurement unit 12 included in the temperature measurement device 1 measures the surface temperature of the CFRP laminate to be estimated before changing the surface temperature of the laminate. Next, the temperature measurement unit 12 generates an image before the temperature change that shows the distribution of the measured surface temperature. Next, the temperature measurement unit 12 sends the image before the temperature change to the image generation unit 13.

In step S11, the temperature change unit 11 of the temperature measurement device 1 changes the temperature of the CFRP laminate to be estimated. Specifically, a fixed external force similar to that in the learning process is applied to the CFRP laminate to be estimated.

In step S12-2, the temperature measurement unit 12 included in the temperature measurement device 1 changes the surface temperature of the CFRP laminate to be estimated, and then measures the surface temperature of the laminate. Next, the temperature measurement unit 12 generates a post-temperature change image that shows the distribution of the measured surface temperature. Next, the temperature measurement unit 12 sends the post-temperature change image to the image generation unit 13.

In step S13, the image generating unit 13 included in the temperature measuring device 1 receives the pre-temperature change image and the post-temperature change image from the temperature measuring unit 12. Next, the image generating unit 13 generates a temperature change image representing the surface temperature change of the CFRP laminate to be estimated based on the pre-temperature change image and the post-temperature change image. Next, the image generating unit 13 transmits the temperature change image to the defect estimation device 2.

In step S21, the image receiving unit 21 included in the defect estimation device 2 receives a temperature change image relating to the CFRP laminate to be estimated from the temperature measuring device 1. Next, the image receiving unit 21 sends the temperature change image received from the temperature measuring device 1 to the feature calculation unit 23.

In step S23, the feature amount calculation unit 23 included in the defect estimation device 2 receives the temperature change image from the image reception unit 21. Next, the feature amount calculation unit 23 calculates the surface principal stress sum distribution based on the temperature change image. Then, the feature amount calculation unit 23 sends the calculated surface principal stress sum distribution to the model learning unit 24.

In step S26, the defect position estimation unit 26 included in the defect estimation device 2 receives the surface principal stress sum distribution from the feature amount calculation unit 23. Next, the defect position estimation unit 26 reads out the estimation model from the model storage unit 25. The defect position estimation unit 26 then inputs the received surface principal stress sum distribution into the read out estimation model to calculate an estimate of the position information of the internal defect in the CFRP laminate to be estimated. The defect position estimation unit 26 then sends the estimate of the position information of the internal defect to the result output unit 27.

In step S27, the result output unit 27 included in the defect estimation device 2 receives an estimate of the position information of the internal defect from the defect position estimation unit 26. Next, the result output unit 27 outputs the estimate result of the three-dimensional position of the internal defect to the display device 506 or the like. The estimation result includes the estimate of the position information of the internal defect calculated by the defect position estimation unit 26.

Effects of the First Embodiment
The defect estimation system in this embodiment learns an estimation model that receives as input the surface principal stress sum distribution of a CFRP laminate to be learned and outputs an estimated value of position information representing the three-dimensional position of an internal defect. The defect estimation system in this embodiment also inputs the surface principal stress sum distribution based on an image representing a change in the surface temperature of the CFRP laminate to be learned into the learned estimation model, thereby estimating the three-dimensional position of an internal defect in the laminate. Therefore, the defect estimation system in this embodiment can estimate the three-dimensional position of an internal defect in a CFRP laminate.

In addition, the position information in this embodiment includes the depth of the layer in which the internal defect of the CFRP laminate exists and the range in which the internal defect exists in that layer. Therefore, the defect estimation system in this embodiment can precisely estimate the three-dimensional coordinates and size of the internal defect in the CFRP laminate.

[Second embodiment]
The defect estimation device 2 in the first embodiment uses the surface principal stress sum distribution as feature data to learn an estimation model and estimate the three-dimensional position of an internal defect. The defect estimation device 2 in the second embodiment is configured to use the surface temperature change distribution as feature data.

<Functional configuration of the defect estimation system>
The feature amount calculation unit 23 in this embodiment calculates the surface temperature change distribution based on the temperature change image received from the image receiving unit 21. The feature amount calculation unit 23 in the first embodiment calculates the surface temperature change distribution based on the temperature change image, and calculates the surface principal stress sum distribution by applying Kelvin's theory to the surface temperature change distribution. The feature amount calculation unit 23 in this embodiment uses the surface temperature change distribution before applying Kelvin's theory as feature data.

The model learning unit 24 in this embodiment generates learning data by associating the position information of the internal defect received by the position information receiving unit 22 with the surface temperature change distribution calculated by the feature calculation unit 23. Furthermore, the model learning unit 24 in this embodiment uses the generated learning data to learn a convolutional neural network (estimation model) that receives the surface temperature change distribution of the CFRP laminate as input and outputs an estimate of the position information of the internal defect in the laminate.

In this embodiment, the defect position estimation unit 26 reads out the estimation model from the model storage unit 25, and inputs the surface temperature change distribution received from the feature calculation unit 23 into the estimation model to calculate an estimate of the position information of the internal defect in the CFRP laminate to be estimated.

Effects of the Second Embodiment
The defect estimation system in this embodiment learns an estimation model using the surface temperature change distribution calculated based on the temperature change image, and estimates the position of an internal defect. This eliminates the need to calculate the surface principal stress sum distribution from the surface temperature change distribution, thereby improving the processing speed of the feature calculation unit 23.

[Modification 1]
In the first and second embodiments, the defect estimation device 2 is configured to calculate feature data (i.e., surface principal stress sum distribution or surface temperature change distribution) from the temperature change image generated by the temperature measurement device 1. In the first modification, the temperature measurement device 1 is configured to calculate feature data based on the pre-temperature change image and the post-temperature change image, and transmit the feature data to the defect estimation device 2.

The following describes this modified example, focusing on the differences from the first embodiment. The configuration of this modified example can also be applied to the second embodiment in the same way. When applied to the second embodiment, the surface principal stress sum distribution in this modified example can be read as the surface temperature change distribution.

<Functional configuration of the defect estimation system>
First, the functional configuration of the defect estimation system in this modified example will be described with reference to Fig. 8. Fig. 8 is a block diagram showing an example of the functional configuration of the defect estimation system in this modified example.

<Functional configuration of temperature measuring device>
8, the temperature measuring device 1 in this modification includes a temperature change unit 11 and a temperature measuring unit 12, similar to the first embodiment, and further includes a feature amount calculation unit 14. Note that the temperature measuring device 1 in this modification does not include the image generation unit 13 that the temperature measuring device 1 in the first embodiment includes.

The feature amount calculation unit 14 calculates the surface principal stress sum distribution based on two surface temperature images generated before and after the temperature change unit 11 changes the surface temperature of the CFRP laminate. The feature amount calculation unit 14 also transmits the calculated surface principal stress sum distribution to the defect estimation device 2.

<Functional configuration of the defect estimation device>
8, the defect estimation device 2 in this modification includes a position information receiving unit 22, a model learning unit 24, a model storage unit 25, a defect position estimation unit 26, and a result output unit 27, similar to the first embodiment, and includes a feature amount receiving unit 31 instead of the image receiving unit 21. It should be noted that the defect estimation device 2 in this modification does not include the feature amount calculation unit 23 that the defect estimation device 2 in the first embodiment includes.

The feature quantity receiving unit 31 receives the surface principal stress sum distribution from the temperature measuring device 1. When the feature quantity receiving unit 31 receives the surface principal stress sum distribution for the CFRP laminate to be learned, it sends the surface principal stress sum distribution to the model learning unit 24. When the feature quantity receiving unit 31 receives the surface principal stress sum distribution for the CFRP laminate to be estimated, it sends the surface principal stress sum distribution to the defect position estimation unit 26.

<Processing Procedure of Defect Estimation System>
Next, a processing procedure of a defect estimation method executed by the defect estimation system in this modified example will be described with reference to FIGS.

<How to learn>
FIG. 9 is a flowchart showing an example of a processing procedure of a learning method in this modified example.

In step S14, the feature amount calculation unit 14 included in the temperature measurement device 1 receives the image before the temperature change and the image after the temperature change from the temperature measurement unit 12. Next, the feature amount calculation unit 14 calculates the surface principal stress sum distribution of the CFRP laminate to be learned based on the image before the temperature change and the image after the temperature change. Then, the feature amount calculation unit 14 transmits the calculated surface principal stress sum distribution to the defect estimation device 2.

In step S31, the feature quantity receiving unit 31 included in the defect estimation device 2 receives the surface principal stress sum distribution for the CFRP laminate to be learned from the temperature measurement device 1. Next, the feature quantity receiving unit 31 sends the surface principal stress sum distribution received from the temperature measurement device 1 to the model learning unit 24.

<Estimation method>
FIG. 10 is a flowchart showing an example of a processing procedure of the estimation method in this modified example.

In step S14, the feature amount calculation unit 14 included in the temperature measurement device 1 receives the image before the temperature change and the image after the temperature change from the temperature measurement unit 12. Next, the feature amount calculation unit 14 calculates the surface principal stress sum distribution of the CFRP laminate to be estimated based on the image before the temperature change and the image after the temperature change. Then, the feature amount calculation unit 14 transmits the calculated surface principal stress sum distribution to the defect estimation device 2.

In step S31, the feature amount receiving unit 31 included in the defect estimation device 2 receives the surface principal stress sum distribution for the CFRP laminate to be estimated from the temperature measurement device 1. Next, the feature amount receiving unit 31 sends the surface principal stress sum distribution received from the temperature measurement device 1 to the defect position estimation unit 26.

<Effects of Modification 1>
In the defect estimation system of this embodiment, the image acquisition device calculates the surface principal stress sum distribution and transmits it to the defect estimation device. An image acquisition device capable of installing a program for calculating the surface principal stress sum distribution from infrared thermography has been realized. By combining with such an image acquisition device, the load of the defect estimation device for calculating feature data can be reduced.

[Third embodiment]
The defect estimation device 2 in the first embodiment generates learning data using temperature change images obtained by actually measuring changes in the surface temperature of a CFRP laminate to be learned. The defect estimation device 2 in the third embodiment is configured to generate learning data by numerical analysis without measuring changes in the surface temperature of a CFRP laminate.

<Functional configuration of the defect estimation system>
The functional configuration of the defect estimation system in this embodiment will be described with reference to Fig. 11. Fig. 11 is a block diagram showing an example of the functional configuration of the defect estimation system in this embodiment.

As shown in FIG. 11, the defect estimation system in this embodiment differs from the defect estimation system in the first embodiment in the following respects. First, the defect estimation device 2 does not include a position information receiving unit 22. Second, the defect estimation device 2 includes a learning data generating unit 28.

The learning data generation unit 28 calculates the surface principal stress sum distribution in each of a plurality of CFRP laminates having different internal defects by numerical analysis based on the finite element method (FEM). The learning data generation unit 28 also generates learning data by associating position information of the internal defects in each CFRP laminate with the calculated surface principal stress sum distribution. Furthermore, the learning data generation unit 28 sends the generated learning data to the model learning unit 24.

The training data generating unit 28 may include in the training data the distribution of the sum of surface principal stresses in a CFRP laminate that has no internal defects. By including training data on a CFRP laminate that has no internal defects, it is possible to improve the estimation accuracy when the internal defects are small.

Specifically, the learning data generation unit 28 calculates the surface principal stress sum distribution as follows. First, the CFRP laminate to be learned is discretized into a mesh of finite elements. Next, the element stiffness matrix and equivalent nodal force vector are calculated for each element. Next, the element stiffness matrix and equivalent nodal force vector for each element are combined into the global coordinate system. Next, simultaneous linear equations are solved to calculate the nodal displacements. Then, the strain and stress for each element are calculated from the nodal displacements.

The element stiffness matrix and global stiffness matrix can be solved by discretizing the principle of virtual work at the nodes. Also, the equivalent nodal force vector can be derived by integrating the surface stress vector over the area (see Reference 2).

[Reference 2] Onate, E. "Structural Analysis with the Finite Element Method", Artes Graficas Torres S.L., 2009.

<Effects of the Third Embodiment>
The defect estimation system in this embodiment is configured to learn an estimation model using learning data generated by numerical analysis. In order to prepare learning data by actual measurement, it is necessary to collect CFRP laminates having various internal defects and measure the temperature change for each of them with the temperature measuring device 1. Therefore, it takes a huge amount of time to collect the learning data.

By generating learning data through numerical analysis, it is possible to learn an estimation model without actually measuring the surface temperature change of the CFRP laminate. Therefore, the defect estimation device 2 in this embodiment can significantly reduce the time and financial costs required to learn an estimation model.

[Test results]
An evaluation test was conducted to evaluate the performance of the defect estimation system in one embodiment. In this test, the surface principal stress sum distribution was calculated for a CFRP laminate to be learned by numerical analysis based on the finite element method described in the third embodiment. In addition, the surface principal stress sum distribution was also calculated for a CFRP laminate to be estimated by numerical analysis based on the finite element method.

Specifically, 2,496 stack models with different positions and sizes of internal defects and one stack model with no internal defects were prepared. Of these, one stack model with no internal defects and 1,996 stack models with internal defects were used as training data, and the remaining 500 stack models were used as validation data.

The laminate model used in this experiment was a 10-layer laminate with a length of 140 mm, width of 50 mm, and thickness of 2 mm. The size of this laminate was changed within the range of length 10 to 70 mm and width 10 to 50 mm, and the center coordinates were also changed to generate a laminate model with an internal defect in any of the second to ninth layers.

Under the above test conditions, the 3D position of the internal defect was accurately estimated for 497 out of 500 sets of validation data. Even for the three sets where estimation failed, the general shape and layer number of the internal defect were able to be roughly identified.

Figure 12 shows the test results. Figure 12 (A) shows three examples of successful estimation. Figure 12 (B) shows three examples of failure. In each test result, the top row is the estimation result, and the bottom row is the correct answer. Each rectangle corresponding to channel 1 to channel 8 indicates the presence or absence of internal defects in each mesh in layers 2 to 9. Note that channel 9 is a channel added so that the sum in the stacking direction is 1. Light areas indicate the presence of internal defects, and dark areas indicate the absence of internal defects.

As shown in Figure 12 (A), it can be seen that the range and layer depth of the internal defect were accurately estimated in the successful cases. As shown in Figure 12 (B), it can be seen that even in the unsuccessful cases, the range and layer depth of the internal defect were generally estimated.

[Application example]
The defect estimation system in the above-described embodiment can be used in CFRP manufacturing facilities or maintenance and inspection facilities for products using CFRP.

In CFRP manufacturing facilities, there is a risk of defective products being produced due to the intrusion of foreign matter between the layers when laminating prepregs. Because the intrusion of foreign matter cannot be detected by appearance, 100% inspection of finished products is carried out using non-destructive testing such as ultrasonic measurement or radiography. However, as mentioned above, these inspection methods are economically and time-consuming and are one of the factors that reduce productivity.

The defect estimation system in one embodiment can be used for the primary screening of defective products. That is, the defect estimation system estimates internal defects for all finished products, and conventional non-destructive testing is then performed on only those finished products for which the location of internal defects has been estimated (in other words, those products that are highly likely to have internal defects). This makes it possible to significantly reduce the number of times that conventional non-destructive testing, which is costly, is performed.

In addition, according to the defect estimation system of one embodiment, the detailed three-dimensional position of the internal defect can be estimated, so that in conventional non-destructive testing, it is only necessary to focus on inspecting positions where there is a high possibility of an internal defect being present. Therefore, by applying the defect estimation system of one embodiment to a CFRP manufacturing facility, it becomes possible to inspect finished products for defects at low cost.

Products made using CFRP are subject to damage such as delamination, fiber breakage, and cracking of the base material due to vibrations that occur during use or the impact of foreign objects. For this reason, non-destructive testing such as ultrasonic measurement is performed on all components made of CFRP at maintenance and inspection facilities. As mentioned above, ultrasonic measurement requires a contact part, and the test results are affected by the skill of the inspection technician.

As in the case of a manufacturing facility, if a primary inspection is performed on the entire product using the defect estimation system in one embodiment, it is possible to extract locations where internal defects are likely to exist. This also allows ultrasonic inspection to focus only on locations where internal defects are likely to exist. Therefore, by applying the defect estimation system in one embodiment to a maintenance and inspection facility for products that use CFRP, it becomes possible to perform damage analysis of products at low cost.

[supplement]
Although the embodiments of the present invention have been described in detail above, the present invention is not limited to these embodiments, and various modifications and changes are possible within the scope of the gist of the present invention described in the claims.

Reference Signs List 100 Defect Estimation System 1 Temperature Measuring Device 2 Defect Estimation Device 3 User Terminal 11 Temperature Change Unit 12 Temperature Measurement Unit 13 Image Generation Unit 21 Image Receiving Unit 22 Position Information Receiving Unit 23 Feature Calculation Unit 24 Model Learning Unit 25 Model Storage Unit 26 Defect Position Estimation Unit 27 Result Output Unit 28 Learning Data Generation Unit 31 Feature Receiving Unit

Claims (10)

a feature amount receiving unit configured to receive input of feature data based on a difference between a temperature distribution before a temperature change occurs on a surface of a fiber-reinforced plastic laminate and a temperature distribution after the temperature change occurs;
a defect position estimation unit configured to estimate a three-dimensional position of the internal defect of the laminate by inputting the feature data accepted by the feature amount acceptance unit into a model that has learned a relationship between the feature data and position information representing a three-dimensional position of an internal defect;
Equipped with
The position information is information obtained by dividing each layer of the laminate into a predetermined number of meshes and indicating whether or not the internal defect exists in each of the meshes.
Defect estimation device. The defect estimation device according to claim 1 ,
The mesh is divided into orthogonal equal intervals.
Defect estimation device. The defect estimation device according to claim 1 ,
The mesh is divided finely in a region near a predetermined position, and roughly in a region different from the region near the predetermined position.
Defect estimation device. The computer
a feature amount receiving step of receiving an input of feature data based on a difference between a temperature distribution before a temperature change occurs on a surface of the fiber-reinforced plastic laminate and a temperature distribution after the temperature change occurs;
a defect position estimation step of estimating a three-dimensional position of the internal defect in the laminate by inputting the feature data received in the feature amount receiving step into a model that has learned a relationship between the feature data and position information that represents a three-dimensional position of the internal defect;
Run
The position information is information obtained by dividing each layer of the laminate into a predetermined number of meshes and indicating whether or not the internal defect exists in each of the meshes.
Defect estimation methods. On the computer,
a feature amount receiving step of receiving an input of feature data based on a difference between a temperature distribution before a temperature change occurs on a surface of the fiber-reinforced plastic laminate and a temperature distribution after the temperature change occurs;
a defect position estimation step of estimating a three-dimensional position of the internal defect in the laminate by inputting the feature data received in the feature amount receiving step into a model that has learned a relationship between the feature data and position information that represents a three-dimensional position of the internal defect;
Run the command,
The position information is information obtained by dividing each layer of the laminate into a predetermined number of meshes and indicating whether or not the internal defect exists in each of the meshes.
program. A defect estimation device that estimates an internal defect of an estimation target laminate by an estimation model generated from a learning target laminate, comprising:
A support portion that supports the laminate and applies stress to the laminate;
a temperature measuring unit for measuring the temperature before and after applying stress to the laminate and the temperature difference therebetween;
an image generating unit that generates an image of the temperature difference as a temperature change image;
a feature amount calculation unit that calculates feature data from a temperature change distribution of the temperature change image;
a model learning unit that creates an estimation model of an internal defect of the laminate from a calculation result obtained by repeatedly executing the measurement of the temperature difference, the imaging of the temperature change image, and the calculation of the feature data, and from position information indicating whether or not the internal defect exists in each of the meshes by dividing each layer of the laminate into a predetermined number of meshes;
A defect position estimation unit that estimates a three-dimensional position of an internal defect of the estimation target laminate;
Equipped with
The estimation target stack is supported on the support portion,
The temperature measuring unit measures the temperatures before and after applying stress to the estimation target laminate and the temperature difference therebetween as an estimation target laminate temperature difference;
the image generating unit visualizes the estimation target stack temperature difference as an estimation target stack temperature change image,
the feature amount calculation unit calculates the estimation target laminate feature data from the temperature change distribution of the estimation target laminate temperature change image,
the defect position estimation unit estimates a three-dimensional position of an internal defect of the estimation target laminate by inputting the estimation target laminate characteristic data to the estimation model;
Defect estimation device. The defect estimation device according to claim 6,
The feature data calculated by the feature amount calculation unit is a surface principal stress sum distribution of the laminate.
Defect estimation device. The defect estimation device according to claim 6,
The feature data calculated by the feature amount calculation unit is a surface temperature change distribution of the laminate.
Defect estimation device. The computer
applying stress to a support portion that supports the laminate;
A procedure for measuring the temperature before and after applying stress to the laminate and the temperature difference therebetween;
imaging the temperature difference as a temperature change image;
calculating feature data from a temperature change distribution of the temperature change image;
a step of creating an estimation model of an internal defect of the laminate from a calculation result obtained by repeatedly performing the measurement of the temperature difference, the imaging of the temperature change image, and the calculation of the feature data, and from position information indicating whether or not an internal defect exists in each of the meshes by dividing each layer of the laminate into a predetermined number of meshes;
A procedure of measuring the temperatures before and after applying stress to the estimation target laminate and the temperature difference therebetween as the estimation target laminate temperature difference;
A step of imaging the estimation target laminate temperature difference as an estimation target laminate temperature change image;
A step of calculating estimation target laminate characteristic data from a temperature change distribution of the estimation target laminate temperature change image;
a step of estimating a three-dimensional position of an internal defect of the estimation target laminate by inputting the estimation target laminate characteristic data into the estimation model;
A defect estimation method that performs On the computer,
applying stress to a support portion that supports the laminate;
A procedure for measuring the temperature before and after applying stress to the laminate and the temperature difference therebetween;
imaging the temperature difference as a temperature change image;
calculating feature data from a temperature change distribution of the temperature change image;
a step of creating an estimation model of an internal defect of the laminate from a calculation result obtained by repeatedly performing the measurement of the temperature difference, the imaging of the temperature change image, and the calculation of the feature data, and from position information indicating whether or not an internal defect exists in each of the meshes by dividing each layer of the laminate into a predetermined number of meshes;
A procedure of measuring the temperatures before and after applying stress to the estimation target laminate and the temperature difference therebetween as the estimation target laminate temperature difference;
A step of imaging the estimation target laminate temperature difference as an estimation target laminate temperature change image;
A step of calculating estimation target laminate characteristic data from a temperature change distribution of the estimation target laminate temperature change image;
a step of estimating a three-dimensional position of an internal defect of the estimation target laminate by inputting the estimation target laminate characteristic data into the estimation model;
A program for executing.

Images