Inspection program for a shaped body region, inspection method for a shaped body region, inspection apparatus for a shaped body region, and recording medium

Abstract

An inspection device (1) is provided with a storage unit (14) for storing a mechanical property inference model generated by machine learning based on mechanical property information and nondestructive inspection information of a first molded body region which is reinforced with reinforcing fibers and of which the mechanical property information and nondestructive inspection information are known, the mechanical property inference model taking nondestructive inspection information of a second molded body region, which is reinforced with reinforcing fibers and of which mechanical property information is unknown, as an input and inferring the mechanical property information of the second molded body region; wherein the inspection device (1) acquires the nondestructive inspection information of the second molded body region, inputs the nondestructive inspection information into the mechanical property inference model, acquires the mechanical property information of the second molded body region from the mechanical property inference model, and performs output based on the mechanical property information.

Description

Technical Field

[0001] The present invention relates to an inspection program, an inspection method, an inspection apparatus, and a recording medium for a fiber-reinforced molded body region.

Background Art

[0002] A molded body fiber-reinforced with carbon fibers can reinforce the vulnerability of the matrix resin with high-strength fibers. Therefore, it is widely adopted as a lightweight material with excellent mechanical properties. Conventionally, in the production process of fiber-reinforced composite materials, ultrasonic inspection has been performed to inspect defective products during the production of the composite materials. For example, in Patent Document 1, in the process of impregnating carbon fibers with a thermoplastic resin, the following inspection is performed. First, an ultrasonic transmitter with directivity and a receiver are opposed to each other at a certain distance from the object to be inspected (a composite material obtained by impregnating carbon fibers with a thermoplastic resin). Then, ultrasonic waves are emitted from one ultrasonic transmitter, the object to be inspected receives the ultrasonic waves with the opposed receiver, the propagation time of the ultrasonic waves is measured by a signal processing circuit, and internal defects of the object to be inspected are detected non-contact. Here, the data of the inspection using ultrasonic waves is converted into an image, and based on the image, it is possible to determine whether the object to be inspected is acceptable or not. Patent Documents 2 and 3 disclose apparatuses that automatically perform high-precision searches in order to efficiently select blood streaks and feathers in the production process of processed foods.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Patent Document 2

Patent Document 3

Summary of the Invention

Problems to be Solved by the Invention

[0004] In recent years, due to factors such as soaring raw material prices and labor costs, it has become an issue to suppress production costs while maintaining high quality. There is a demand to realize low-cost and high-precision inspection for a formed body region, which is a region obtained by forming a fiber-reinforced composite material in a mold (molding die). The sorting operation of composite materials by ultrasonic inspection described in Patent Document 1 relies on visual inspection of the obtained images. Therefore, it is difficult to grasp the state of the composite material in detail. In particular, in the pass / fail determination performed by visually inspecting the images, it is difficult to establish objective evaluation criteria, and it is difficult to determine what to use to calculate the pass / fail criteria of the material. Also, the food inspection systems described in Patent Documents 2 and 3 are devices for humans to search for the positions of bones, and are different from the technologies for inspecting formed bodies. This food inspection system merely replaces what can be judged by humans looking at images or the quality of the measurement object with a neural network.

[0005] An object of the present invention is to provide an inspection program, an inspection method, an inspection apparatus, and a recording medium for a formed body region that can estimate the mechanical properties of a formed body region obtained by forming a composite material in a mold without measuring the mechanical properties and can be used for evaluating the formed body region.

Means for Solving the Problems

[0006] The above object can be solved by the following aspects. The inspection program for the molded body region according to one aspect of the present invention is to machine-learn the mechanical property information and the non-destructive inspection information of the fiber-reinforced first molded body region where the mechanical property information and the non-destructive inspection information are known, thereby inputting the non-destructive inspection information of the fiber-reinforced second molded body region with unknown mechanical property information to generate a mechanical property estimation model for estimating the mechanical property information of the second molded body region; a step of acquiring the non-destructive inspection information of the second molded body region; and a step of inputting the acquired non-destructive inspection information into the mechanical property estimation model, obtaining the mechanical property information of the second molded body region from the mechanical property estimation model, and performing an output based on the mechanical property information. The inspection program for the molded body region causes a processor to execute the steps. The first molded body region and the second molded body region are each obtained by molding a plate-shaped composite material in a mold. The projected area of the composite material is S1, and the projected area of the portion corresponding to each of the first molded body region and the second molded body region in the mold cavity of the mold is S2. The value obtained by the calculation of (S1 / S2)×100 is defined as the charge rate. The first molded body region and the second molded body region are each obtained by molding the composite material such that the charge rate is 10% or more and 500% or less. However, the non-destructive inspection information is vibration characteristic information or acoustic characteristic information, and the sampling frequency when acquiring the non-destructive inspection information is more than 0 Hz.

[0007] The inspection method for a molded body region according to one aspect of the present invention includes: training a machine learning model with mechanical property information and non-destructive inspection information of a fiber-reinforced first molded body region, where the mechanical property information and the non-destructive inspection information are known, so as to generate a mechanical property estimation model for estimating the mechanical property information of a fiber-reinforced second molded body region with unknown mechanical property information using the non-destructive inspection information of the second molded body region as an input; a step of obtaining the non-destructive inspection information of the second molded body region; and a step of inputting the obtained non-destructive inspection information into the mechanical property estimation model, obtaining the mechanical property information of the second molded body region from the mechanical property estimation model, and performing an output based on the mechanical property information. The first molded body region and the second molded body region are each obtained by molding a plate-shaped composite material in a mold. Let the projected area of the composite material be S1, and the projected area of the portion corresponding to each of the first molded body region and the second molded body region in the mold cavity of the mold be S2. The value obtained by the calculation of (S1 / S2)×100 is defined as the charge ratio. The first molded body region and the second molded body region are each obtained by molding the composite material such that the charge ratio is 10% or more and 500% or less. However, the non-destructive inspection information is vibration characteristic information or acoustic characteristic information, and the sampling frequency when obtaining the non-destructive inspection information is more than 0 Hz.

[0008] The inspection apparatus for a molded body region according to one aspect of the present invention includes a processor that can access a model storage unit that stores a mechanical property estimation model generated by machine learning based on the mechanical property information and non-destructive inspection information of a fiber-reinforced first molded body region. The mechanical property estimation model estimates the mechanical property information of a fiber-reinforced second molded body region with unknown mechanical property information by using the non-destructive inspection information of the second molded body region as an input. The first molded body region and the second molded body region are each obtained by molding a plate-shaped composite material in a mold. Let the projected area of the composite material be S1, and the projected area of the portion corresponding to each of the first molded body region and the second molded body region in the mold cavity of the mold be S2. The value obtained by the calculation of (S1 / S2)×100 is defined as the charge ratio. The first molded body region and the second molded body region are each obtained by molding the composite material such that the charge ratio is 10% or more and 500% or less. The processor acquires the non-destructive inspection information of the second molded body region, inputs the non-destructive inspection information into the mechanical property estimation model, acquires the mechanical property information of the second molded body region from the mechanical property estimation model, and performs an output based on the mechanical property information. However, the non-destructive inspection information is vibration characteristic information or acoustic characteristic information, and the sampling frequency when acquiring the non-destructive inspection information is more than 0 Hz.

[0009] The inspection program for the molded body region according to one aspect of the present invention is to make a machine learning of the mechanical property information and the temperature distribution information of the fiber-reinforced first molded body region where the mechanical property information and the temperature distribution information are known, and input the temperature distribution information of the fiber-reinforced second molded body region where the mechanical property information is unknown, and generate a mechanical property estimation model for estimating the mechanical property information of the second molded body region; a step of acquiring the temperature distribution information of the second molded body region; and a step of inputting the acquired temperature distribution information into the mechanical property estimation model, acquiring the mechanical property information of the second molded body region from the mechanical property estimation model, and performing an output based on the mechanical property information, and is an inspection program for the molded body region that causes a processor to execute, wherein the first molded body region and the second molded body region are each obtained by molding a plate-shaped composite material in a mold, the projected area of the composite material is S1, the projected area of the portion corresponding to each of the first molded body region and the second molded body region in the mold cavity of the mold is S2, and the value obtained by the calculation of (S1 / S2)×100 is defined as the charge ratio, and the first molded body region and the second molded body region are each obtained by molding the composite material so that the charge ratio is 10% or more and 500% or less.

Effect of the Invention

[0010] According to the present invention, it is possible to estimate the mechanical properties of the molded body region only by non-destructive inspection information without measuring the mechanical properties, and use it for the evaluation of the molded body region. According to the present invention, it is possible to instantaneously estimate with high accuracy the mechanical property information that cannot be estimated by a person no matter how hard they try from the non-destructive inspection information. As a result, the waste loss in the production of the molded body including the molded body region can be reduced, and a high-quality molded body can be provided at low cost.

Brief Description of the Drawings

[0011]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8A

Figure 8B

Figure 8C

Figure 8D

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Figure 16

Figure 17

Figure 18A

Figure 18B

Embodiments for Carrying Out the Invention

[0012] Hereinafter, a test system including a test apparatus according to an embodiment of the present invention will be described, but the present invention is not limited thereto. [Overview of the Test System] The test system of this embodiment uses, as an object to be inspected, a molded body region (second molded body region) having a predetermined shape (for example, plate-like) reinforced with fibers whose mechanical properties are unknown, and estimates the mechanical properties of this second molded body region without actually measuring them. Here, the molded body region is a region obtained by molding a composite material containing reinforcing fibers using a mold.

[0013] The mechanical property information of the molded body region is information indicating the mechanical physical properties of the molded body region, and is, for example, information related to fracture or elasticity such as the strength of the molded body region. Examples of the mechanical property information include information related to fracture strength such as tensile strength or bending strength, and information related to the elastic modulus related to each of fracture strength, compressive strength, or shear strength.

[0014] The information related to the elastic modulus may be the elastic modulus itself (for example, tensile elastic modulus or bending elastic modulus), or may be the rank (hereinafter referred to as mechanical property rank) when the elastic modulus (for example, tensile elastic modulus or bending elastic modulus) is ranked. The information related to the elastic modulus preferably further includes any one of information indicating that the elastic modulus (or its rank) corresponds to a defective product, information indicating that the elastic modulus (or its rank) corresponds to a non-defective product, and information indicating that it was difficult to estimate the elastic modulus (or its rank).

[0015] Information on the breaking strength may be the breaking strength itself (e.g., tensile strength or yield strength), or the rank when the breaking strength (e.g., tensile strength or yield strength) is ranked (hereinafter referred to as the mechanical property rank). The information on the breaking strength preferably further includes any one of information indicating that the breaking strength (or its rank) corresponds to a defective product, information indicating that the breaking strength (or its rank) corresponds to a non-defective product, and information indicating that it was difficult to estimate the breaking strength (or its rank).

[0016] The computer included in the inspection system acquires the non-destructive inspection information of the second molded body region, inputs this non-destructive inspection information into a mechanical property estimation model that was previously generated and stored in the model storage unit, performs estimation by this mechanical property estimation model, and outputs the estimated mechanical property information of the second molded body region or information indicating that estimation was difficult. Examples of the estimated mechanical property information of the second molded body region include the mechanical property rank, information indicating whether it is a non-defective product or a defective product, and the like. Examples of the method of outputting the information include displaying the information on a display unit, streaming the information as a message from a speaker, causing the information to be printed by a printer, and the like.

[0017] In the present invention, the non-destructive inspection information is, for example, vibration characteristic information or temperature distribution information. The vibration characteristics are information on the molded body region itself, such as displacement, acceleration, and natural frequency, obtained by vibrating the molded body region in a fixed state or a non-fixed state. The vibration characteristic information may be numerical data such as the natural frequency, or may be a vibration inspection image or an acoustic characteristic image using a wavelet image. As the vibration inspection image, for example, the wavelet images of FIGS. 8A to 8D can be exemplified. Here, the wavelet image is a two-dimensional image in which the frequency of vibration is shown on the vertical axis, time is shown on the horizontal axis, and the amplitude of vibration is shown in shades after wavelet transformation of the vibration of the molded body region. The temperature distribution information is information indicating the temperature distribution of the molded body region to be inspected. As the temperature distribution information, for example, an image (thermography) obtained by imaging the molded body region with an infrared camera can be used.

[0018] The mechanical property prediction model is a model that outputs mechanical property information with non-destructive inspection information as input, which is generated by subjecting data of mechanical property information and non-destructive inspection information in a fiber-reinforced molded body region (first molded body region) where the data is known to machine learning (including supervised learning or deep learning of unsupervised learning). For the mechanical property prediction model, for example, a neural network or a support vector machine is used.

[0019] The computer of the inspection system constitutes the inspection device. This computer may include a processor, a storage unit composed of a device capable of storing information such as a hard disk device or an SSD (Solid State Drive), and a RAM (Random Access Memory) and a ROM (Read Only Memory). By executing the inspection program stored in the ROM, this processor performs processes such as acquisition of non-destructive inspection information of the molded body region of the object to be inspected, input of the acquired non-destructive inspection information to the mechanical property prediction model, acquisition of mechanical property information from the mechanical property prediction model, and output based on the acquired mechanical property information. The storage medium for storing the inspection program is not limited to the ROM, and a known storage medium can be used, but it is preferably a non-temporary storage medium, and a hard disk device or an SSD (Solid State Drive) may be used. The storage medium is sometimes referred to as a recording medium. Also, the inspection program may be stored in a server (cloud) on the network, and the processor may download the program from the server and execute the program.

[0020] The non-destructive inspection information of the molded body region is usually used to determine whether there are defects, voids, or foreign substances inside the molded body region, and if so, to what extent the degree of existence is. However, even if there are many defects, voids, or foreign substances inside the molded body region, depending on the distribution state of the defects, voids, or foreign substances, the mechanical properties may be good in some cases. In such a case, if the non-destructive inspection information is visually confirmed and it is determined that the molded body region is a defective product because there are many defects, voids, or foreign substances, the molded body region that should have been a good product will be discarded, resulting in a decrease in production efficiency. On the other hand, the opposite can also occur. That is, even if the non-destructive inspection information is visually confirmed and it is determined that the product is a good product because there are few defects, voids, or foreign substances, the mechanical properties may correspond to those of a defective product in some cases.

[0021] As a result of verification based on the above viewpoints, the inventors of the present invention found that there is a correlation between non-destructive inspection information and mechanical property information, and by machine learning a large number of measured data of non-destructive inspection information and mechanical property information into a model such as a neural network or a support vector machine, they successfully inferred the mechanical property information of the molded body region with high accuracy from the non-destructive inspection information of the molded body region. Obtaining mechanical property information from non-destructive inspection information has not been considered conventionally. Therefore, it has not been easy for those skilled in the art to construct a machine learning model that outputs mechanical property information with non-destructive inspection information as input. Hereinafter, a detailed example of the inspection system will be described. In the following, an example in which the mechanical property inference model is a neural network will be described.

[0022] [Reinforcing fiber] The type of reinforcing fiber used in the present invention can be appropriately selected according to the use of the molded body region a (corresponding to the second molded body region with unknown mechanical property information) which is the object to be inspected, and is not particularly limited. As the reinforcing fiber, either inorganic fiber or organic fiber can be preferably used. Examples of the inorganic fiber include carbon fiber, activated carbon fiber, graphite fiber, glass fiber, tungsten carbide fiber, silicon carbide fiber (silicon carbide fiber), ceramic fiber, alumina fiber, natural mineral fiber (such as basalt fiber), boron fiber, boron nitride fiber, boron carbide fiber, and metal fiber.

[0023] [Carbon fiber] When carbon fiber is used as the fiber, generally known carbon fibers include polyacrylonitrile (PAN)-based carbon fiber, petroleum and coal pitch-based carbon fiber, rayon-based carbon fiber, cellulose-based carbon fiber, lignin-based carbon fiber, phenol-based carbon fiber, and vapor-grown carbon fiber. In the present invention, any of these carbon fibers can be preferably used.

[0024] [Form of reinforcing fiber] In the present invention, the form of the reinforcing fiber is not particularly limited. Hereinafter, the continuous fiber, which is a specific example conducted by the present inventors, will be described. However, the present invention is not limited to continuous fiber. The continuous fiber means a reinforcing fiber in which a bundle of reinforcing fibers is aligned continuously without cutting the reinforcing fiber into a short fiber state. For the purpose of obtaining a molded body region a having excellent mechanical properties, it is preferable to use continuous reinforcing fibers. More specifically, the continuous fiber is preferably a fiber having a length of 1 m or more, and after being processed into a woven fabric such as a woven fabric or a knitted fabric, it is impregnated with a resin by hand lay-up or the like, or used as a prepreg impregnated with an uncured resin in the continuous fiber.

[0025] [Molded body region a] The molded body region a is reinforced with reinforcing fibers. Hereinafter, an example of the embodiment conducted by the present inventors will be described. However, the present invention is not limited to the molded body region a described below. 1. Molded body The molded body region a is a molded body after molding a plate-shaped composite material, and may be a molded body using a thermoplastic resin or a molded body using a thermosetting prepreg. A prepreg is a material for creating a molded article. It is made by arranging continuous carbon fibers in one direction to form a sheet (unidirectional prepreg), impregnating a base material made of carbon fibers such as a carbon fiber fabric with a thermosetting resin, or a molding intermediate material in which a part of the thermosetting resin is impregnated and the remaining part is arranged on at least one surface.

[0026] 2. Unidirectional material The molded article region a is preferably a unidirectional material. A unidirectional material refers to a material in which continuous reinforcing fibers with a length of 100 mm or more are arranged in one direction inside the molded article region a. The unidirectional material may be a laminate of a plurality of continuous reinforcing fibers. In particular, when the molded article region a is a unidirectional material and the molded article is made using a thermosetting prepreg, the influence on the mechanical properties due to the fiber orientation is small. Therefore, the accuracy of estimating the mechanical property information by the model described later can be improved.

[0027] [Preferred molded article region] The molded article region contains reinforcing fibers and a matrix resin as essential components, and other components as optional components. It is preferable that the porosity Vr of the molded article region obtained by the following formulas (A) and (B) is 10% or less. Vr = (t2 - t1) / t2 × 100 ··· Formula (A) t1 = (Wf / Df + Wm / Dm + Wz / Dz) ÷ unit area (mm 2 ) ··· Formula (B) t1: Theoretical thickness of the molded article region (mm) t2: Measured thickness of the molded article region (mm) Df: Density of the reinforcing fiber (mg / mm 3 ) Dm: Density of the matrix resin (mg / mm 3 ) Dz: Density of other components (mg / mm 3 ) Wf: Mass of the reinforcing fiber (mg) Wm: Mass of the matrix resin (mg) Wz: Mass of other components (mg) The void ratio Vr is more preferably 5% or less, and even more preferably 3% or less. If the void ratio is within this range, the accuracy of predicting the mechanical properties of the present invention is improved.

[0028] [Manufacture of the molded body region a] For example, the molded body region a can be prepared as follows. 1. Materials · Reinforcing fiber: Carbon fiber "Tenax (registered trademark)" STS40 - 24K (tensile strength 4,300 MPa, tensile modulus 240 GPa, number of filaments 24,000, fineness 1,600 tex, elongation 1.8%, density 1.78 g / cm 3 , manufactured by Teijin Limited) · Base resin: A thermosetting resin composition mainly composed of an epoxy resin

[0029] 2. Preparation of unidirectional prepreg The unidirectional prepreg was prepared by the hot melt method as follows. First, the above thermosetting resin composition was applied onto a release paper using a coater to produce a resin film. Next, the above carbon fiber bundle was fed out from a creel, passed through a comb to align the pitch between the carbon fiber bundles, then widened through a fiber opening bar, and aligned in one direction so as to form a sheet-like shape with a fiber mass per unit area (fiber basis weight) of 100 g / m 2 . Then, the above resin film was overlaid on both sides of the carbon fiber, heated and pressurized to impregnate the thermosetting resin composition, and wound up with a winder to produce a unidirectional prepreg. The resin content of the obtained unidirectional prepreg was set to 30 wt.%.

[0030] 3. Preparation of the molded body region a Eleven unidirectional prepregs were manually laminated in the 0° direction to obtain a prepreg laminate with a laminate structure of

[0011] T . The above prepreg laminate was placed in a bag film, arranged in a mold, and the temperature was raised in an autoclave, heated at 130 °C for 120 minutes, and cured to produce a CFRP molded body (a molded body region a which is a unidirectional carbon fiber reinforced thermosetting resin molded body) with a thickness of 1 mm. The charge ratio described below during autoclave molding was 100%.

[0031] [Measurement of Tensile Elastic Modulus and Tensile Strength] As specific examples of the breaking strength or elastic modulus of the present invention, the inventors measured the tensile elastic modulus and tensile strength of the molded body region a as described below. The CFRP molded body was processed into a test piece shape (length 250 mm × width 15 mm) by a water jet, and tabs made of a glass fiber-reinforced resin matrix composite material were adhered. In accordance with ASTM D3039 method, a tensile test in the 0° direction was carried out using a universal testing machine under the condition of a test speed of 2 mm / min, and the tensile elastic modulus and tensile strength of the CFRP molded body (molded body region a) were calculated.

[0032] [Vibration Characteristic Information] In the present invention, the non-destructive inspection information is preferably vibration characteristic information. There is no particular limitation on the vibration inspection method used for obtaining the vibration characteristic information, and any inspection method that can detect internal defects, voids, or foreign matters in the molded body region without destroying the molded body region may be used. Further, the finite element method (FEA) may be used for obtaining the vibration characteristic information. The vibration characteristic information is preferably an image converted from the information obtained by the vibration inspection or the finite element method, and it is particularly preferable that the converted image is a wavelet image.

[0033] Specific wavelet images are shown in FIGS. 8A to 8D. In obtaining this image, the vibration measurement was carried out by impulse excitation under free-free boundary conditions. A hole with a diameter of Φ2 mm was opened in the molded body, and the free-free boundary condition was achieved by suspending the molded body from a beam through a nylon tegus (22-8231 manufactured by Takagi Tsuna Kogyo Co., Ltd.) inserted into this hole. An acceleration pickup sensor (356A01 manufactured by PCB PIEZOTRONICS) was installed in the molded body region, and excitation was carried out using an impulse hammer (GK-3100 manufactured by Ono Sokki Co., Ltd.). Data measurement and analysis were performed using a real-time acoustic vibration analysis system (DS-3000 manufactured by Ono Sokki Co., Ltd.). At this time, the sampling frequency was set to 2000 Hz.

[0034] The obtained vibration data was analyzed using wavelet analysis software (BIOMAS manufactured by ELMEC), and a wavelet image was obtained with the vibration frequency on the vertical axis, time on the horizontal axis, and the vibration amplitude represented by brightness or hue. Figures 8A and 8B are wavelet images of the molded body region without defects, and Figures 8C and 8D are wavelet images of the molded body region with defects. Note that Figures 8A to 8D are binary images, and the more white parts there are, the larger the vibration amplitude at that frequency. The reference numeral 803 in Figure 8C and the reference numeral 804 in Figure 8D are spherical patterns, indicating that vibrations at 450 Hz occurred intermittently. It was confirmed that the portions of the reference numeral 803 in Figure 8C and the reference numeral 804 in Figure 8D are clearly different from the portions of the reference numeral 801 in Figure 8A and the reference numeral 802 in Figure 8B.

[0035] [Vibration attenuation of vibration characteristics] The image preferably contains information on vibration attenuation. Vibration attenuation is a vibration phenomenon in which the amplitude decreases with time in the time history data of vibration. For example, looking at the patterns of the reference numeral 801 in Figure 8A and the reference numeral 802 in Figure 8B, the length in the vertical axis direction of the white pattern at approximately 450 Hz in Figures 8A and 8B becomes smaller with the passage of time, indicating that the vibration at approximately 450 Hz is attenuating. At this time, in the reference numeral 801 of Figure 8A, the speed at which the vibration at approximately 450 Hz attenuates is faster than that of the reference numeral 802 in Figure 8B. In the molded body regions where the measurement data of Figures 8A to 8D were measured, the faster the vibration at approximately 450 Hz attenuated, the smaller the tensile elastic modulus and the tensile strength. Note that Figures 8A to 8D are merely examples. Depending on the shape and natural vibration frequency of the test piece, the frequency at which vibration attenuation occurs is different. Also, for mechanical properties other than the tensile elastic modulus and the tensile strength, it is not necessarily the case that the smaller the speed at which the vibration attenuates, the smaller the value.

[0036] [Acoustic characteristics] The non-destructive inspection information in the present invention may be acoustic characteristic information. [Frequency] 1. Sampling frequency Generally, the sampling frequency refers to the frequency of taking samples per unit time in sampling, which is a process necessary to convert an analog waveform such as audio into digital data. In order to correctly sample a certain waveform, it is necessary to sample at a frequency that is at least twice the bandwidth of the frequency components of the waveform. In the present invention, the non-destructive inspection information is vibration characteristic information or acoustic characteristic information, and the sampling frequency when acquiring the non-destructive inspection information of the second molded body region is greater than 0 Hz. A preferable value of the lower limit is 50 Hz or more, a more preferable value is 250 Hz or more, still more preferably 5,000 Hz or more, and even more preferably 10,000 Hz or more. On the other hand, a preferable value of the upper limit is 50,000 Hz or less, a more preferable value is 40,000 Hz or less, and a still more preferable value is 30,000 Hz or less. Therefore, the sampling frequency is preferably 50 Hz or more and 50,000 Hz or less, more preferably 250 Hz or more and 50,000 Hz or less, still more preferably 5,000 Hz or more and 40,000 Hz or less, and even more preferably 10,000 Hz or more and 30,000 Hz or less. From another perspective, the sampling frequency is preferably at least twice the frequency at which the above-described vibration attenuation occurs. Also, from another perspective, the sampling frequency when acquiring the vibration characteristics of the second molded body region is preferably at least twice the natural frequency of the primary mode described later, and more preferably at least twice the natural frequency of the secondary mode. 2. Natural Frequency of the Primary Mode When the non-destructive inspection information is vibration characteristic information, the natural frequency of the primary mode of the first molded body region and the second molded body region is preferably greater than 0 Hz and less than or equal to 1,000 Hz. Within this range, for example, when the molded body is assembled in an automobile, the comfort is enhanced without resonance with vibrations from the outside or the engine room. A more preferable natural frequency of the primary mode of the first molded body region and the second molded body region is greater than 0 Hz and less than or equal to 500 Hz. On the one hand, when the non-destructive inspection information is acoustic characteristic information, the natural frequencies of the primary modes of the first molded body region and the second molded body region are preferably more than 0 Hz and 20,000 Hz or less, and more preferably more than 0 Hz and 10,000 Hz or less.

[0037] [Finite Element Method] The mechanical property information or non-destructive inspection information of the first molded body region is preferably obtained by the finite element method. Here, the finite element method (Finite Element Method, FEM) is one of the numerical analysis methods, and it is possible to numerically obtain an approximate solution of a differential equation that is difficult to solve analytically. By creating an analysis model in which material parameters previously identified by comparing actual measurement and analysis are applied to the shape of the first molded body region, these information can be obtained by the finite element method without actually measuring the mechanical properties or non-destructive inspection of the first molded body region.

[0038] [Temperature Distribution Image] The temperature distribution information is preferably a temperature distribution image. The temperature distribution image refers to an image obtained by imaging the temperature distribution of the molded body region a, which is the object to be inspected, with an infrared camera. Specific temperature distribution images are shown in FIGS. 18A and 18B. FIG. 18A is an active thermography image of a molded body region including a "void". FIG. 18B is an active thermography image of a molded body region formed by fluid molding a sheet molding compound in a mold. In obtaining these images, in the measurement by active thermography, the molded body region a, which is the object to be inspected, was fixed with a clamp, and two halogen lamps (CHP500 manufactured by Caster) were respectively installed so that heat could be irradiated from an angle of 45° to the left and right with respect to the surface of the molded body region a. The heat irradiated from these two halogen lamps is reflected on the surface of the molded body region a, and this reflected heat was measured with an infrared camera (A615 manufactured by FLIR Systems). The heat source frequency was 0.1 Hz, the frame rate was 25 Hz, and the shooting time was 125 seconds. It was confirmed that the defect in the molded body region was visualized at reference numeral 1801 in FIG. 18A. When such a defect exists, the mechanical properties such as the breaking strength and elastic modulus tend to decrease in the defective part of the molded body region.

[0039] A sheet molding compound is obtained by impregnating a chopped fiber bundle mat formed by depositing chopped fiber bundles obtained by cutting fiber bundles of continuous reinforcing fibers in a sheet shape with a thermosetting resin as a matrix resin. When a molded body region is formed by fluid molding a sheet molding compound in a mold, the orientation and distribution of the chopped fiber bundles change according to the flow in the mold, and the mechanical properties of the molded body region may change according to the orientation and distribution of the chopped fiber bundles.

[0040] The chopped fiber bundles contained in the sheet molding compound and the matrix resin have different thermal conductivities. Therefore, in the molded body region formed from the sheet molding compound, the temperature distribution may change according to the chopped fiber bundles distributed in the matrix resin. Therefore, as shown in FIG. 18B, the temperature distribution image of the molded body region formed from the sheet molding compound has a complex pattern according to the distribution and orientation of the chopped fiber bundles.

[0041] As described above, since the temperature distribution image and the mechanical properties can change according to the orientation and distribution of the chopped fiber bundles contained in the sheet molding compound, there is a correlation between the temperature distribution image and the mechanical properties of the molded body region.

[0042] It is clear that no matter how hard a person tries, the mechanical property information of the molded body region cannot be inferred from the temperature distribution images of FIGS. 18A and 18B. In the inspection apparatus 1 of the present embodiment, a mechanical property estimation model is generated using teacher data in which the temperature distribution image of the first molded body region is associated with the mechanical property information obtained by measuring the first molded body region, which is the source of the temperature distribution image. By using the inspection apparatus 1 of the present embodiment, the mechanical property information of the second molded body region can be inferred from the temperature distribution image of the second molded body region. Therefore, the mechanical property information that even a skilled worker cannot infer no matter how hard they try can be instantly inferred without actual measurement.

[0043] The method for acquiring the temperature distribution image is not particularly limited. For example, infrared thermography can be used. Infrared thermography is a non-destructive inspection method for knowing the internal situation from the surface temperature and temperature change of an object. When measuring an object that does not generate heat by exciting it from the outside, it is called active thermography. The excitation method used in infrared thermography is not particularly limited, and it may be heating or cooling, but it is preferable to heat from the outside and capture an infrared thermography image. For heating, radiant heat, vibration energy by ultrasonic waves, electromagnetic force, sunlight, etc. can be used, and for cooling, means such as natural cooling, heat of vaporization, and blowing of low-temperature gas can be used. A temperature change occurs on the heated or cooled surface, and the internal situation can be estimated by detecting the temperature change with an infrared sensor. When heating a molded body region containing fibers and a thermoplastic resin, the heating temperature is preferably below the softening point of the thermoplastic resin. If it is heated and observed below the softening point, it is less likely to have an adverse effect on the molded body.

[0044] [Inspection System] Hereinafter, the data converted into a format that can be input to the input layer of the neural network is described as input data. In the inspection system, the non-destructive inspection information of the sample in the first molded body region a (hereinafter referred to as the molded body region sample b) (hereinafter referred to as the non-destructive inspection information sample) and the mechanical property information actually measured from the molded body region sample b (hereinafter referred to as the mechanical property information sample) are acquired and used as the second input data, and the neural network is trained using this second input data. When the learning of the neural network is completed, the non-destructive inspection information of the molded body region a is input to the neural network as the first input data, and the mechanical property information of the molded body region a is inferred based on the reaction value in the output layer from the neural network. Based on the inferred mechanical property information, the molded body region a may be sorted into good products and defective products.

[0045] In order for the inspection system to perform effective learning and highly accurate inference, non-destructive inspection information optimized for the learning process and the inference process can be used. For example, various image processes may be performed on the vibration inspection image so that it is easy to detect feature amounts from the vibration inspection image, the acoustic characteristic image, or the temperature distribution image.

[0046] [Inspection Device] FIG. 1 is a block diagram showing a configuration example of the inspection device 1. The inspection device 1 performs image processing, generation of input data, learning of the neural network, inference of mechanical property information using the neural network, and the like. The inspection device 1 includes one or more processors configured by a CPU (Central Processing Unit) or the like, a storage unit, and a communication unit, and is an information processing device such as a computer on which an OS (Operating System) and an application operate. The inspection device 1 may be a physical computer, or may be realized by a virtual machine (VM), a container, or a combination thereof. More specifically, the structure of the processor is an electric circuit combining circuit elements such as semiconductor elements.

[0047] The inspection device 1 includes an image storage unit 11 that stores non-destructive inspection information and non-destructive inspection information samples, a processing unit 12 that processes non-destructive inspection information and non-destructive inspection information samples, an input data generation unit 13, a teacher data storage unit 14, a learning unit 15, a model storage unit 16, an inference unit 17, a display unit 18, and an operation unit 19. The processing unit 12, the input data generation unit 13, the learning unit 15, and the inference unit 17 are functional blocks realized by the processor of the inspection device 1 executing a program. This program includes an inspection program for the molded body region.

[0048] The image storage unit 11 is a storage area for storing non-destructive inspection information (preferably a vibration inspection image, an acoustic characteristic image, or a temperature distribution image). The image storage unit 11 may be a volatile memory such as SRAM (Static Random Access Memory) or DRAM (Dynamic Random Access Memory), or a non-volatile memory such as NAND type flash memory, magnetoresistive random access memory (MRAM), or ferroelectric random access memory (FeRAM).

[0049] The processing unit 12 performs image processing on non-destructive inspection information (preferably a vibration inspection image, an acoustic characteristic image, or a temperature distribution image), and stores the image after the image processing in the image storage unit 11. Examples of image processing include generating an image in which the luminance of each of the red, green, and blue (RGB) colors of the pixels in the image is extracted, generating an image in which the luminance of green (G) is subtracted from the luminance of red (R) of each pixel, generating an image in which only the red component is extracted after conversion to the HSV color space, etc., but other types of image processing may also be performed.

[0050] The processing unit 12 may also perform, among other things, image enlargement, reduction, cropping, noise removal, rotation, inversion, color depth change, contrast adjustment, brightness adjustment, sharpness adjustment, color correction, etc.

[0051] The input data generation unit 13 generates input data that is input to the input layer of the neural network from the non-destructive inspection information or non-destructive inspection information sample stored in the image storage unit 11. For example, when performing learning described later using a vibration inspection image, it is preferable to cut out a desired part from the vibration inspection image or remove an extra part to obtain the second input data.

[0052] When the inspection device 1 is executing a learning process, the input data generation unit 13 stores the input data in the teacher data storage unit 14. When the inspection device 1 is inspecting the molded body area a, the input data is transferred to the estimation unit 17. Note that when performing the learning process, the input data generation unit 13 may generate input data using, for example, an image (non-destructive inspection information sample) captured by an external device or system.

[0053] The teacher data storage unit 14 is a storage area that stores a plurality of input data used for the learning of the neural network. The input data stored in the teacher data storage unit 14 is used as the teacher data of the learning unit 15. The input data (second input data) used as the teacher data is stored in association with the mechanical property information sample obtained by measuring from the molded body area sample b that is the acquisition source of the input data. This mechanical property information sample preferably includes, in addition to at least one of the mechanical property rank and mechanical property value (for example, elastic modulus) of the molded body area sample b, information indicating that the mechanical property value corresponds to a good product, information indicating that the mechanical property value corresponds to a defective product, information indicating that it is difficult to estimate the mechanical property value, and the like.

[0054] For example, the association of the mechanical property information sample with the second input data obtained from the molded body area sample b (hereinafter, this association is also referred to as labeling) can directly input the mechanical property value (for example, elastic modulus) of the molded body area sample b by the user operating the operation unit 19. After the input, the inspection device 1 classifies the mechanical property value into mechanical property ranks. For example, the tensile elastic modulus can be divided into the following mechanical property ranks. Mechanical property rank 1: The tensile elastic modulus in the molded body region is 30 GPa or more Mechanical property rank 2: The tensile elastic modulus in the molded body region is 25 - 30 GPa Mechanical property rank 3: The tensile elastic modulus in the molded body region is 25 GPa or less This mechanical property rank is not in units of 5 GPa as described above, and it may output the mechanical property rank in units of 3 GPa or 1 GPa. If the mechanical property information sample corresponding to the second input data obtained from the molded body region sample b is known, the mechanical property rank can be automatically labeled by a program, script, etc., rather than by user operation. The labeling of the mechanical property rank may be performed before the conversion of the non-destructive inspection information sample obtained from the molded body region sample b to the second input data, or after the conversion to the second input data.

[0055] The learning unit 15 uses the input data (second input data) stored in the teacher data storage unit 14 to perform neural network learning. The learning unit 15 stores the learned neural network in the model storage unit 16. The learning unit 15 can learn, for example, a three-layer neural network including an input layer, a hidden layer, and an output layer. By learning a three-layer neural network, it is possible to ensure real-time response performance during the inspection of the molded body region a. The number of units included in each of the input layer, the hidden layer, and the output layer is not particularly limited. The number of units included in each layer can be determined based on the required response performance, the object to be inferred, the discrimination performance, etc. Note that the three-layer neural network is an example, and it does not prevent the use of a multi-layer neural network with a larger number of layers. When using a multi-layer neural network, various neural networks such as a convolutional neural network can be used.

[0056] The model storage unit 16 is a storage area that stores the neural network learned by the learning unit 15. A plurality of neural networks may be stored in the model storage unit 16 according to the type of the molded body area a to be inspected. Since the model storage unit 16 is set to be referable by the inference unit 17, the inference unit 17 can perform the inspection (inference of mechanical property information) of the molded body area a using the neural network stored in the model storage unit 16. The model storage unit 16 may be a volatile memory such as a RAM or a DRAM, or a non-volatile memory such as a NAND, an MRAM, or a FRAM (registered trademark). Note that the model storage unit 16 only needs to be in a location accessible by the processor of the inspection device 1 and does not have to be built into the inspection device 1. For example, the model storage unit 16 may be an external storage attached to the inspection device 1, or a network storage connected to a network accessible from the inspection device 1.

[0057] The inference unit 17 uses the neural network stored in the model storage unit 16 to infer the mechanical property information of the molded body area a. The inference unit 17 infers the mechanical property rank of the molded body area a based on the response value output from the unit of the output layer. Examples of the units of the output layer include a unit of mechanical property rank 1, a unit of mechanical property rank 2, a unit of mechanical property rank 3, a unit of difficult to infer, etc., but other types of units may be prepared. For example, those with a low inferred mechanical property rank may have a high possibility of containing a lot of foreign substances, etc. The mechanical property rank of the molded body area a may be inferred using the difference or ratio of the response values of multiple units.

[0058] The display unit 18 is a display that displays images and texts. The display unit 18 may display the captured image, the image after image processing, and the inference result by the inference unit 17. The operation unit 19 is a device that provides means for the user to operate the inspection device 1. The operation unit 19 is, for example, a keyboard, a mouse, a button, a switch, a voice recognition device, etc., but is not limited thereto.

[0059] [Learning process] Before estimating the mechanical property rank of the molded body region a by the inspection device 1, it is necessary to train a neural network using non-destructive inspection information samples and mechanical property information samples of the molded body region sample b of the same type as the molded body region a. FIG. 2 is a flowchart of the training process. First, the processor of the inspection device 1 acquires non-destructive information samples of each of the plurality of molded body region samples b (step S201). The non-destructive inspection information samples here should include those with high and low mechanical property ranks. When providing a unit that outputs a reaction value with difficult estimation to the output layer of the neural network, non-destructive inspection information samples with difficult-to-estimate mechanical property information may be prepared. Examples of non-destructive inspection information samples with difficult estimation include, when the non-destructive inspection information sample is an image, an image in which the molded body region sample b is not sufficiently shown, an image in which the brightness adjustment due to illumination or exposure is inappropriate and the molded body region sample b is not clearly shown, and the like.

[0060] The processor of the inspection device 1 generates second input data from each of the acquired non-destructive inspection information samples (step S201). Next, the processor of the inspection device 1 acquires mechanical property information samples of each of the plurality of molded body region samples b, and stores the acquired mechanical property information samples in association with the respective second input data (step S203). Note that step S203 may be performed before step S202. In this case, even after each non-destructive inspection information sample is converted into second input data, the mechanical property information sample associated with the non-destructive inspection information sample may be carried over to the second input data.

[0061] Next, the processor of the inspection device 1 starts training by the neural network based on the second input data (step S204). Figure 3 shows an example of a neural network that outputs three reaction values. The neural network 301 in Figure 3 is a neural network having three layers: an input layer 302, a hidden layer 303, and an output layer 304. The output layer 304 includes units 311, 312, and 313 that estimate the mechanical property rank. Although there are three units 311, 312, and 313 in Figure 3, the number can be appropriately increased or decreased according to the rank of the mechanical properties.

[0062] In a neural network, the value input to the input layer is propagated through the hidden layer and the output layer, and the reaction value of the output layer is obtained. In step S205, when the second input data is input to the neural network, the number of hidden layers 303, the number of units included in each of the input layer 302 and the hidden layer 303, and the coupling coefficient between the units included in each of the input layer 302 and the hidden layer 303 are adjusted so that the mechanical property information sample associated with this second input data or information close to it is output from the neural network with a high probability. In this way, a mechanical property estimation model is generated and stored in the model storage unit 16.

[0063] Figure 4 shows the arithmetic processing between the units of the neural network. Figure 4 shows the unit of the (m - 1)-th layer and the unit of the m-th layer. For the sake of explanation, it is assumed that only a part of the units of the neural network is shown in Figure 4. The unit numbers in the (m - 1)-th layer are k = 1, 2, 3, ···. The unit numbers in the m-th layer are j = 1, 2, 3, ···. Let the reaction value of the unit number k in the (m - 1)-th layer be a k m-1 Then, the reaction value a j m of the unit number j in the m-th layer is obtained using the following formula (2).

[0064]

Equation

[0065] Here, Wjk m is the weight and indicates the strength of the connection between units. b j m is the bias. f(···) is the activation function. From Equation (2), it can be seen that the response value of any unit in the m-th layer is the output value when the response values of all units (k = 1, 2, 3, ···) in the (m - 1)-th layer are weighted and added and input as the variable of the activation function. Next, an example of the activation function will be described. The following Equation (3) is the normal distribution function.

[0066]

Equation

[0067] Here, μ is the mean value and indicates the center position of the bell-shaped peak drawn by the normal distribution function. σ is the standard deviation and indicates the width of the peak. Since the value of Equation (3) depends only on the distance from the center of the peak, it can be said that the Gaussian function (normal distribution function) is a type of radial basis function (RBF). The Gaussian function (normal distribution function) is just an example, and other RBFs can also be used. The following Equation (4) is the sigmoid function. The sigmoid function approaches 1.0 in the limit as x → ∞. Also, it approaches 0.0 in the limit as x → -∞. That is, the sigmoid function takes values in the range (0.0, 1.0).

[0068]

Equation

[0069] Note that using functions other than the Gaussian function and the sigmoid function as the activation function is not prohibited. For example, the inventors used Relu in the convolutional layer and softmax in the output layer. For the learning of the neural network, after inputting the input data into the input layer, the weight W, which is the strength of the connection between units, is adjusted so that the correct output can be obtained.jk Adjustment is performed. The correct output (the response value of the units in the output layer) expected when inputting input data labeled with a certain mechanical property rank in a neural network is also called a teacher signal. For example, when inputting input data labeled with a mechanical property rank of 311 into the neural network 301, in the teacher signal, the response value of unit 311 is 1, the response value of unit 312 is 0, and the response value of unit 313 is 0. When inputting input data labeled with a mechanical property rank of 312 into the neural network 301, in the teacher signal, the response value of unit 311 is 0, the response value of unit 312 is 1, and the response value of unit 313 is 0. For example, the weight W jk Adjustment can be performed using the backpropagation method (error backpropagation method: Back Propagation Method). In the backpropagation method, in order to reduce the deviation between the output of the neural network 310 and the teacher signal, starting from the output layer side in order, the weight W jk is adjusted. The following formula (5) shows the improved backpropagation method.

[0070]

Number

[0071] When using a Gaussian function as the activation function, not only the weight W jk but also σ and μ in formula (3) are targets for adjustment as parameters in the improved backpropagation method. By adjusting the values of the parameters σ and μ, the learning convergence of the neural network is assisted. The following formula (6) shows the value adjustment process for the parameter σ.

[0072]

Number

[0073] Here, t is the number of learning times, η is the learning constant, δ k is the generalization error, O j is the response value of unit number j, α is the sensitivity constant, and β is the oscillation constant. ΔW jk , Δσ jk , Δμ jk are the correction amounts of the weights W jk , σ, and μ, respectively. Here, taking the improved backpropagation method as an example, the adjustment process of the weights W jk and parameters has been described. However, the general backpropagation method can also be used instead. In the following, when simply referred to as the backpropagation method, it is assumed to include both the improved backpropagation method and the general backpropagation method. The number of times of adjusting the weights W jk and parameters by the backpropagation method may be once or multiple times, and is not particularly limited. Generally, based on the estimation accuracy of the mechanical property rank when using test data, it is possible to determine whether to repeat the adjustment of the weights W jk and parameters by the backpropagation method. Repeating the adjustment of the weights W jk and parameters may improve the estimation accuracy of the mechanical property rank. By using the above method, in step S205, the values of the weights W jk , the parameters σ, and μ can be determined. Once the values of the weights W jk , the parameters σ, and μ are determined, it becomes possible to perform the estimation process using the neural network.

[0074] FIG. 5 is a flowchart for explaining the inference operation of mechanical property information by an inspection apparatus 1 that operates according to an inspection program for a molded body region. The processor of the inspection apparatus 1 acquires non-destructive inspection information of the molded body region a (preferably, captures a vibration inspection image, an acoustic characteristic image, or a temperature distribution image) (step S501). When using the vibration inspection image, the acoustic characteristic image, or the temperature distribution image as the non-destructive inspection information, there may be a step of performing image processing on the vibration inspection image, the acoustic characteristic image, or the temperature distribution image between step S501 and step S502.

[0075] Next, the processor of the inspection apparatus 1 generates first input data from the non-destructive inspection information (step S502). The first input data has N elements equal to the number of units in the input layer of the neural network and is in a format that can be input to the neural network.

[0076] Next, the processor of the inspection apparatus 1 inputs the first input data to the neural network (step S503). The first input data is transmitted in the order of the input layer, the hidden layer, and the output layer. The processor of the inspection apparatus 1 performs an inference of the mechanical property rank based on the response value in the output layer of the neural network (step S504).

[0077] The inference process using the neural network is equivalent to the process of finding the position in the discrimination space of the first input data. FIG. 6 shows an example of the discrimination space when using a Gaussian function as the activation function. When using an RBF such as a Gaussian function as the activation function, the discrimination surface that divides the discrimination space into regions for each rank of mechanical properties becomes a closed surface. Also, for each category of the mechanical property rank, by adding an index in the height direction, the regions related to each category in the discrimination space can be localized.

[0078] FIG. 7 shows an example of the discrimination space when the sigmoid function is used as the activation function. When the activation function is the sigmoid function, the discrimination surface is an open surface. Note that the above-described learning process of the neural network corresponds to the process of learning the discrimination surface in the discrimination space. Only the mechanical property rank 311 and the mechanical property rank 312 are shown in the regions in FIGS. 6 and 7, but there may be a distribution of three or more mechanical property ranks.

[0079] As described above, by using the inspection system of the present embodiment, the mechanical property information of the molded body region a can be inferred from the non-destructive inspection information (preferably, a vibration inspection image, an acoustic characteristic image, or a temperature distribution image) of the molded body region a. In the inventions described in Patent Document 2 (International Publication No. 2019 / 151393) and Patent Document 3 (International Publication No. 2019 / 151394), what can be judged by a human by simply looking at an image or a measurement object is merely replaced by a neural network. That is, in these inventions, since the inspection target is a food that has been photographed, a human can easily judge the presence or absence of a foreign object or the like in the food. On the other hand, the mechanical property information is a numerical value or a rank equivalent thereto, and the non-destructive inspection information (preferably, a vibration inspection image, an acoustic characteristic image, or a temperature distribution image) visualizes or quantifies the internal state of the molded body region. That is, even if a skilled worker looks at the non-destructive inspection information, it is impossible to infer the mechanical property information therefrom. For example, it is obvious that no matter how much a human tries, the mechanical property information cannot be inferred from the vibration inspection images in FIGS. 8A to 8D and the temperature distribution images in FIGS. 18A to 18B. By using the inspection apparatus 1 of the present embodiment, it is possible to instantaneously infer the mechanical property information that cannot be inferred by a skilled worker no matter how much they try, without actually measuring it.

[0080] In the above embodiment, an example of inferring mechanical property information from the non-destructive inspection information of a plate-shaped molded body region has been described. However, the present invention is not limited to a plate-shaped molded body region, and mechanical property information can also be inferred from the non-destructive inspection information for a molded body region having a three-dimensional shape. Hereinafter, an example of forming a plate-shaped composite material in a mold to obtain a formed body region will be described. Hereinafter, the plate-shaped composite material before being formed in the mold will be collectively referred to as composite material MX. In the following description, as the composite material MX, composite material M, composite material Ms, composite material M2, and composite material M3 are exemplified. The composite material MX preferably contains discontinuous fibers. The composite material MX may contain a thermoplastic resin as the matrix resin. The composite material MX is particularly preferably a sheet molding compound in which a chopped fiber bundle obtained by cutting a fiber bundle of continuous reinforcing fibers is deposited in a sheet shape and impregnated with a thermosetting resin as the matrix resin.

[0081] [Forming method of composite material MX] The forming method of the composite material MX is not particularly limited, and press molding (compression molding), autoclave molding, vacuum molding, etc. are used, but press molding is preferred.

[0082] [Mold] FIG. 9 is a side view schematically showing an example of a mold used for press molding of the composite material MX. FIG. 10 is a schematic plan view of the fixed mold 20 in the mold shown in FIG. 9 as viewed from the movable mold 30 side. The mold shown in FIG. 9 includes a fixed mold 20 and a movable mold 30 configured to be movable relative to the fixed mold 20. A recess 22 is formed on the upper surface 21 of the fixed mold 20 on the side of the movable mold 30. The recess 22 is a region defined by a bottom surface 22A and a pair of side surfaces 22B connecting the bottom surface 22A and the upper surface 21. The movable mold 30 is configured to be movable in a direction D including a direction D1 approaching the fixed mold 20 and a direction D2 moving away from the fixed mold 20. The direction D1 is a direction in which pressure is applied to the composite material MX when forming the composite material MX. In addition, if the forming method is vacuum molding, the direction D1 becomes the suction direction. In the state where the movable mold 30 is closest to the fixed mold 20 (the state where the movable mold 30 is at the position of the dashed line in the figure), a mold cavity (the space SP shown by the oblique lines in the figure) is formed between the movable mold 30 and the fixed mold 20. A forming cavity is a space for forming a formed body. Among the formed bodies obtained by forming the composite material MX using a mold, unnecessary portions may be trimmed at the ends or the like. In this case, among the spaces SP shown in FIG. 9, the space that forms the formed body that is trimmed and finally remains becomes the forming cavity. For example, assume a case where, among the formed body MD1 shown in FIG. 13 obtained by forming the composite material M using the mold shown in FIG. 9, the portion indicated by the dashed line is trimmed to obtain a final product. In this case, among the space SP, the portion excluding the range indicated by the dashed line becomes the forming cavity.

[0083] FIG. 11 is a schematic diagram showing an example of a method for forming a single composite material M using the mold shown in FIG. 9. The plate-shaped composite material M is disposed in a state covering the recess 22 on the upper surface 21 of the fixed mold 20. FIG. 12 is a schematic plan view of the fixed mold 20 and the composite material M shown in FIG. 11 as viewed in the direction D1. When the movable mold 30 is moved in the direction D1 from the state shown in FIG. 11, the composite material M flows due to the pressure from the movable mold 30 and deforms along the shape of the space SP, and as shown in FIG. 13, the formed body MD1 is obtained. In the examples of FIGS. 11 to 13, the entire formed body region obtained by forming the composite material M constitutes the formed body MD1. When the composite material M contains a thermoplastic resin as a matrix resin, the composite material M may be formed by cold pressing. When the forming method is cold pressing, the composite material M is heated and preformed before pressing. The preformed composite material M is disposed on the upper surface and the recess of the fixed mold 20 as shown in FIG. 14. Thereafter, when the movable mold 30 is moved in the direction D1 from the state shown in FIG. 14, the preformed composite material M flows due to the pressure from the movable mold 30 and deforms along the shape of the space SP, and as shown in FIG. 13, the formed body MD1 is obtained.

[0084] When manufacturing a molded body using a molding die, the composite material MX that is the basis of the molded body may be one or a plurality. For example, as shown by the dashed line in FIGS. 11 and 12, there may be a case where another plate-shaped composite material Ms is further arranged on the composite material M and then molding is performed. Also, as shown in FIG. 15, there may be a case where a plurality of pre-shaped plate-shaped composite materials (composite material M2 and composite material M3) are arranged on the fixed die 20 so as not to overlap each other and then molding is performed. When the movable die 30 is moved in the direction D1 from the state shown in FIG. 15, the pre-shaped composite materials M2 and M3 flow due to the pressure from the movable die 30 and deform along the shape of the space SP. Then, as shown in FIG. 16, a molded body MD2 composed of a molded body region MA2 obtained by molding the composite material M2 and a molded body region MA3 obtained by molding the composite material M3 is obtained. In the examples of FIGS. 15 and 16, the molded body region MA2 obtained by molding the composite material M2 and the molded body region MA3 obtained by molding the composite material M3 each constitute a part of the molded body MD2.

[0085] [Projected area S1 of the composite material MX constituting the molded body] (1) When the molded body is composed of a single composite material MX, the planar area of the composite material MX when viewed in the thickness direction of the single composite material MX (in the case of adopting cold press molding, the state before pre-shaping) is defined as the projected area S1. In the example of FIG. 11, the planar area of the composite material M is the projected area S1.

[0086] (2) Assume a case where the molded body is composed of a plurality of composite materials MX and is molded in a state where these plurality of composite materials MX are stacked. In this case, the molded body is regarded as being composed of a single molded body region. Then, the total planar area of the plurality of composite materials MX when viewed in their respective thickness directions (in the case of cold press molding, the state before pre-shaping) is defined as the projected area S1.

[0087] (i) In the example of FIG. 12, the whole of the composite material Ms overlaps the composite material M. Therefore, the projected area S1 is the same as the planar area of the composite material M.

[0088] (ii) Assume a case where, for example, in FIG. 12, only a part of the composite material Ms is arranged so as to overlap the composite material M. In this case, the sum of the planar area of the composite material M and the planar area of the region in the composite material Ms that does not overlap the composite material M becomes the projected area S1.

[0089] (3) Assume a case where the molded body is composed of a plurality of composite materials MX, and after arranging them so that the plurality of composite materials MX do not overlap, they are molded. In this case, when each of the plurality of composite materials MX (in the case of cold press molding, in the state before preforming) is viewed in its respective thickness direction, the planar area of each composite material MX is defined as the projected area S1. In the example of FIG. 15, the planar area of the composite material M2 when the composite material M2 before preforming is viewed in the thickness direction becomes the projected area S1. Also, the planar area of the composite material M3 when the composite material M3 before preforming is viewed in the thickness direction becomes the projected area S1.

[0090] [Projected area S2 of the portion corresponding to the molded body region in the mold cavity] Assume a case where the molded body formed by the mold cavity is composed of (1) a single composite material MX, or (2) is regarded as being composed of a single molded body region obtained by molding a plurality of overlapping composite materials MX. In this case, the entire mold cavity (space SP) becomes the portion corresponding to this single molded body region. The planar area of this portion when viewed in the direction D1 is defined as the projected area S2. In the example of FIG. 13, the planar area of the space SP when viewed in the direction D1 becomes the projected area S2.

[0091] (3) Assume a case where the molded body formed by the mold cavity is composed of a plurality of molded body regions obtained by molding a plurality of composite materials MX arranged in a non-overlapping state. In this case, the portion of the mold cavity where each molded body region exists is the portion corresponding to each molded body region. The planar area of each of these portions when viewed in the direction D1 is defined as the projected area S2.

[0092] In the example of FIG. 16, in the space SP, when looking at the portion where the molded body region MA2 exists in the direction D1, the planar area is the projected area S2 of the molded body region MA2. Further, in the space SP, when looking at the portion where the molded body region MA3 exists in the direction D1, the planar area is the projected area S2 of the molded body region MA3. In addition, when a plurality of composite materials MX are arranged apart from each other and cold pressed to form a molded body, during molding, each composite material MX flows and a weld is formed at the boundary. Therefore, by observing the completed molded body, the "molded body region obtained by molding each composite material MX" can be easily discriminated.

[0093] [Charge rate] For a molded body manufactured by a mold, the value obtained by the calculation of the following formula (C) is defined as the charge rate. Charge rate [%] = 100 × {(projected area S1 of the composite material MX constituting the molded body) / (projected area S2 of the portion corresponding to the molded body region in the mold cavity)} ··· (C) In addition, as in the example shown in FIG. 16, when the molded body MD2 includes a plurality of molded body regions MA2 and MA3, by substituting the projected area S1 and the projected area S2 for each molded body region MA2 and MA3 into the formula (C), two charge rates can be obtained for one molded body MD2. That is, by substituting the planar area of the composite material M2 as the projected area S1 and the planar area of the portion where the molded body region MA2 exists in the mold cavity as the projected area S2 of the molded body region MA2 into the formula (C), the first charge rate is calculated. Further, by substituting the planar area of the composite material M3 as the projected area S1 and the planar area of the portion where the molded body region MA3 exists in the mold cavity as the projected area S2 of the molded body region MA3 into the formula (C), the second charge rate is calculated.

[0094] The higher this charging rate value is, the shorter the flow distance of the composite material MX during molding. For example, taking the case of manufacturing only the molded body region MA2 shown in FIG. 16 using the mold shown in FIG. 9 as a product. In this case, as shown in FIG. 17, when the projected area S1 of the composite material M2 is made smaller than the example shown in FIG. 15 to lower the charging rate, the flow distance of the composite material M2 becomes longer than the example shown in FIG. 15. Here, the "flow distance" can be represented by the distance from the end of the composite material arranged in the mold to the end of the molded body region after molding the composite material.

[0095] When the composite material MX contains discontinuous fibers and the molded body region obtained by molding this composite material MX is fiber-reinforced by the discontinuous fibers, when flowing the composite material MX to generate the molded body region, the fibers are oriented in the flow direction simultaneously with the flow of the composite material MX. Therefore, when the flow distance of the composite material MX is short, it becomes easier to control the flow and fiber orientation during molding, and the mechanical property quality of the molded body region when mass-produced becomes stable. Here, the "mechanical property quality" is the quality indicating the mechanical properties of the molded body region. Examples of the mechanical properties include breaking strengths such as tensile strength, flexural strength, and shear strength, and elastic moduli related to each of tensile elastic modulus, flexural elastic modulus, breaking strength, compressive strength, or shear strength.

[0096] By generating the above-described mechanical property estimation model using the teacher data obtained from the molded body region with as little variation in mechanical property quality as possible, the estimation accuracy of the mechanical properties can be improved. Specifically, as the first molded body region that is the measurement source of the teacher data (non-destructive inspection information and mechanical property information) to be learned by the mechanical property estimation model, and the second molded body region that is the target for estimating the mechanical properties by this mechanical property estimation model, it is preferable to apply those obtained by molding the composite material MX in a mold so that the charging rate is 10% or more and 500% or less.

[0097] As the structure of each of the first formed body region and the second formed body region, for example, the formed body MD1 shown in FIG. 13, the formed body region MA2 shown in FIG. 16, the formed body region MA3 shown in FIG. 16, or the formed body MD2 shown in FIG. 16 can be adopted. When each of the first formed body region and the second formed body region is the formed body MD2 shown in FIG. 16, either one of the charge rate obtained for the composite material M2 and the formed body region MA2 and the charge rate obtained for the composite material M3 and the formed body region MA3 may be 10% or more and 500% or less. When the charge rate is less than 10%, the flow distance of the composite material MX increases. For this reason, it becomes difficult to stabilize the mechanical property quality of each of the first formed body region and the second formed body region. Therefore, the lower limit of the charge rate is set to 10%. The upper limit of the charge rate can be determined according to the use of the product manufactured by the mold. If this upper limit is set to 500%, products suitable for many uses can be targeted. In order to further stabilize the mechanical property quality of each of the first formed body region and the second formed body region, it is preferable that the charge rate is 50% or more, more preferably 70% or more, and still more preferably 80% or more. When the reinforcing fiber is a discontinuous fiber and it is desired to improve the mechanical strength in a specific direction, it is preferable that the charge rate is 10% or more and less than 50% because anisotropy is likely to occur in the formed body region.

[0098] [Isotropy of the formed body region] When the flow of the composite material MX during molding is small, the fiber orientation of the composite material MX and the formed body region after molding thereof is approximate. In particular, in the case of an isotropic base material in which the composite material MX is reinforced with discontinuous fibers, if the flow during molding is small (in other words, if the charge rate is large), isotropy is ensured also in the formed body region. Here, the isotropy can be evaluated by measuring the tensile elastic modulus in an arbitrary direction of the formed body region and a direction orthogonal thereto, and taking the ratio of the larger value to the smaller value among the measured tensile elastic moduli. The isotropy (ratio of tensile elastic modulus) of each of the first formed body region and the second formed body region is preferably 1.5 or less, and more preferably 1.3 or less.

[0099] [Coefficient of Variation (CV) of Areal Density of Formed Body Region] The coefficient of variation (CV) of the areal density of each of the first formed body region and the second formed body region is preferably 10% or less. Here, the areal density is the mass of the formed body region per unit area. If the coefficient of variation (CV) of the areal density is 10% or less, the physical properties of each of the first formed body region and the second formed body region are made uniform. For this reason, it becomes easy to estimate the mechanical properties by the mechanical property estimation model.

[0100] [Reinforcing Fibers Contained in Formed Body Region] When the fiber length of the reinforcing fiber bundle is Li, the single fiber diameter of the reinforcing fiber constituting the reinforcing fiber bundle is Di, and the number of single fibers contained in the reinforcing fiber bundle is Ni, for each of the first formed body region and the second formed body region, Li is 1 mm or more and 100 mm or less, and Li / (Ni×Di 2 ) is 8.0×10 1 or more and 3.3×10 3 or less, it is preferable to contain the reinforcing fiber bundle A. The volume ratio of the reinforcing fiber bundle A is preferably 50 to 100 vol%, more preferably 70 to 90 vol%, based on the total reinforcing fibers contained in the formed body.

[0101] This application claims priority based on Japanese Patent Application No. 2021-210611 filed on December 24, 2021.

Claims

1. By causing a processor to machine-learn the mechanical property information and the non-destructive inspection information of a first fiber-reinforced formed body region where the mechanical property information and the non-destructive inspection information are known, the mechanical property information of a second fiber-reinforced formed body region where the mechanical property information is unknown is used as an input to generate a mechanical property estimation model for estimating the mechanical property information of the second formed body region; obtaining non-destructive inspection information of the second formed body region; inputting the obtained non-destructive inspection information into the mechanical property estimation model, obtaining the mechanical property information of the second formed body region from the mechanical property estimation model, and performing an output based on the mechanical property information, which is an inspection program for a formed body region for causing a processor to execute; the first formed body region and the second formed body region are each obtained by molding a plate-shaped composite material using a mold; wherein the projected area of the composite material is S1, the projected area of a portion corresponding to each of the first formed body region and the second formed body region in the mold cavity of the mold is S2, and a value obtained by an operation of (S1 / S2)×100 is defined as a charge ratio; the first formed body region and the second formed body region are each obtained by molding the composite material such that the charge ratio is 10% or more and 500% or less; the mechanical property information is information related to tensile strength or bending strength, and information related to an elastic modulus related to each of fracture strength, compressive strength, or shear strength; the non-destructive inspection information is vibration characteristic information or acoustic characteristic information, and the sampling frequency when obtaining the non-destructive inspection information is more than 0 Hz, which is an inspection program for a formed body region.

2. An inspection program for a formed body region according to Claim 1, wherein the first formed body region and the second formed body region are each obtained by molding the composite material such that the charge ratio is 50% or more.

3. An inspection program for a formed body region according to Claim 1, wherein the first formed body region and the second formed body region are each fiber-reinforced by discontinuous fibers.

4. An inspection program for a formed body region according to Claim 1, wherein the first formed body region and the second formed body region each have an isotropy of 1.5 or less.

5. An inspection program for a formed body region according to Claim 1, The inspection program for the formed body region, wherein the first formed body region and the second formed body region each have a coefficient of variation of basis weight of 10% or less.

6. The inspection program for the formed body region according to Claim 1, For each of the first formed body region and the second formed body region, the reinforcing fiber has a fiber length of the reinforcing fiber bundle of Li, a single fiber diameter of the reinforcing fiber constituting the reinforcing fiber bundle of Di, and a number of single fibers included in the reinforcing fiber bundle of Ni. When Li is 1 mm or more and 100 mm or less, and Li / (Ni × Di ２ ), it contains a reinforcing fiber bundle that is 8.0 × 10 １ or more and 3.3 × 10 ３ or less. An inspection program for the formed body region.

7. The inspection program for the formed body region according to Claim 1, The inspection program for the formed body region, wherein the forming is press forming.

8. The inspection program for the formed body region according to Claim 1, The inspection program for the formed body region, wherein the composite material is a sheet molding compound in which a chopped fiber bundle mat is impregnated with a thermosetting resin as a matrix resin.

9. The inspection program for the formed body region according to Claim 1, The inspection program for the formed body region, wherein the non-destructive inspection information is an image or numerical data.

10. The inspection program for the formed body region according to Claim 1, The inspection program for the formed body region, wherein the mechanical property information is information regarding the elastic modulus or the fracture strength of the formed body region.

11. The inspection program for the formed body region according to Claim 10, The information regarding the elastic modulus includes the elastic modulus or the rank when the elastic modulus is ranked, The inspection program for the formed body region, wherein the information regarding the fracture strength includes the fracture strength or the rank when the fracture strength is ranked.

12. The inspection program for the formed body region according to Claim 11, The inspection program for the formed body region, wherein the information regarding the elastic modulus or the fracture strength includes at least one of information indicating that the elastic modulus or the fracture strength is difficult to estimate, information indicating that the elastic modulus or the fracture strength corresponds to a defective product, and information indicating that the elastic modulus or the fracture strength corresponds to a non-defective product.

13. The inspection program for the formed body region according to any one of Claims 1 to 12, The inspection program, wherein the mechanical property information or the non-destructive inspection information of the first formed body region includes at least one piece of information obtained by the finite element method.

14. The inspection program for the formed body region according to Claim 9, The inspection program, wherein the image includes a waveform of vibration attenuation.

15. The inspection program for the formed body region according to any one of Claims 1 to 12, The inspection program, wherein the non-destructive inspection information is vibration characteristic information.

16. The inspection program for the formed body region according to Claim 15, An inspection program in which the sampling frequency is 250 Hz or more and 50,000 Hz or less

17. An inspection program for the molded body region according to claim 15, wherein The natural frequencies of the primary modes of the first molded body region and the second molded body region are inspection programs that are more than 0 Hz and 1000 Hz or less.

18. An inspection program for the molded body region according to any one of claims 1 to 12, wherein The non-destructive inspection information is acoustic characteristic information.

19. An inspection program for the molded body region according to claim 18, wherein The natural frequencies of the primary modes of the first molded body region and the second molded body region are inspection programs that are more than 0 Hz and 20,000 Hz or less.

20. By machine-learning the mechanical property information and the non-destructive inspection information of the fiber-reinforced first molded body region, where the mechanical property information and the non-destructive inspection information are known, a step of generating a mechanical property estimation model for estimating the mechanical property information of the fiber-reinforced second molded body region with the non-destructive inspection information of the second molded body region having unknown mechanical property information as an input; A step of obtaining non-destructive inspection information of the second molded body region; A step of inputting the obtained non-destructive inspection information into the mechanical property estimation model, obtaining the mechanical property information of the second molded body region from the mechanical property estimation model, and performing an output based on the mechanical property information, and The first molded body region and the second molded body region are each obtained by molding a plate-shaped composite material in a mold, Let the projected area of the composite material be S1, and the projected area of the portions corresponding to each of the first molded body region and the second molded body region in the mold cavity of the mold be S2. The value obtained by the calculation of (S1 / S2) × 100 is defined as the charge ratio, The first molded body region and the second molded body region are each obtained by molding the composite material so that the charge ratio is 10% or more and 500% or less, The mechanical property information is information related to tensile strength or bending strength, and information related to the elastic modulus related to each of the fracture strength, compressive strength, or shear strength, The non-destructive inspection information is vibration characteristic information or acoustic characteristic information, and the sampling frequency when obtaining the non-destructive inspection information is more than 0 Hz. A method for inspecting a molded body region.

21. A processor accessible to a model storage unit that stores a mechanical property estimation model generated by machine learning based on the mechanical property information and non-destructive inspection information of a fiber-reinforced first molded body region. The mechanical property estimation model estimates the mechanical property information of the fiber-reinforced second molded body region by using, as an input, the non-destructive inspection information of the fiber-reinforced second molded body region with unknown mechanical property information. The first molded body region and the second molded body region are each obtained by molding a plate-shaped composite material using a mold. Let the projected area of the composite material be S1, and the projected area of the portion corresponding to each of the first molded body region and the second molded body region in the mold cavity of the mold be S2. A value obtained by the calculation of (S1 / S2) × 100 is defined as the charge ratio. The first molded body region and the second molded body region are each obtained by molding the composite material such that the charge ratio is 10% or more and 500% or less. The processor acquires the non-destructive inspection information of the second molded body region, inputs the non-destructive inspection information into the mechanical property estimation model, acquires the mechanical property information of the second molded body region from the mechanical property estimation model, and performs an output based on the mechanical property information. The mechanical property information is information related to tensile strength or bending strength, and information related to the elastic modulus related to each of the fracture strength, compressive strength, or shear strength. The non-destructive inspection information is vibration characteristic information or acoustic characteristic information, and is an inspection device for a molded body region in which the sampling frequency when acquiring the non-destructive inspection information is more than 0 Hz.

22. A step of generating a mechanical property estimation model that estimates the mechanical property information of a fiber-reinforced second molded body region by inputting, as an input, the temperature distribution information of the fiber-reinforced second molded body region with unknown mechanical property information, by performing machine learning on the mechanical property information and the temperature distribution information of a fiber-reinforced first molded body region where the mechanical property information and the temperature distribution information are known. A step of acquiring the temperature distribution information of the second molded body region. A step of inputting the acquired temperature distribution information into the mechanical property estimation model, acquiring the mechanical property information of the second molded body region from the mechanical property estimation model, and performing an output based on the mechanical property information, which is an inspection program for a molded body region that causes a processor to execute. The first molded body region and the second molded body region are each obtained by molding a plate-shaped composite material using a mold. The mechanical property information is information regarding tensile strength or flexural strength, and information regarding the elastic modulus related to each of the fracture strength, compressive strength, or shear strength. Let S1 be the projected area of the composite material, and let S2 be the projected area of the portions corresponding to each of the first formed body region and the second formed body region in the mold cavity of the mold. The value obtained by the calculation of (S1 / S2) × 100 is defined as the charge ratio. The inspection program for the formed body regions, wherein the first formed body region and the second formed body region are each obtained by molding the composite material such that the charge ratio is 10% or more and 500% or less.

23. The inspection program according to claim 22, wherein the temperature distribution information is a temperature distribution image.

24. The inspection program according to claim 22, wherein the temperature distribution information is an active thermography image.

25. A recording medium storing the inspection program according to any one of claims 1 to 12.

Images