PREDICTION SYSTEM, PREDICTION METHOD, AND NON-TRANSITORY STORAGE MEDIUM

Abstract

A prediction system configured to predict a defect of a target product includes a first pre-trained model trained based on a defect characteristic value indicating a defect associated with a location in an existing product, a feature of a three-dimensional shape of the existing product, and conditional information indicating a manufacturing condition of the existing product. The first pre-trained model is configured to, when a feature of a three-dimensional shape of the target product is input, output a defect characteristic value indicating a defect associated with a location in the target product.

Claims

1. A prediction system configured to predict a defect of a target product, the prediction system comprising a first pre-trained model trained based on a defect characteristic value indicating a defect associated with a location in an existing product, a feature of a three-dimensional shape of the existing product, and conditional information indicating a manufacturing condition of the existing product, wherein the first pre-trained model is configured to, when a feature of a three-dimensional shape of the target product is input, output a defect characteristic value indicating a defect associated with a location in the target product.

2. The prediction system according to claim 1, wherein: the product is a casting; and the defect of the product, indicated by the defect characteristic value, includes at least one of seizure, shrinkage, flow line, galling, raw material deformation, die cracking, or entrapment of the product.

3. The prediction system according to claim 1, wherein the defect characteristic value includes a value indicating a degree of the defect of the product.

4. The prediction system according to claim 1, wherein, when the product is a casting, the first pre-trained model is further trained by at least one of a die volume of the casting, a casting volume, a casting surface area, or a thickness of the casting.

5. The prediction system according to claim 1, further comprising a second pre-trained model configured to, when shape information indicating the three-dimensional shape of the existing product is input, output the feature of the three-dimensional shape of the existing product, wherein the first pre-trained model is trained by using the feature output from the second pre-trained model.

6. The prediction system according to claim 1, wherein, when the product is a casting, the manufacturing condition includes at least one of molten metal type, molten metal temperature, internal cooling temperature, water flow time, die temperature, die surface treatment, cycle time, die time, die opening sequence, spray application amount, spray time, or air blow sequence.

7. The prediction system according to claim 1, further comprising a display device, wherein the defect characteristic value indicating the defect associated with the location in the target product is displayed on the display device.

8. A prediction method of predicting a defect of a target product, the prediction method comprising inputting, by a computer, a feature of a three-dimensional shape of the target product to a first pre-trained model, trained based on a defect characteristic value indicating a defect associated with a location in an existing product, shape information indicating a three-dimensional shape of the existing product, and conditional information indicating a manufacturing condition of the existing product, and causing the first pre-trained model to output a defect characteristic value indicating a defect associated with a location in the target product.

9. The prediction method according to claim 8, wherein, when the product is a casting, the defect of the product, indicated by the defect characteristic value, includes at least one of seizure, shrinkage, flow line, galling, raw material deformation, die cracking, or entrapment of the product.

10. The prediction method according to claim 8, wherein the defect characteristic value includes a value indicating a degree of the defect of the product.

11. The prediction method according to claim 8, wherein, when the product is a casting, the first pre-trained model is further trained by at least one of a die volume of the casting, a casting volume, a casting surface area, or a thickness of the casting.

12. The prediction method according to claim 8, further comprising: inputting, by the computer, the shape information indicating the three-dimensional shape of the existing product to a second pre-trained model and causing the second pre-trained model to output a feature of the three-dimensional shape of the existing product; and training, by the computer, the first pre-trained model by using the feature output from the second pre-trained model.

13. The prediction method according to claim 8, wherein, when the product is a casting, the manufacturing condition includes at least one of molten metal type, molten metal temperature, internal cooling temperature, water flow time, die temperature, die surface treatment, cycle time, die time, die opening sequence, spray application amount, spray time, or air blow sequence.

14. The prediction method according to claim 8, further comprising displaying the defect characteristic value indicating the defect associated with the location in the target product on a display device.

15. A non-transitory storage medium storing a program that is a pre-trained model configured to predict a defect of a target product, wherein: the pre-trained model is trained based on a defect characteristic value indicating a defect associated with a location in an existing product, a feature of a three-dimensional shape of the existing product, and conditional information indicating a manufacturing condition of the existing product; and the pre-trained model is configured to, when a feature of a three-dimensional shape of the target product is input, output a defect characteristic value indicating a defect associated with a location in the target product.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Features, advantages, and technical and industrial significance of exemplary embodiments of the disclosure will be described below with reference to the accompanying drawings, in which like signs denote like elements, and wherein:

[0018] FIG. 1 is a block diagram showing the configuration of a prediction apparatus according to an embodiment of the disclosure;

[0019] FIG. 2 is a flowchart showing an example of a process to be executed by the prediction apparatus according to the embodiment of the disclosure;

[0020] FIG. 3 is a flowchart showing another example of a process to be executed by the prediction apparatus according to the embodiment of the disclosure; and

[0021] FIG. 4 is a view showing an example of an image showing a prediction result according to the embodiment of the disclosure.

DETAILED DESCRIPTION OF EMBODIMENTS

[0022] Hereinafter, an embodiment of the disclosure will be described with reference to the accompanying drawings. FIG. 1 is a block diagram showing the configuration of a prediction apparatus 10 according to the embodiment of the disclosure. The prediction apparatus 10 predicts a defect of a target product. Specific examples of the prediction apparatus 10 include information processing apparatuses, such as a server and a personal computer (PC); however, the prediction apparatus 10 is not limited thereto. The prediction apparatus 10 may be regarded as a prediction system. Examples of the target product include castings for use in vehicles, such as automobiles.

[0023] The prediction apparatus 10 includes an arithmetic unit 100, a storage device 110, and a display device 120. The arithmetic unit 100 is an arithmetic unit, such as a central processing unit (CPU) and a micro processing unit (MPU). The arithmetic unit 100 executes a prediction method of predicting a defect of a target product by reading and running a program saved in the storage device 110.

[0024] The storage device 110 is a storage device in which the program to be run by the arithmetic unit 100 and various data such as information on an existing product and information on the target product are saved. Specifically, the information on the existing product includes a defect characteristic value indicating a defect associated with a location in the existing product, shape information indicating the three-dimensional shape of the existing product, and conditional information indicating the manufacturing condition of the existing product. The defect characteristic value and shape information of the existing product are obtained by means of, for example, the computer-aided engineering (CAE) analysis of the existing product. The information on the target product includes shape information indicating the three-dimensional shape of the target product and conditional information indicating the manufacturing condition of the target product.

[0025] When the product is a casting, the defect of the product, indicated by the defect characteristic value, may be a general casting defect. Specific examples of the casting defect include seizure, shrinkage, flow line, galling, raw material deformation, die cracking, and entrapment. The raw material deformation means an undesirable deformation that can occur when a casting is cooled to an ordinary temperature after the casting is formed in a casting process. The defect of the product, indicated by the defect characteristic value, is not limited to these examples.

[0026] The defect characteristic value is a quantitative variable and is a value indicating the degree of a defect of the product. The defect characteristic value indicates the degree of a defect of the product by its magnitude.

[0027] When the product is a casting, the manufacturing condition indicated by the conditional information may be a condition that is set in a general casting process. Specific examples of the manufacturing condition indicated by the conditional information include molten metal type, molten metal temperature, internal cooling temperature, water flow time, die temperature, die surface treatment, cycle time, die time, die opening sequence, spray application amount, spray time, and air blow sequence in order to manufacture a casting. The manufacturing condition is not limited to these examples.

[0028] The molten metal type is a type of molten metal. The molten metal temperature is the temperature of molten metal. The internal cooling temperature is the temperature of water that passes through the inside of a die for cooling a casting. The die temperature is the temperature of a die at the time of forming a casting. The water flow time is a time taken by water to pass through the inside of the die. The die surface treatment is a heat treatment or the like that is carried out to reduce abrasion of a die surface.

[0029] The cycle time is a time required in a casting process at the time when castings are continuously produced. The cycle of the casting process consists of die closing, pouring, solidification, die opening, removal of a casting, mold release agent application, air blow, and die closing.

[0030] The die time is a time included in a cycle time and is a time from solidification, that is, completion of pouring, to die opening. The die opening sequence is the sequence of opening a die made up of a plurality of die components. The spray application amount is the amount of application of die release agent used to easily remove a casting from a die. The spray time is a time to apply die release agent. The air blow sequence is the sequence of removing die release agent remaining on a die with air blow.

[0031] The programs to be run by the arithmetic unit 100 include a separation unit 101, a model control unit 102, a first model 103, a second model 104, a prediction accuracy determining unit 105, and a prediction unit 106. In another embodiment, an integrated circuit, such as field-programmable gate array (FPGA) and application specific integrated circuit (ASIC), may run these programs. A server, a PC, an arithmetic unit, and an integrated circuit may be regarded as a computer.

[0032] The separation unit 101 is a program that acquires information on the existing product from the storage device 110 and that separates the information on the existing product into the defect characteristic value of the existing product and the shape information of the existing product.

[0033] The first model 103 is a program trained based on the defect characteristic value indicating a defect associated with a location in the existing product, a feature of the three-dimensional shape of the existing product, and the conditional information indicating the manufacturing condition of the existing product. The first model 103 may be trained by using machine learning, such as deep learning. In the case of, for example, deep learning, the first model 103 can be implemented by a neural network. Machine learning is not limited to deep learning, and another technology may be adopted.

[0034] The second model 104 is a program that, when the shape information indicating the three-dimensional shape of the existing product is input, outputs the feature of the three-dimensional shape of the existing product. The second model 104 may be implemented by a convolutional neural network. The feature of the three-dimensional shape can be expressed in the form of a feature vector.

[0035] The model control unit 102 is a program that controls the first model 103 and the second model 104. The model control unit 102 is capable of training the second model 104 by inputting the shape information indicating the three-dimensional shape of the existing product to the second model 104. The model control unit 102 is capable of training the first model 103 by using the feature output from the second pre-trained model 104, the defect characteristic value of the existing product, obtained by the separation unit 101, and one or more pieces of the conditional information on the existing product, saved in the storage device 110.

[0036] In another embodiment, the model control unit 102 may input not only the feature output from the second pre-trained model 104 but also another feature of the existing product to the first model 103. When the existing product is a casting, examples of the other feature include a die volume of the casting, a casting volume, a casting surface area, and a thickness of the casting. In other words, the first model 103 may be further trained by at least one of the die volume of the casting, the casting volume, the casting surface area, and the thickness of the casting.

[0037] The prediction accuracy determining unit 105 is a program that compares the defect characteristic value output from the first model 103 with the defect characteristic value of the existing product and that determines whether the defect prediction accuracy of the first model 103 is higher than or equal to a set accuracy. With this determination, the defect characteristic value at each location in the existing product, obtained by simulation, such as CAE analysis, can be used as the defect characteristic value of the existing product.

[0038] A defect characteristic value is able to indicate the degree of defect of a product. The prediction accuracy determining unit 105 is able to, when the difference between the defect characteristic value output from the first model 103 and the defect characteristic value of the existing product, obtained by simulation, is less than or equal to a prescribed value, determine that the defect prediction accuracy of the first model 103 is higher than or equal to the set accuracy.

[0039] The prediction unit 106 is a program that predicts a defect of the target product by using the first pre-trained model 103. Specifically, the prediction unit 106 is capable of predicting a defect of the target product by inputting, to the first pre-trained model 103, a feature indicating a feature quantity of the three-dimensional shape of the target product. The prediction unit 106 displays a defect characteristic value indicating a defect associated with a location in the target product on the display device 120 based on the defect characteristic value of the target product, output from the first pre-trained model 103.

[0040] In another embodiment, the prediction unit 106 may predict a defect of the target product by inputting not only a feature indicating a feature quantity of the three-dimensional shape of the target product but also conditional information indicating one or more manufacturing conditions of the target product to the first pre-trained model 103.

[0041] FIG. 4 is a view showing an example of an image showing a prediction result of defects of the target product. In the example shown in FIG. 4, the defects of the target product are shown location by location. In the example shown in FIG. 4, the defects are represented by circle marks for the sake of convenience. Alternatively, the defects of the target product may be represented location by location by using color indication or various shapes. In this case, the degree of defect of the target product may be expressed by the type of color. The degree of defect of the target product may also be expressed by the size of the shape indicating each defect.

[0042] FIG. 2 is a flowchart showing an example of a process of training the first model 103 and the second model 104. In step S101, the separation unit 101 of the prediction apparatus 10 separates the information on the existing product into the defect characteristic values and the shape information. In step S102, the model control unit 102 inputs the shape information on the existing product to the second model 104.

[0043] In step S103, the second model 104 executes a process of convolving the three-dimensional shape of the existing product by using the shape information on the existing product, and generates a feature of the three-dimensional shape of the existing product. In step S104, the second model 104 outputs the generated feature of the three-dimensional shape of the existing product.

[0044] In step S105, the model control unit 102 inputs the defect characteristic values of the existing product, obtained in step S101, the feature of the existing product, output in step S104, and the conditional information on the existing product to the first model 103.

[0045] In step S106, the first model 103 associates the defect characteristic values, feature, and conditional information on the existing product with one another. In step S107, the first model 103 constructs a regression expression based on the association among the defect characteristic values, feature, and conditional information on the existing product. In step S108, the first model 103 outputs the defect characteristic values respectively associated with locations in the existing product.

[0046] In step S109, the prediction accuracy determining unit 105 compares the defect characteristic values output from the first model 103 with the defect characteristic values of the existing product, obtained by simulation, and determines whether the defect prediction accuracy of the first model 103 is higher than or equal to a set accuracy. When the defect prediction accuracy of the first model 103 is lower than the set accuracy (NO), the process returns to step S102, and the first model 103 and the second model 104 are repeatedly trained. On the other hand, when the defect prediction accuracy of the first model 103 is higher than or equal to the set accuracy (YES), the process of FIG. 2 ends.

[0047] FIG. 3 is a flowchart showing an example of a process of predicting a defect of a target product. In step S201, the prediction unit 106 of the prediction apparatus 10 inputs a feature indicating the feature quantity of the three-dimensional shape of the target product to the first pre-trained model 103 trained through the process shown in FIG. 2. In step S202, the first pre-trained model 103 outputs a defect characteristic value indicating a defect associated with a location in the target product. In step S203, the prediction unit 106 displays the defect characteristic value of the target product, output by the first pre-trained model 103, on the display device 120, and the process of FIG. 3 ends.

[0048] In the above embodiment, the first model 103 is trained based on the defect characteristic value indicating the defect associated with the location in the existing product, the feature of the three-dimensional shape of the existing product, and the conditional information indicating the manufacturing condition of the existing product. When a feature of the three-dimensional shape of the target product and conditional information indicating a manufacturing condition of the target product are input to the first pre-trained model 103, the first pre-trained model 103 outputs a defect characteristic value indicating a defect associated with a location in the target product.

[0049] The defect characteristic value of the existing product correlates with the defect characteristic value of the target product. The three-dimensional shape of a product correlates with a defect of the product. The manufacturing condition of a product correlates with a defect of the product. Therefore, a defect associated with the three-dimensional shape of the target product is able to be predicted by using the first pre-trained model 103 trained based on the defect characteristic value, the feature of the three-dimensional shape, and the conditional information indicating the manufacturing condition, of the existing product. Therefore, a defect due to the shape of the target product, such as a new product, is able to be predicted location by location.

[0050] When the product is a casting, the defect of the product, indicated by the defect characteristic value, includes at least one of seizure, shrinkage, flow line, galling, raw material deformation, die cracking, and entrapment of the product. Thus, it is possible to predict seizure, shrinkage, flow line, galling, raw material deformation, die cracking, and entrapment of the target product. Particularly, it is possible to predict seizure, shrinkage, flow line, galling, raw material deformation, die cracking, and entrapment of the target product location by location.

[0051] The defect characteristic value includes a value indicating the degree of defect of the target product. Therefore, it is possible to predict the degree of defect of the target product location by location.

[0052] When the product is a casting, the first pre-trained model 103 may be further trained by at least one of a die volume of the casting, a casting volume, a casting surface area, and a thickness of the casting. Thus, the first pre-trained model 103 is capable of predicting a defect with considerations to the die volume of a casting, a casting volume, a casting surface area, and the thickness of the casting.

[0053] When the product is a casting, the manufacturing condition includes at least one of molten metal type, molten metal temperature, internal cooling temperature, water flow time, die temperature, die surface treatment, cycle time, die time, die opening sequence, spray application amount, spray time, and air blow sequence. Thus, it is possible to predict a defect of the target product location by location based on these various manufacturing conditions.

[0054] In the above example, a program includes a command set (or software code) for causing a computer to execute one or more functions described in the embodiment when the program is loaded onto the computer. The program may be stored in a non-transitory computer-readable medium or a tangible non-transitory storage medium. Nonrestrictive examples of the computer-readable medium or tangible non-transitory storage medium include memory technologies, such as a random-access memory (RAM), a read-only memory (ROM), a flash memory, a solid-state drive (SSD), and others, optical disk storages, such as a CD-ROM, a digital versatile disc (DVD), a Blu-ray (registered trademark) disc, and others, and magnetic storage devices, such as a magnetic cassette, a magnetic tape, a magnetic disk storage, and others. The program may be transmitted on a temporary computer-readable medium or communication medium. Nonrestrictive examples of the temporary computer-readable medium or communication medium include an electrical, optical, acoustic, or other-type propagation signals.

[0055] The disclosure is not limited to the above-described embodiment and may be modified as needed without departing from the scope of the disclosure. For example, in the above embodiment, the prediction apparatus 10 that is a single apparatus may be regarded as a prediction system. Alternatively, in another embodiment, a prediction system may be implemented by a plurality of apparatuses. For example, the separation unit 101, model control unit 102, first model 103, second model 104, prediction accuracy determining unit 105, and prediction unit 106 implemented in the arithmetic unit 100 may be implemented in a distributed manner among a plurality of apparatuses.