Prediction system, prediction method, and nonvolatile storage medium

Abstract

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

Description

Technical Field

[0001] This invention relates to prediction systems, prediction methods, and non-volatile storage media. Background Technology

[0002] Typically, technologies for predicting product defects are used across various industries, including the automotive sector. As an example of such technology, a characteristic prediction device is described in WO2018 / 062398. When parameters indicating the manufacturing conditions of an aluminum product are input, the characteristic prediction device predicts the characteristics of the aluminum product by using a neural network that outputs characteristic values ​​of the aluminum product manufactured under those conditions. Summary of the Invention

[0003] However, the feature prediction apparatus described in WO 2018 / 062398 predicts product defects based solely on the product's manufacturing conditions, and therefore cannot predict defects caused by the product's shape.

[0004] This invention provides a prediction system, a prediction method, and a non-volatile storage medium capable of predicting defects caused by the shape of a product.

[0005] A first aspect of the present invention relates to a prediction system configured to predict defects in a target product. The prediction system includes a pre-trained first model trained based on defect characteristic values ​​indicating a defect associated with a location in an existing product, features of the three-dimensional shape of the existing product, and condition information indicating the manufacturing conditions of the existing product. The pre-trained first model is configured to output defect characteristic values ​​indicating a defect associated with a location in the target product when features of the three-dimensional shape of the target product are input.

[0006] When the product is a casting, the defects of the product indicated by the defect characteristic value may include at least one of the following: seizure, shrinkage, burn marks, scratches, raw material deformation, mold cracking, and interception.

[0007] Defect characteristic values ​​can include values ​​that indicate the degree of defect in a product.

[0008] When the product is a casting, the pre-trained first model can also be trained using at least one of the following: the mold volume, the casting volume, the casting surface area, and the casting thickness.

[0009] The prediction system may also include a pre-trained second model configured to output features of the three-dimensional shape of the existing product when given shape information indicating the three-dimensional shape of the existing product, and to train the pre-trained first model using the features output from the pre-trained second model.

[0010] When the product is a casting, the manufacturing conditions may include at least one of the following: molten metal type, molten metal temperature, internal cooling temperature, water flow time, mold temperature, mold surface treatment, cycle time, molding time, mold opening sequence, spray application amount, spraying time, and blower sequence.

[0011] The prediction system may also include a display device, and defect characteristic values ​​indicating the location of the defect in the target product may be displayed on the display device.

[0012] A second aspect of the present invention relates to a prediction method for predicting defects in a target product. The prediction method includes: inputting features of the three-dimensional shape of the target product into a pre-trained first model by a computer, the pre-trained first model being trained based on defect characteristic values ​​indicating defects associated with a location in an existing product, shape information indicating the three-dimensional shape of the existing product, and condition information indicating the manufacturing conditions of the existing product; and causing the pre-trained first model to output defect characteristic values ​​indicating defects associated with a location in the target product. This method is applicable to both the configuration of the prediction system according to the first aspect and the prediction method according to the second aspect.

[0013] The prediction method may also include having a computer input shape information indicating the three-dimensional shape of an existing product into a pre-trained second model, and having the pre-trained second model output features of the three-dimensional shape of the existing product; and having a computer train a pre-trained first model by using features output from the pre-trained second model.

[0014] A third aspect of the invention relates to a non-volatile storage medium storing a program that is a pre-trained model configured to predict defects in a target product. The pre-trained model is trained based on defect characteristic values ​​indicating a location of a defect in an existing product, features of the three-dimensional shape of the existing product, and conditional information indicating the manufacturing conditions of the existing product. The pre-trained model is configured to output defect characteristic values ​​indicating a location of a defect in the target product when features of the three-dimensional shape of the target product are input.

[0015] According to the above aspects of the present invention, a prediction system, a prediction method, and a non-volatile storage medium capable of predicting defects caused by the shape of a product can be provided. Attached Figure Description

[0016] The features, advantages, and technical and industrial significance of exemplary embodiments of the present invention will now be described with reference to the accompanying drawings, in which similar symbols denote similar elements, and wherein: Figure 1 This is a block diagram illustrating the configuration of a prediction device according to an embodiment of the present invention; Figure 2This is a flowchart illustrating an example of a process performed by a prediction device according to an embodiment of the present invention; Figure 3 This is a flowchart illustrating another example of a process performed by a prediction device according to an embodiment of the present invention; and Figure 4 This is an example image illustrating a prediction result according to an embodiment of the present invention. Detailed Implementation

[0017] In the following description, embodiments of the invention will be illustrated with reference to the accompanying drawings. Figure 1 This is a block diagram illustrating the configuration of a prediction device 10 according to an embodiment of the present invention. The prediction device 10 predicts defects in a target product. Specific examples of the prediction device 10 include information processing devices such as servers and personal computers (PCs); however, the prediction device 10 is not limited thereto. The prediction device 10 can be considered as a prediction system. Examples of target products include castings used in vehicles such as automobiles.

[0018] The prediction device 10 includes a processing unit 100, a storage device 110, and a display device 120. The processing unit 100 is a processing unit such as a central processing unit (CPU) or a microprocessor unit (MPU). The processing unit 100 executes a prediction method for predicting defects in a target product by reading and running programs stored in the storage device 110.

[0019] Storage device 110 is a storage device that stores programs to be executed by processing unit 100 and various data such as information about existing products and information about target products. Specifically, the information about existing products includes defect characteristic values ​​indicating defects associated with locations in existing products, shape information indicating the three-dimensional shape of existing products, and condition information indicating the manufacturing conditions of existing products. The defect characteristic values ​​and shape information of existing products are obtained, for example, through computer-aided engineering (CAE) analysis of existing products. The information about target products includes shape information indicating the three-dimensional shape of target products and condition information indicating the manufacturing conditions of target products.

[0020] When the product is a casting, the defects indicated by the defect characteristic value can be general casting defects. Specific examples of casting defects include seizure, shrinkage, flow lines, galling, raw material deformation, mold cracking, and entrapment. Raw material deformation refers to undesirable deformation that may occur when the casting is cooled to room temperature after it has been formed during the casting process. Defects indicated by the defect characteristic value are not limited to these examples.

[0021] Defect characteristic values ​​are quantitative variables and indicate the degree of defect in a product. The magnitude of the defect characteristic value indicates the extent of the product's defect.

[0022] When the product is a casting, the manufacturing conditions indicated by the condition information can be conditions set in a typical casting process. Specific examples of manufacturing conditions indicated by the condition information include the type of molten metal, molten metal temperature, internal cooling temperature, water flow time, mold temperature, mold surface treatment, cycle time, molding time, mold opening sequence, coating application amount, coating time, and blower sequence for manufacturing the casting. Manufacturing conditions are not limited to these examples.

[0023] Molten metal type refers to the type of molten metal. Molten metal temperature refers to the temperature of the molten metal. Internal cooling temperature is the temperature of the water that passes through the inside of the mold to cool the casting. Mold temperature is the temperature of the mold at the moment the casting is formed. Water flow time is the time it takes for water to pass through the inside of the mold. Mold surface treatment includes heat treatment to reduce wear on the mold surface.

[0024] Cycle time is the time required during the casting process when castings are produced continuously. The casting cycle consists of mold closing, pouring, curing, mold opening, casting removal, release agent application, air blowing, and mold closing again.

[0025] Molding time is the time included in the cycle time and is the time from curing (i.e., completion of pouring) to mold opening. Mold opening sequence is the order in which the mold, consisting of multiple mold components, is opened. Spray application amount is the amount of release agent applied to facilitate easy removal of the casting from the mold. Spraying time is the time taken to apply the release agent. Blowing sequence is the order in which release agent remaining on the mold is removed using blowers.

[0026] The program to be executed by the computing unit 100 includes a separation unit 101, a model control unit 102, a first model 103, a second model 104, a prediction accuracy determination unit 105, and a prediction unit 106. In another embodiment, integrated circuits such as field-programmable gate arrays (FPGAs) and application-specific integrated circuits (ASICs) can run these programs. The server, PC, computing unit, and integrated circuit can be considered as a computer.

[0027] The separation unit 101 is a program that obtains information about existing products from the storage device 110 and separates the information about existing products into defect characteristic values ​​and shape information of existing products.

[0028] The first model 103 is trained based on a procedure that includes defect characteristic values ​​indicating the location of defects in an existing product, features of the three-dimensional shape of the existing product, and conditional information indicating the manufacturing conditions of the existing product. The first model 103 can be trained using machine learning, such as deep learning. In the case of deep learning, for example, the first model 103 can be implemented using a neural network. Machine learning is not limited to deep learning and other techniques can be employed.

[0029] The second model 104 is a program that outputs the features of the three-dimensional shape of an existing product when shape information indicating the three-dimensional shape of the existing product is input. The second model 104 can be implemented by a convolutional neural network. The features of the three-dimensional shape can be expressed in the form of feature vectors.

[0030] The model control unit 102 is a program that controls the first model 103 and the second model 104. The model control unit 102 can train the second model 104 by inputting shape information indicating the three-dimensional shape of the existing product into the second model 104. The model control unit 102 can train the first model 103 by using features output from the pre-trained second model 104, defect characteristic values ​​of the existing product obtained by the separation unit 101, and one or more conditional information about the existing product stored in the storage device 110.

[0031] In another embodiment, the model control unit 102 can input not only the features output from the pre-trained second model 104 into the first model 103, but also other features of the existing product. When the existing product is a casting, examples of other features include the mold volume, casting volume, casting surface area, and casting thickness. In other words, the first model 103 can be further trained using at least one of the mold volume, casting volume, casting surface area, and casting thickness.

[0032] The prediction accuracy determination unit 105 is a procedure that compares the defect characteristic values ​​output from the first model 103 with the defect characteristic values ​​of existing products and determines whether the defect prediction accuracy of the first model 103 is equal to or higher than a set accuracy. Using this determination, the defect characteristic values ​​of various parts of existing products obtained through simulations such as CAE analysis can be used as the defect characteristic values ​​of existing products.

[0033] Defect characteristic values ​​can indicate the degree of defects in a product. The prediction accuracy determination unit 105 can determine that the defect prediction accuracy of the first model 103 is equal to or higher than the set accuracy when the difference between the defect characteristic value output from the first model 103 and the defect characteristic value of the existing product obtained through simulation is less than or equal to a predetermined value.

[0034] The prediction unit 106 is a procedure that predicts defects in a target product using a pre-trained first model 103. Specifically, the prediction unit 106 is capable of predicting defects in the target product by inputting features indicating the three-dimensional shape of the target product into the pre-trained first model 103. Based on the defect characteristic values ​​of the target product output from the pre-trained first model 103, the prediction unit 106 displays the defect characteristic values ​​indicating the defects associated with their location in the target product on the display device 120.

[0035] In another embodiment, the prediction unit 106 can predict defects in the target product by inputting not only features indicating the three-dimensional shape of the target product into the pre-trained first model 103, but also condition information indicating one or more manufacturing conditions of the target product.

[0036] Figure 4 This is an example image showing the predicted defects of the target product. Figure 4 In the example shown, defects in the target product are illustrated position by position. Figure 4 In the example shown, for convenience, defects are represented by circular markers. Alternatively, defects in the target product can be represented positionally by using color indicators or various shapes. In this case, the severity of the defect can be expressed by the type of color. The severity of the defect can also be expressed by indicating the size of the shape representing each defect.

[0037] Figure 2 This is a flowchart illustrating an example of the process of training the first model 103 and the second model 104. In step S101, the separation unit 101 of the prediction device 10 separates information about the existing product into defect characteristic values ​​and shape information. In step S102, the model control unit 102 inputs the shape information about the existing product into the second model 104.

[0038] In step S103, the second model 104 performs convolution processing on the three-dimensional shape of the existing product using shape information about the existing product, and generates features of the three-dimensional shape of the existing product. In step S104, the second model 104 outputs the generated features of the three-dimensional shape of the existing product.

[0039] In step S105, the model control unit 102 inputs the defect characteristic value of the existing product obtained in step S101, the characteristics of the existing product output in step S104, and the condition information about the existing product into the first model 103.

[0040] In step S106, the first model 103 correlates the defect characteristic values, features, and conditional information of the existing product with each other. In step S107, the first model 103 constructs a regression expression based on the correlation between the defect characteristic values, features, and conditional information of the existing product. In step S108, the first model 103 outputs the defect characteristic values ​​associated with their respective locations in the existing product.

[0041] In step S109, the prediction accuracy determination unit 105 compares the defect characteristic value output from the first model 103 with the defect characteristic value of the existing product obtained through simulation, and determines whether the defect prediction accuracy of the first model 103 is equal to or higher than the 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 trained repeatedly. On the other hand, when the defect prediction accuracy of the first model 103 is equal to or higher than the set accuracy (Yes), Figure 2 The processing is now complete.

[0042] Figure 3 This is a flowchart illustrating an example of a process for predicting defects in a target product. In step S201, the prediction unit 106 of the prediction device 10 inputs a feature quantity indicating the three-dimensional shape of the target product into the system... Figure 2 The pre-trained first model 103 shown is used for processing training. In step S202, the pre-trained first model 103 outputs defect characteristic values ​​indicating the location of the defect in the target product. In step S203, the prediction unit 106 displays the defect characteristic values ​​of the target product output by the pre-trained first model 103 on the display device 120, and Figure 3 The processing is now complete.

[0043] In the above embodiments, the first model 103 is trained based on defect characteristic values ​​indicating the location of defects in the existing product, the features of the three-dimensional shape of the existing product, and condition information indicating the manufacturing conditions of the existing product. When the features of the three-dimensional shape of the target product and the condition information indicating the manufacturing conditions of the target product are input into the pre-trained first model 103, the pre-trained first model 103 outputs defect characteristic values ​​indicating the location of defects in the target product.

[0044] The defect characteristic values ​​of the existing product are related to the defect characteristic values ​​of the target product. The three-dimensional shape of the product is related to the product's defects. The manufacturing conditions of the product are related to the product's defects. Therefore, a pre-trained first model 103 trained based on the defect characteristic values ​​of the existing product, the features of the three-dimensional shape, and condition information indicating the manufacturing conditions is used to predict defects associated with the three-dimensional shape of the target product. Thus, it is possible to predict defects caused by the shape of the target product, such as a new product, at each location.

[0045] When the product is a casting, the defects indicated by the defect characteristic value include at least one of the following: seizure, shrinkage, burn marks, scratches, raw material deformation, mold cracking, and interception. Therefore, it is possible to predict seizure, shrinkage, burn marks, scratches, raw material deformation, mold cracking, and interception of the target product. Specifically, it is possible to predict seizure, shrinkage, burn marks, scratches, raw material deformation, mold cracking, and interception of the target product at individual locations.

[0046] Defect characteristic values ​​include values ​​that indicate the degree of defect in the target product. Therefore, it is possible to predict the degree of defect in the target product location by location.

[0047] When the product is a casting, the pre-trained first model 103 can be further trained using at least one of the mold volume, casting volume, casting surface area, and casting thickness. Therefore, the pre-trained first model 103 is able to predict defects by taking into account the mold volume, casting volume, casting surface area, and casting thickness.

[0048] When the product is a casting, the manufacturing conditions include at least one of the following: molten metal type, molten metal temperature, internal cooling temperature, water flow time, mold temperature, mold surface treatment, cycle time, molding time, mold opening sequence, coating application amount, coating time, and blower sequence. Therefore, it is possible to predict defects in the target product location by location based on these various manufacturing conditions.

[0049] In the above examples, the program includes a set of commands (or software code) for causing the computer to perform one or more functions described in this embodiment when the program is loaded onto the computer. The program may be stored in a non-volatile computer-readable medium or a tangible non-volatile storage medium. Non-limiting examples of computer-readable media or tangible non-volatile storage media include memory technologies such as random-access memory (RAM), read-only memory (ROM), flash memory, solid-state drives (SSDs), and others; optical disc storage such as CD-ROMs, digital versatile discs (DVDs), Blu-ray discs, and others; and magnetic storage devices such as magnetic tape cassettes, magnetic tapes, disk drives, and others. The program may be transmitted on a volatile computer-readable medium or a communication medium. Non-limiting examples of volatile computer-readable media or communication media include electrical, optical, acoustic, or other types of propagating signals.

[0050] This invention is not limited to the above embodiments and can be modified as needed without departing from the scope of the invention. For example, in the above embodiments, the prediction device 10, as a single device, can be considered as a prediction system. Alternatively, in another embodiment, the prediction system can be implemented by multiple devices. For example, the separation unit 101, model control unit 102, first model 103, second model 104, prediction accuracy determination unit 105, and prediction unit 106 implemented in the computing unit 100 can be implemented in a distributed manner among multiple devices.

Claims

1. A prediction system configured to predict defects in a target product, characterized in that, The prediction system includes a pre-trained first model and a pre-trained second model. The pre-trained second model is configured to output the features of the three-dimensional shape of the existing product when shape information indicating the three-dimensional shape of the existing product is input. The pre-trained first model is trained based on defect characteristic values ​​indicating defects associated with their location in the existing product, features of the three-dimensional shape of the existing product output from the pre-trained second model, and condition information indicating the manufacturing conditions of the existing product. The first pre-trained model is configured to output defect characteristic values ​​that indicate the location of defects in the target product when the three-dimensional shape features of the target product are input.

2. The prediction system according to claim 1, characterized in that: The product is a casting; and The defects of a product indicated by the defect characteristic value include at least one of the following: seizure, shrinkage, burn marks, scratches, raw material deformation, mold cracking, and interception.

3. The prediction system according to claim 1 or 2, characterized in that, The defect characteristic values ​​include values ​​that indicate the degree of defect in the product.

4. The prediction system according to claim 1 or 2, characterized in that, When the product is a casting, the pre-trained first model is also trained using at least one of the mold volume, casting volume, casting surface area, and casting thickness.

5. The prediction system according to claim 1 or 2, characterized in that, When the product is a casting, the manufacturing conditions include at least one of the following: molten metal type, molten metal temperature, internal cooling temperature, water flow time, mold temperature, mold surface treatment, cycle time, molding time, mold opening sequence, spray application amount, spraying time, and blower sequence.

6. The prediction system according to claim 1 or 2, characterized in that, It also includes a display device, wherein The defect characteristic value, indicating the location of the defect in the target product, is displayed on the display device.

7. A method for predicting defects in a target product, characterized in that, The prediction method includes: The computer inputs shape information indicating the three-dimensional shape of the existing product into a pre-trained second model, and the pre-trained second model outputs the features of the three-dimensional shape of the existing product. The computer inputs the features of the three-dimensional shape of the target product into a pre-trained first model, which is trained based on defect characteristic values ​​indicating defects associated with locations in the existing product, features of the three-dimensional shape of the existing product output from the pre-trained second model, and condition information indicating the manufacturing conditions of the existing product; and The pre-trained first model outputs defect characteristic values ​​that indicate the location of the defect in the target product.

8. The prediction method according to claim 7, characterized in that, When the product is a casting, the defects of the product indicated by the defect characteristic value include at least one of the following: seizing, shrinkage, burn marks, scratches, raw material deformation, mold cracking, and interception.

9. The prediction method according to claim 7 or 8, characterized in that, The defect characteristic values ​​include values ​​that indicate the degree of defect in the product.

10. The prediction method according to claim 7 or 8, characterized in that, When the product is a casting, the pre-trained first model is also trained using at least one of the mold volume, casting volume, casting surface area, and casting thickness.

11. The prediction method according to claim 7 or 8, characterized in that, When the product is a casting, the manufacturing conditions include at least one of the following: molten metal type, molten metal temperature, internal cooling temperature, water flow time, mold temperature, mold surface treatment, cycle time, molding time, mold opening sequence, spray application amount, spraying time, and blower sequence.

12. The prediction method according to claim 7 or 8, characterized in that, It also includes displaying defect characteristic values ​​on a display device that indicate the location of the defect in the target product.

13. A non-volatile storage medium storing a program, said program being a pre-trained first model configured to predict defects in a target product, characterized in that: The pre-trained first model is trained based on defect characteristic values ​​indicating defects associated with locations in an existing product, features of the three-dimensional shape of the existing product output from a pre-trained second model, and condition information indicating the manufacturing conditions of the existing product. The pre-trained second model is configured to output features of the three-dimensional shape of the existing product when shape information indicating the three-dimensional shape of the existing product is input. The first pre-trained model is configured to output defect characteristic values ​​that indicate the location of defects in the target product when the three-dimensional shape features of the target product are input.

Images