Core quality inference system, core quality inference method, non-temporary storage medium, learning model, machine learning device

Abstract

The present invention is a core quality inference system for inferring the quality of a core, which is molded by filling mixed sand mixed in a mixing tank into a mold and heating it. The core quality inference system is equipped with a computer, which performs the following operations: obtaining mold temperature information of the mold; obtaining environmental information related to the surrounding environment of molding the core; and inferring the quality of the core based on the mold temperature information and the environmental information.

Description

Technical Field

[0001] The present invention relates to a core quality inference system, a core quality inference method, a non-temporary storage medium, a learning-completed model, and a machine learning device. Background Art

[0002] Generally, in a core molding device for molding a core for casting, a core is molded by filling a mold with kneaded sand obtained by kneading core sand, water glass, water, a surfactant, and the like, which are core raw materials, and heating the mold.

[0003] Japanese Patent Application Laid-Open No. 2020-110811 discloses a technology in which, in such a core molding device, the temperature of the mold when the mixed sand is filled into the mold is measured, and an abnormality is reported when the difference from the sintering temperature, which is the appropriate temperature for molding the core, is greater than a specified value, thereby knowing the possibility of reduced quality of the core. Summary of the Invention

[0004] The present inventors have discovered the following problem in quality management of core molding equipment. In the core molding equipment, core sand, water glass, water, surfactant, and the like are kneaded in a pre-calculated optimal ratio. However, due to evaporation during kneading, the amount of water contained in the kneaded sand when actually filled into the mold is affected by environmental factors such as temperature and humidity. Furthermore, the amount of water in the kneaded sand causes its dynamic viscosity to vary, resulting in the behavior of the kneaded sand when filled into the mold being affected by the surrounding environment.

[0005] However, in the quality management of the core molding device described in Japanese Patent Application Laid-Open No. 2020-110811, the influence of the surrounding environment is not taken into consideration, and there is room for further improvement in the quality management of the core molding device.

[0006] The present invention provides a core quality inference system, a core quality inference method, a non-temporary storage medium, a learned model, and a machine learning device capable of learning the quality of a core, which can infer the quality of a core with higher accuracy.

[0007] The core quality inference system involved in the first embodiment of the present invention infers the quality of the core. The core is molded by filling the mixed sand mixed in the mixing tank into the mold and heating it. The core quality inference system has a computer, which performs the following operations: obtaining mold temperature information of the mold; obtaining environmental information related to the surrounding environment of molding the core; and inferring the quality of the core based on the mold temperature information and the environmental information.

[0008] With this configuration, the quality of the molded core is estimated in consideration of environmental information related to the surrounding environment of the molded core. Therefore, the quality of the core can be estimated with higher accuracy based on changes in the amount of water contained in the kneaded sand.

[0009] In the core quality estimation system described above, the environmental information may include at least one of humidity information and air pressure information of the surrounding environment. With this configuration, by taking into account at least one of the humidity information and air pressure information, it is possible to more accurately estimate the quality of the core based on changes in the amount of water contained in the mixed sand.

[0010] In the core quality estimation system described above, the computer may also obtain the environmental information based on the position information of the core. This configuration eliminates the need for sensors for obtaining environmental information, thereby reducing costs.

[0011] In the core quality estimation system described above, the computer may estimate the quality of the core molded from the kneaded sand before the kneaded sand is completely filled into the mold. With this configuration, by performing the estimation before the kneaded sand is completely filled, the mold temperature can be controlled based on the estimation result, thereby preventing defects in the core.

[0012] In the core quality estimation system of the above embodiment, the computer may estimate the quality of the core molded from the kneaded sand before kneading of the kneaded sand in the kneading tank is completed. With this configuration, by estimating the quality of the core molded from the kneaded sand before kneading of the kneaded sand in the kneading tank is completed, the amount of water added during kneading in the kneading tank can be controlled, thereby preventing defects in the core.

[0013] In the core quality estimation system described above, the computer can also estimate the quality of the core by analyzing the behavior of the kneaded sand within the mold using CAE. With this configuration, the behavior of the kneaded sand within the mold can be analyzed using CAE, and the quality of the core can be estimated based on the behavior within the mold.

[0014] In the core quality estimation system described above, the computer may also estimate the core quality based on mold temperature information of the mold during filling or heating of the mixed sand, obtained using the behavior analysis. With this configuration, core quality can be estimated based on mold temperature information by analyzing the behavior of the mixed sand within the mold, thereby enabling estimation of core quality, which is affected by mold temperature.

[0015] In the core quality estimation system described above, the computer may also estimate the core quality based on the behavior information of the kneaded sand during filling, obtained using the behavior analysis. With this configuration, the core quality can be estimated based on the behavior information during filling by analyzing the behavior of the kneaded sand within the mold, thereby enabling estimation of the core quality, which is influenced by the behavior of the kneaded sand.

[0016] In the core quality estimation system described above, the computer may estimate the quality of the core based on the mold temperature information and the environmental information using a learned model. The learned model is machine-learned using supervised data that takes the mold temperature information and environmental information as input and outputs quality information related to the core. With this configuration, the quality of the molded core is estimated using a learned model that has previously learned the relationship between the mold temperature information, environmental information, and core quality information. This allows for more accurate estimation of the core quality based on changes in the amount of water contained in the mixed sand.

[0017] In the core quality estimation system described above, the computer may further perform the following operations: accepting corrections to the estimation results; and, if the corrections are accepted, updating the learned model based on the corrected results. This configuration enables appropriate relearning of the learned model, enabling higher-precision estimation of core quality.

[0018] In the core quality estimation system of the above embodiment, the computer may further execute: controlling the molding conditions of the molding core based on the estimation result. With such a structure, defects in the core can be prevented by controlling the molding conditions based on the estimation result.

[0019] In the core quality estimation system described above, the computer may also control at least one of the mold temperature and the amount of water added during kneading in the kneading tank. With this configuration, core defects can be prevented by adjusting the mold temperature and the amount of water added.

[0020] The core quality inference method involved in the second embodiment of the present invention infers the quality of the core by using a computer. The core is molded by filling the mixed sand mixed in the mixing tank into the mold and heating it. The core quality inference method includes: obtaining mold temperature information of the mold; obtaining environmental information related to the surrounding environment of molding the core; and inferring the quality of the core based on the mold temperature information and the environmental information.

[0021] A non-transitory storage medium according to a third aspect of the present invention stores one or more computer-executable instructions for inferring the quality of a core molded by filling a mold with mixed sand kneaded in a kneading tank and heating the mold. The instructions cause the one or more computers to perform the following functions. The instructions include: obtaining mold temperature information of the mold; obtaining environmental information related to the environment surrounding the molding of the core; and inferring the quality of the core based on the mold temperature information and the environmental information.

[0022] In addition, the learned model involved in the fourth embodiment of the present invention enables the computer to perform a function, and the aforementioned function is to output quality information related to the quality of the aforementioned core based on the mold temperature information of the mold filled with mixed sand and the environmental information related to the surrounding environment of the molding core, and use the aforementioned mold temperature information and the aforementioned environmental information as input and the aforementioned quality information as output to perform machine learning. Supervisory data.

[0023] With this configuration, the quality of the molded core is estimated in consideration of environmental information related to the surrounding environment of the molded core. Therefore, the quality of the core can be estimated with higher accuracy based on changes in the amount of water contained in the kneaded sand.

[0024] In addition, the machine learning device involved in the fifth embodiment of the present invention learns the quality of the core, and the core is molded by filling the mixed sand mixed in the mixing tank into the mold and heating it. The machine learning device has a computer, which performs the following operations: obtaining mold temperature information of the mold; obtaining environmental information related to the surrounding environment of the molding of the core; obtaining quality information related to the quality of the core; and using supervision data with the mold temperature information and the environmental information as input and the quality information as output to learn the quality of the molded core.

[0025] With such a configuration, the quality of the core of the mold is learned, and a learned model capable of estimating the quality of the core of the mold can be generated in consideration of environmental information related to the surrounding environment.

[0026] According to the present invention, a core quality estimation system, a core quality estimation method, a non-transitory storage medium, a learned model, and a machine learning device capable of learning the quality of a core can be provided, which can estimate the quality of a core with higher accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Features, advantages, and technical and industrial significance of exemplary embodiments of the present invention will be described below with reference to the accompanying drawings, in which like numerals refer to like elements.

[0028] Figure 1This is a schematic cross-sectional view of the entire core molding device during the mixing of mixed sand.

[0029] Figure 2 This is a schematic cross-sectional view of the entire core molding device during the filling of mixed sand.

[0030] Figure 3 This is a diagram schematically showing the hardware configuration of the core quality estimation system according to the first embodiment.

[0031] Figure 4 It is a diagram showing the functional configuration of the core quality estimation system according to the first embodiment.

[0032] Figure 5 This is a flowchart of the core quality estimation process according to the first embodiment.

[0033] Figure 6 It is a diagram showing the functional configuration of a core quality estimation system according to the second embodiment.

[0034] Figure 7 This is a diagram showing the functional configuration of a machine learning device according to the second embodiment.

[0035] Figure 8 This is a diagram showing a neural network used in machine learning according to the second embodiment.

[0036] Figure 9 This is a flowchart showing the flow of the learning process according to the second embodiment.

[0037] Figure 10 This is a flowchart of the core quality estimation process according to the second embodiment. DETAILED DESCRIPTION

[0038] Hereinafter, specific embodiments of the present invention will be described in detail with reference to the accompanying drawings. However, the present invention is not limited to the following embodiments. In addition, the following description and drawings are appropriately simplified for clarity of description.

[0039] First, before describing a core quality estimation system according to an embodiment to which the present invention is applied, a core molding device for molding a core whose quality is estimated by the core quality estimation system will be briefly described. Figure 1 、 Figure 2 It is an overall cross-sectional view schematically showing a core molding device.

[0040] like Figure 1 and Figure 2 As shown, the core molding device 100 includes a kneading tank 110 , a raw material supply device 120 , a base 130 , a piston 140 , a cylinder 150 , and a mold 160 .

[0041] The kneading tank 110 is a storage tank for producing kneaded sand by kneading raw materials. The kneading tank 110 is a bottomed container with an open top and a bottom, such as a cylindrical shape. The kneading tank 110 has an inner diameter of approximately 150 to 350 mm and a height of approximately 150 to 350 mm, for example, an inner diameter of approximately 250 mm and a height of approximately 250 mm. In the kneading tank 110, core sand, water glass, water, surfactants, and other liquid additives serving as the raw materials for the core are supplied from the open top via a raw material supply device 120 (core sand supply device 121, water glass supply device 122, water supply device 123, and surfactant supply device 124). The water glass acts as a binder. The binder is not limited to water glass; inorganic substances such as clay and cement can also be used.

[0042] A through-hole 111 is provided at the bottom of the kneading tank 110 for injecting the kneaded material S produced by kneading the raw materials inside the kneading tank 110 by kneading blades (not shown). In addition, a rubber-like valve with a cutout is provided in the through-hole 111 to prevent the kneaded material S from leaking out of the kneading tank 110 and to open it during injection.

[0043] The base 130 is a member for supporting the kneading tank 110. A convex portion is formed on the upper surface of the base 130 to fit into the through hole 111 provided at the bottom of the kneading tank 110, and the valve provided in the through hole 111 is supported and pressed from the lower side.

[0044] In this way, after the raw materials are kneaded in the kneading tank 110 placed on the base 130, the kneading tank 110 containing the kneaded material S is transferred from the base 130 to the mold 160. The transfer can be performed by a conveying device having a driving device such as a motor, a holding device for holding the kneading tank 110, etc. Figure 1 In FIG, the kneading tank 110 is shown as being placed on the base 130, and the kneading tank 110 on the mold 160 is shown as a two-dot chain line. Figure 2 In FIG. 1 , the kneading tank 110 is shown in a state where it is placed on the mold 160 after being transferred, and the kneading tank 110 on the base 130 is indicated by a two-dot chain line.

[0045] like Figure 2 As shown, after the kneading tank 110 is placed on the mold 160, the kneaded material S in the kneading tank 110 is injected into the mold 160 by the piston 140 to fill the mold. Here, the piston 140 can be moved in the vertical direction by the cylinder 150. When the cylinder 150 is lowered by a control unit (not shown), the piston 140 is lowered, and the kneaded material S is injected into the mold 160 to fill the mold.

[0046] The mixed material S filled in the mold 160 is heated inside the mold 160 through the mold 160 whose temperature is regulated by the mold thermal controller 161, and hardens along the inner wall of the mold 160, thereby molding the core of the mold shape. Here, the mold thermal controller 161 heats, cools down, and maintains the temperature of each part of the mold 160, and can be set inside or outside the mold 160. Specifically, the mold temperature is adjusted by a heater, a cooling water circulator, etc. In addition, a temperature sensor 162 is provided on the mold 160 to continuously or discontinuously measure the mold temperature of the mold 160. The temperature sensor 162 can be, for example, a contact temperature sensor such as a thermocouple, or a non-contact temperature sensor. In addition, the location and number of the temperature sensor 162 are appropriately determined according to the shape of the mold 160.

[0047] The molded core is taken out from the mold 160 . For example, the left mold 163 is configured to be movable relative to the right mold 164 , and the left mold 163 is separated from the right mold 164 , thereby allowing the core to be taken out from the mold 160 .

[0048] Here, an example of a core molding device for molding a core whose quality is inferred by the core quality inference system involved in the present invention is described, but its structure and molding process can also be changed in various ways. In the core quality inference system involved in the present invention, the core molding device for molding a core that is an inference object is not limited to the structure of the above-mentioned core molding device.

[0049] First embodiment

[0050] Next, the core quality estimation system according to the first embodiment will be described. Figure 3 This is a block diagram schematically showing the hardware configuration of the core quality estimation system 200 according to the present embodiment.

[0051] like Figure 3 As shown, the core quality estimation system 200 includes computer resources common to typical information processing devices such as personal computers. Specifically, it includes a CPU (Central Processing Unit) 201, ROM (Read Only Memory) 202, RAM (Random Access Memory) 203, an HDD (Hard Disk Drive) 204, a communication interface (I / F) 205, and an input / output interface (I / F) 206. Furthermore, these components are interconnected and communicative via a data bus 207.

[0052] While this example illustrates the various components of the core quality estimation system 200 being implemented by an information processing device embedded with the control functions of the core molding apparatus 100, the core quality estimation system can also be implemented by an information processing device separate from the control device that controls the core molding apparatus 100. Furthermore, some or all of these functions can be implemented on external devices such as edge devices or servers (cloud). Specifically, information related to various devices within a factory can be aggregated and implemented on an in-factory server that functions as a platform for data management and analysis, or on an inter-factory server that manages devices across multiple factories.

[0053] The CPU 201 is a microprocessor that comprehensively controls the core quality estimation system 200. Specifically, it reads various control programs executed in this embodiment stored in the ROM 202 and HDD 204 and executes these programs loaded onto the RAM 203. Here, an SSD (Solid State Drive) may be used as a storage device in place of the HDD 204 or in addition to the HDD 204.

[0054] Communication interface 205 enables communication between the core quality estimation system 200 and external devices. Communication can be achieved using various communication technologies, whether wireless or wired. In this embodiment, for example, the system receives mold temperature information from temperature sensor 162 installed on the core molding device 100.

[0055] The input / output interface 206 performs input / output between the core quality estimation system 200 and the outside world, and includes, for example, an output device such as a display that displays information related to the estimated core quality, and an input device for an operator to use.

[0056] Next, the functional configuration of the core quality estimation system 200 according to the present embodiment will be described. Figure 4 2 is a block diagram showing the functional configuration of the core quality estimation system 200 according to the present embodiment.

[0057] like Figure 4 As shown in FIG, the core quality estimation system 200 includes a mold temperature information acquisition unit 301, an environmental information acquisition unit 302, an estimation unit 303, a control unit 304, and a display unit 305 as its functional structure. These functions are realized by the CPU 201 executing various control programs stored in the ROM 202 and other devices. Furthermore, some or all of the functions of the core quality estimation system 200 may be implemented using hardware circuits.

[0058] The mold temperature information acquisition unit 301 acquires mold temperature information of the mold 160 filled with kneaded sand S. Specifically, the mold temperature information is acquired from a temperature sensor 162 installed in the mold 160 to measure the mold temperature. Furthermore, if multiple temperature sensors 162 are installed in the mold 160, for example, the mold temperature and position information of the temperature sensors 162 in the mold 160, which is associated with the mold temperature, can be acquired as the mold temperature information. Alternatively, only the mold temperature can be acquired from the temperature sensor 162, and the mold temperature information acquisition unit 301 can associate the mold temperature with the position information and acquire the mold temperature information as the mold temperature information.

[0059] The environmental information acquisition unit 302 acquires environmental information related to the surrounding environment of the molding core. Specifically, based on the location information of the core molding apparatus 100, it can acquire environmental information such as temperature, humidity, and air pressure using externally provided weather information and weather forecast information corresponding to the location information. For example, upon acquiring the location information of the core molding apparatus 100, the environmental information acquisition unit 302 transmits this location information, along with a weather information request, to a weather information providing server connected via a communication network such as the Internet. The weather information providing server reads weather information and other information for the location identified by the received location information from a database and transmits it to the environmental information acquisition unit 302. In this manner, the environmental information acquisition unit 302 can acquire weather information and other information corresponding to the location information.

[0060] Regarding location information, the core molding apparatus 100 and the core quality estimation system 200 may include a GPS function or other function for acquiring location information, and the GPS function may be used to acquire location information. Alternatively, location information input by a user of the core quality estimation system 200 may be utilized. Acquiring environmental information based on location information in this manner eliminates the need for sensors for acquiring environmental information, thereby reducing costs. Alternatively, a configuration may be employed in which environmental information is acquired through wired or wireless communication methods using sensors located around the core molding apparatus 100 that measure environmental information such as temperature, humidity, and air pressure.

[0061] The inference unit 303 infers the quality of the molded core based on the mold temperature information obtained by the mold temperature information acquisition unit 301 and the environmental information obtained by the environmental information acquisition unit 302. For example, the inference unit 303 compares the quality of the core with a pre-set reference value to determine whether the quality is higher than the reference value (i.e., a qualified product) or lower than the reference value (i.e., a defective product). In this case, the reference value does not necessarily have to be a single one; it may be multiple. Furthermore, the degree of quality may be expressed using a numerical value or ranking, generating information indicating specific quality details.

[0062] Specifically, the quality of the core is estimated by analyzing the behavior of the kneaded sand S within the mold 160 using CAE (Computer Aided Engineering). This CAE analysis makes it possible to estimate quality that would be difficult to estimate using calculations, for example. The software used for this analysis can be appropriately selected from various commercially available analysis software, or software that has been modified and improved for this analysis.

[0063] First, the analysis target is the behavior of the kneaded material S, which is filled from the kneading tank 110 into the mold 160 via the piston 140, within the mold 160 (including the gate portion, which serves as the inlet to the mold 160). An analytical model is constructed using the drawing data (CAD data) of these target locations. The constructed analytical model is stored in the inference unit 303. By inputting analytical conditions to this analytical model, the behavior of the kneaded sand S within the mold 160 is analyzed, and the analytical results are output. The analytical results can be output in various formats using various software.

[0064] The behavior of the mixed material S within the mold 160 is affected by factors such as the amount of water contained in the mixed sand S and its dynamic viscosity. Furthermore, the mixed sand S is obtained by mixing core sand, water glass, water, and a surfactant. However, since the water evaporates during mixing, the amount of water contained in the mixed sand S and its dynamic viscosity when actually filling the mold are affected by the surrounding environmental information. In this embodiment, at least mold temperature information and environmental information are input as analysis conditions for behavior analysis, allowing for high-precision behavior analysis that takes these influences into account. Here, any one of air temperature information, humidity information, and air pressure information can be used as environmental information. For example, a combination of two or more information such as air temperature information and humidity information, humidity information and air pressure information, or air temperature information and air pressure information can also be used. This configuration allows for more precise behavior analysis by carefully considering the influence of the amount of water contained in the mixed sand S and its dynamic viscosity.

[0065] In this embodiment, mold temperature information and, as optional analysis conditions other than environmental information, information related to molding conditions such as the type of raw material used in the kneaded sand S, the raw material ratio (e.g., the amount of water added during kneading), the dynamic viscosity of the kneaded sand S, and the injection pressure applied by the piston 140 are pre-set or stored in the inference unit 303 for behavior analysis. However, behavior analysis can also be performed by inputting this information during analysis. In this case, the core quality estimation system 200 also includes a function for acquiring molding conditions such as the type of raw material used in the kneaded sand S, the raw material ratio (e.g., the amount of water added during kneading), the dynamic viscosity of the kneaded sand S, and the injection pressure, allowing this information to be input during analysis and behavior analysis to be performed.

[0066] In this way, the quality of the molded core is estimated based on information obtained through CAE-based behavior analysis of the kneaded sand S. Specifically, the quality of the core is estimated based on mold temperature information of the mold 160 during filling and heating of the kneaded sand S. Specifically, the mold temperature in the area that comes into contact with the kneaded sand S during filling and heating significantly affects the quality of the molded core. Therefore, if the mold temperature deviates from a predetermined temperature range, defects may occur. For example, if the mold temperature exceeds the specified range, defects such as breakage and cracking of the core may occur. On the other hand, if the mold temperature falls below the specified range, defects such as the core sticking to the mold 160 may occur. Furthermore, in this embodiment, the quality of the core is estimated based on mold temperature information obtained through behavior analysis. This allows the mold temperature to be estimated before the kneaded sand S is actually filled into the mold 160. Furthermore, more detailed temperature distribution information can be estimated compared to temperature sensor 162, enabling high-precision estimation of the core quality.

[0067] Furthermore, the inference unit 303 can also infer the quality of the core based on information about the behavior of the mixed sand S during filling, obtained through analysis of its behavior. Specifically, the core quality is inferred based on information about the flow rate and energy of the mixed sand S within the mold 160. That is, information about the behavior of the mixed sand S within the mold 160 significantly impacts the quality of the resulting core. Therefore, if the flow rate, energy, and other parameters deviate from predetermined ranges, defects can occur. For example, if the dynamic viscosity of the mixed sand S is high and the flow rate is slow, the resulting core will experience defects such as sand clogging and wrinkling. On the other hand, if the dynamic viscosity of the mixed sand S is low and the flow rate is fast, the resulting core will experience defects such as sticking to the mold 160 and deformation. Therefore, the inference unit 303 can infer the quality of the core based on this information. Furthermore, the predetermined values ​​for mold temperature, flow rate, and energy are set in advance by verifying the quality through experiments, etc.

[0068] The control unit 304 controls the molding conditions for molding the core based on the inference result of the inference unit 303. Specifically, if the inference unit 303 infers that a defect has occurred in the molded core, the mold temperature of the mold 160 and the amount of water added during kneading in the kneading tank 110 are adjusted to prevent the occurrence of defects in the molded core. For example, if the inference unit 303 infers that the mold temperature of the mold 160 during the filling of the kneaded sand S deviates from the specified range, the mold thermal controller 161 is controlled to implement temperature control so that the mold temperature falls within the specified range. Furthermore, if the flow rate is inferred to be slower than the specified range, the water supply device 123 is controlled to adjust the amount of water added during kneading in the kneading tank 110, thereby controlling the flow rate to fall within the specified range. Furthermore, the control of molding conditions described here is merely an example. Similarly, the amount of water contained in the kneaded sand S can be adjusted by controlling the core sand supply device 121, the water glass supply device 122, the surfactant supply device 124, and the like, or by adjusting the kneading time. Alternatively, the injection pressure applied by the piston 140 can be adjusted by controlling the operation of the air cylinder 150. By controlling various other molding conditions, core defects can be prevented. The control unit 304 does not necessarily need to be included in the core quality estimation system 200. For example, an operator can operate a separate control device based on the estimation results of the estimation unit 303 to prevent core defects.

[0069] The display unit 305 displays the core quality estimation results estimated by the estimation unit 303, as well as manufacturing conditions for preventing defects determined by the control unit 304. By displaying information and warnings related to the occurrence of core quality defects on the display unit 305, the operator can be notified of this information, allowing the operator to, for example, adjust core molding conditions or suspend molding. While the core quality estimation system 200 is described herein as including the display unit 305 as an information processing device, the display unit 305 may also be provided on the core molding apparatus 100 or on a tablet computer or the like carried by the operator.

[0070] Next, the flow of the estimation process in the core quality estimation system 200 according to the present embodiment, that is, the core quality estimation method will be described. Figure 5 This is a flowchart of the core quality estimation process according to this embodiment. This process is not limited to this. For example, in the core molding device 100, the process starts at the timing of kneading the kneaded sand S in the kneading tank 110 each time a core molding process is performed.

[0071] First, in step S11, the mold temperature information acquisition unit 301 acquires mold temperature information of the mold 160 filled with kneaded sand S. Next, in step S12, the environmental information acquisition unit 302 acquires environmental information related to the surrounding environment of the molding core. Alternatively, the environmental information acquisition unit 302 may acquire environmental information each time the estimation process is performed, or it may acquire environmental information at regular intervals, using the same value for multiple estimation processes. Furthermore, the order of steps S11 and S12 may be changed as appropriate.

[0072] Next, in step S13, the estimation unit 303 estimates the quality of the molded core based on the mold temperature information acquired by the mold temperature information acquisition unit 301 and the environmental information acquired by the environmental information acquisition unit 302. The estimation by the estimation unit 303 is preferably performed before the filling of the kneaded sand S into the mold 160 is completed during the molding process of the core to be estimated. By performing the estimation before the filling of the kneaded sand S is completed, the mold temperature of the mold 160 can be controlled based on the estimation result, thereby preventing defects in the core. Furthermore, the estimation by the estimation unit 303 is preferably performed before the kneading of the kneaded sand S in the kneading tank 110 is completed during the molding process of the core to be estimated. By performing the estimation before the kneading of the kneaded sand S in the kneading tank 110 is completed, the amount of water added during kneading in the kneading tank 110 can be controlled, thereby preventing defects in the core. In addition, in this way, when it is desired to infer the quality of the core earlier than the timing specified in the core molding process, this can be implemented by adjusting the timing of starting this process to suit the analysis time, regressing the analysis software used in the above-mentioned behavior analysis, etc.

[0073] Next, in step S14, the control unit 304 controls the molding conditions of the molded core based on the estimation results from the estimation unit 303, and the present process ends. Although omitted from this process, a step may be added in which the display unit 305 displays the estimation results of the core quality estimated by the estimation unit 303, or steps for preventing poor manufacturing conditions determined by the control unit 304. While this example illustrates an example in which the core molding apparatus 100 performs the estimation process for each core molding process, the estimation process may also be performed once for multiple core molding processes, or in a virtual environment independent of the core molding process. In other words, the timing for starting this process and the timing for the estimation by the estimation unit 303 may be appropriately determined. For example, a configuration may be employed in which the estimation results from the current molding process are used to control the molding conditions in the next molding process.

[0074] According to the above description, the core quality inference system involved in this embodiment has a mold temperature information acquisition unit 301 for acquiring mold temperature information of the mold, an environmental information acquisition unit 302 for acquiring environmental information related to the surrounding environment of the molded core, and an inference unit 303 for inferring the quality of the core based on the mold temperature information and the environmental information. Therefore, the quality of the molded core is inferred based on the environmental information related to the surrounding environment of the molded core. Therefore, the quality of the core can be inferred with high precision based on the change in the amount of water contained in the mixed sand.

[0075] Second embodiment

[0076] Next, a core quality estimation system 200 according to a second embodiment will be described. Since the hardware configuration of the core quality estimation system 200 according to the second embodiment is the same as that of the core quality estimation system 200 according to the first embodiment, a description thereof will be omitted. Furthermore, since the functional configuration also includes common features, only the differences will be described here.

[0077] Figure 6 : is a block diagram showing the functional structure of the core quality estimation system 200 according to this embodiment. Figure 6 As shown, the core quality estimation system 200 includes a mold temperature information acquisition unit 301 , an environment information acquisition unit 302 , an estimation unit 303 , a control unit 304 , a display unit 305 , a correction acceptance unit 306 , and a model updating unit 307 as a functional structure.

[0078] The inference unit 303 of the core quality estimation system 200 according to the second embodiment infers the quality of the molded core using a learned model based on the mold temperature information acquired by the mold temperature information acquisition unit 301 and the environmental information acquired by the environmental information acquisition unit 302. The learned model is machine-learned using supervised data that takes the mold temperature and environmental information as input and outputs quality information related to the core. The learned model is pre-learned in the core quality estimation system 200, an edge device, a server (cloud), or other external device using supervised data that takes the mold temperature and environmental information as input and outputs quality information related to the core, and is then stored in the inference unit 303. The specific learning method will be described in detail later.

[0079] The correction accepting unit 306 accepts corrections to the core quality estimation result from the estimation unit 303. Specifically, if an operator or an inspection device inspects an actual molded core and verifies the quality of the molded core using a device other than the estimation by the estimation unit 303, and if the estimation result from the estimation unit 303 differs from the verification result from the other device, the corrections from the other device are accepted. If the results match, the corrections may be accepted without corrections. Furthermore, in addition to the above, the verification using the other device may also utilize mold temperature information obtained from the temperature sensor 162 when the kneaded sand S is actually filled into the mold 160.

[0080] When the correction accepting unit 306 accepts the correction, the model updating unit 307 updates the learned model stored in the inference unit 303 based on the corrected content. Specifically, the model updating unit 307 relearns the learned model using a learning dataset containing the corrected content, using a learning method described below. This configuration allows the learned model stored in the inference unit 303 to be properly trained, enabling higher-precision inference of core quality.

[0081] Next, the machine learning device 400 according to the second embodiment will be described. A detailed description of the hardware structure of the machine learning device 400 will be omitted here, but similar to the core quality estimation system 200, it includes computer resources typically found in an information processing device. Furthermore, while this description uses an example in which the various components of the machine learning device 400 are implemented by an information processing device independent of the core quality estimation system 200, they can also be implemented as a structure integrated within the core quality estimation system 200, or some or all of their functions can be implemented on an external device such as an edge device or a server (cloud). Specifically, in a factory, etc., where a large number of core molding devices 100 are present, each core molding device 100 and each core quality estimation system 200 in each unit within the factory is connected to a fog server via a network. Furthermore, the fog server provided for each unit is connected to a cloud server via a network. In this configuration, the machine learning device 400 can be installed on either the fog server or the cloud server. In this way, by installing the machine learning device 400 on the server, information can be collected from multiple core molding devices, etc., via the network, and learning can be performed.

[0082] Next, the functional configuration of the machine learning device 400 according to this embodiment will be described. Figure 7 This is a block diagram showing the functional configuration of the machine learning device 400 according to this embodiment.

[0083] The machine learning device 400 includes a mold temperature information acquisition unit 301, an environmental information acquisition unit 302, a quality information acquisition unit 401, and a learning unit 402. As described above, the mold temperature information acquisition unit 301 and the environmental information acquisition unit 302 respectively acquire mold temperature information of the mold 160 filled with kneaded sand S and environmental information related to the surrounding environment of the molding core.

[0084] The quality information acquisition unit 401 acquires quality information related to the quality of the molded cores. Specifically, the mold temperature information acquired by the mold temperature information acquisition unit 301 and the environmental information acquired by the environmental information acquisition unit 302 are correlated to obtain quality information related to the quality of the cores molded under these conditions. While this embodiment describes an example of using mold temperature information of the mold 160 measured during the actual molding of the cores, measured during the filling and heating of the kneaded sand S, as quality information, it is also possible to use information such as measured behavior information during the filling of the kneaded sand S as quality information. Alternatively, an operator may inspect the actual molded cores and use the quality of the cores, input as inspection result information, as quality information.

[0085] The learning unit 402 generates and stores supervisory data, and performs so-called supervised learning to learn the quality of the molded core based on this supervisory data. The supervisory data is input from the mold temperature information acquisition unit 301 and the environmental information acquisition unit 302, and outputs quality information associated with this information. Supervised learning refers to learning that uses supervisory data—that is, a learning dataset consisting of a certain input and the output corresponding to that input—to generate a learned model that infers the output from the input. Various methods used in supervised learning can be used in the present invention.

[0086] Here, taking a neural network as an example, the following describes the learning phase of learning the quality of the molded core through the machine learning device 400 and the application phase of using the learned model to infer the quality of the molded core through the core quality inference system 200.

[0087] First, during the learning phase, a learning dataset is generated. For example, mold temperature information at multiple locations within mold 160, acquired by mold temperature information acquisition unit 301, and ambient temperature, humidity, and pressure information, acquired by environmental information acquisition unit 302, are used as inputs. The mold temperature information of mold 160 during the actual molding of a core under these conditions, when mixed sand S is filled into mold 160, is used as output. A single learning dataset is generated and stored. By preparing multiple data sets representing the relationship between these inputs and outputs, a learning dataset is generated for use in machine learning.

[0088] Next, refer to Figure 8 , illustrating the neural network used in learning. Figure 8 This represents a hierarchical neural network with multiple inputs and outputs. For simplicity, the hidden layer is shown as two layers, but in reality, further layers are possible, and the number of nodes in the input and hidden layers can be arbitrarily changed based on the training dataset.

[0089] For each node of the input layer of such a neural network, the above-mentioned learning data set is used, and input values ​​are input, thereby outputting the mold temperature information of the mold 160 when the mixed sand S is filled into the mold 160 as an inferred value from the output layer. Then, learning is performed in a manner that the inferred value is consistent with the mold temperature information of the mold 160 when the mixed sand S is filled into the mold 160 when the core is actually molded, which is prepared as the learning data set. Specifically, learning is performed using back propagation, etc. until the error of these values ​​converges to less than a predetermined set error, and the weights and biases of the neural network (hereinafter collectively referred to as "weights") are learned. In this way, a neural network (learning completed model) is generated that outputs the mold temperature information of the mold 160 when the mixed sand S is filled into the mold 160, based on the input of the mold temperature information obtained by the mold temperature information acquisition unit 301 and the environmental information obtained by the environmental information acquisition unit 302.

[0090] The neural network that has learned the weights in this way is output to the inference unit 303 of the core quality inference system 200, and the application phase is executed. Figure 8 As shown, the input layer of the learned neural network receives input values ​​of mold temperature information obtained by the mold temperature information acquisition unit 301 and environmental information obtained by the environmental information acquisition unit 302. The output layer then outputs the mold temperature information of mold 160 during the filling of kneaded sand S into mold 160 as an inferred value. The inference unit 303 compares the mold temperature information of mold 160 output by the neural network with a pre-set range to infer the quality of the core. Similarly, a neural network that outputs information on the behavior of kneaded sand S during filling and a neural network that outputs information on the quality of the core can be generated using the same method. Furthermore, while this embodiment describes an example using mold temperature information and environmental information as input values, various molding conditions can also be used as input values, including the type of raw materials used in the kneaded sand S, the raw material ratio (such as the amount of water), the dynamic viscosity of the kneaded sand S, and the injection pressure.

[0091] Next, the flow of the learning phase, that is, the learning method, in the machine learning device 400 according to this embodiment will be described. Figure 9This is a flow chart of the learning process involved in this embodiment. This process begins, for example, when a predetermined amount of learning data sets or more are accumulated in the machine learning device 400. Alternatively, this learning process may be executed to perform re-learning when a predetermined amount of learning data sets or more has been accumulated since the previous learning.

[0092] First, in step S21 , the learning unit 402 reads each data of the learning dataset stored in the machine learning device 400 .

[0093] Next, in step S22 , the learning unit 402 reads the number of nodes in the input layer, hidden layer, and output layer of the neural network, and generates a neural network.

[0094] In step S23, the learning unit 402 uses the read learning dataset to learn the weights of the neural network. Specifically, the learning unit 402 inputs the input values ​​of the learning dataset into the input layer and uses backpropagation to learn the weights of the neural network in such a way that the error between the output values ​​of the neural network and the output values ​​of the learning dataset is minimized. When learning is completed for all the data, the process proceeds to step S24.

[0095] In step S24, the learning unit 402 determines whether the error between the output value of the neural network and the output value of the learning dataset has converged to or below a predetermined set error. If it is determined that it has converged to or below the set error, the process proceeds to step S25 to store the learned weights of the neural network. On the other hand, if it is determined that it has not converged to or below the set error, the learning unit 402 relearns the weights in step S23 and continues learning until the error between the output value of the neural network and the output value of the learning dataset converges to or below the set error.

[0096] Next, the flow of the operation phase in the core quality estimation system 200 according to the present embodiment, that is, the core quality estimation method, will be described. Figure 10 This is a flowchart of the core quality estimation process according to this embodiment. While much of the flowchart is common to the core quality estimation process according to the first embodiment, the differences are described here. This process begins after the aforementioned learning process is completed and the learned model is stored in the estimation unit 303.

[0097] First, in steps S31 and S32, mold temperature information and environmental information are acquired similarly. Then, in step S33, the inference unit 303 uses the learned model, which has undergone the above-described learning process, to estimate the quality of the molded core based on the mold temperature and environmental information. As in the first embodiment, the inference unit 303 preferably performs the inference during the molding process of the core being estimated, before the kneaded sand S is filled into the mold 160, and more preferably before the kneading of the kneaded sand S in the kneading tank 110 is completed. In this way, when it is desired to estimate the quality of the core earlier than a predetermined timing during the core molding process, this can be achieved by selecting a simple structure for the learned model, as described above, and reducing the amount of information used as input.

[0098] Next, in step S34 , the correction receiving unit 306 receives correction of the estimation result of the core quality by the estimation unit 303 .

[0099] Next, in step S35, when the correction accepting unit 306 accepts the correction, the model updating unit 307 updates the learned model stored in the inference unit 303 based on the corrected content, and the present process ends. Furthermore, although omitted from this process, as described in the process of the core quality inference process according to the first embodiment, steps such as controlling the molding conditions by the control unit 304 and displaying the core quality inference results on the display unit 305 may be added.

[0100] In the core quality inference system involved in the present embodiment, in particular, the inference unit 303 infers the quality of the core based on the mold temperature information and the environmental information, and uses a learned model that has been machine-learned using supervised data with the mold temperature information and the environmental information as input and quality information related to the quality of the core as output. Therefore, the quality of the molded core is inferred in advance using a learned model that has learned the relationship between the mold temperature information and the environmental information and the quality information of the core. This has the effect of being able to infer the quality of the core with high precision based on changes in the amount of water contained in the mixed sand.

[0101] Other implementations

[0102] While the first and second embodiments illustrate cases where the core quality estimation system 200 and the machine learning device 400 are provided for each core molding apparatus 100, a configuration may also be employed where the core quality estimation system 200 and the machine learning device 400 are shared among multiple core molding apparatuses 100 and perform estimation and learning of the quality of cores molded by the multiple core molding apparatuses 100. Furthermore, the first and second embodiments illustrate examples where the estimation unit 303 estimates core quality by analyzing the behavior of the kneaded sand S within the mold 160 using CAE, and where the estimation unit 303 estimates core quality using a learned model obtained through machine learning using supervised data. However, the estimation method is not limited to these methods, and various other methods, such as estimation based on precalculated mathematical formulas and multivariate analysis, may also be employed.

[0103] The present invention is not limited to the above-described embodiment, and can be implemented in various forms by appropriately changing the embodiment without departing from the spirit of the invention.

Claims

1. A core quality estimation system for estimating the quality of a core obtained by moving a kneading tank containing kneaded sand to a mold, filling the mold with the kneaded sand from the moved kneading tank, and heating the mold to form a mold, the core quality estimation system comprising: A computer that performs the following operations: obtaining mold temperature information of the mold from a temperature sensor provided on the mold; Obtaining environmental information related to the surrounding environment of the molding core; Based on the mold temperature information and the environmental information, mold temperature information of the mold during filling or heating of the kneaded sand and behavior information of the kneaded sand during filling are obtained by performing a behavior analysis of the kneaded sand in the mold using CAE. If the mold temperature of a portion of the mold in contact with the kneaded sand during filling or heating deviates from a predetermined temperature range, or if the flow rate or energy in the behavior information deviates from a predetermined range, it is inferred that a defect has occurred in the core. as well as If it is inferred based on the inference result that a defect has occurred in the molded core, the molding conditions during the kneading of the kneaded sand, including the amount of water input, the amount of core sand input, the amount of surfactant input, and the kneading time, and the molding conditions during the filling of the kneaded sand into the mold, including the mold temperature of the mold, are controlled.

2. The core quality inference system according to claim 1, characterized in that: The environmental information includes at least one of humidity information and air pressure information of the surrounding environment.

3. The core quality inference system according to claim 1, characterized in that: The computer acquires the environmental information based on the position information of the molding core.

4. The core quality inference system according to any one of claims 1 to 3, characterized in that: The computer infers the quality of the core based on the mold temperature information and the environmental information and uses a learned model, wherein the learned model is machine-learned using supervised data with the mold temperature information and the environmental information as input and quality information related to the quality of the core as output.

5. The core quality inference system according to claim 4, characterized in that: The computer also performs the following operations: To make corrections to the inferred results; and When the correction is accepted, the learned model is updated based on the corrected content.

6. A method for estimating the quality of a core, wherein the core is obtained by moving a kneading tank containing kneaded sand to a mold, filling the mold with the kneaded sand from the moved kneading tank, and heating the mold to form the mold, the method comprising: obtaining mold temperature information of the mold from a temperature sensor provided on the mold; Obtaining environmental information related to the surrounding environment of the molding core; Based on the mold temperature information and the environmental information, mold temperature information of the mold during filling or heating of the kneaded sand and behavior information of the kneaded sand during filling are obtained by performing a behavior analysis of the kneaded sand in the mold using CAE. If the mold temperature of a portion of the mold in contact with the kneaded sand during filling or heating deviates from a predetermined temperature range, or if the flow rate or energy in the behavior information deviates from a predetermined range, it is inferred that a defect has occurred in the core. as well as If it is inferred based on the inference result that a defect has occurred in the molded core, the molding conditions during the kneading of the kneaded sand, including the amount of water input, the amount of core sand input, the amount of surfactant input, and the kneading time, and the molding conditions during the filling of the kneaded sand into the mold, including the mold temperature of the mold, are controlled.

7. A non-transitory storage medium storing one or more computer-executable instructions for inferring the quality of a core obtained by moving a kneading tank containing kneaded sand to a mold, filling the mold with the kneaded sand from the moved kneading tank, and heating the mold to form a mold, wherein the non-transitory storage medium stores one or more computer-executable instructions for inferring the quality of a core obtained by moving a kneading tank containing kneaded sand to a mold, filling the mold with the kneaded sand from the moved kneading tank, and heating the mold to form a mold, wherein the instructions cause the one or more computers to perform the following functions: obtaining mold temperature information of the mold from a temperature sensor provided on the mold; Obtaining environmental information related to the surrounding environment of the molding core; Based on the mold temperature information and the environmental information, mold temperature information of the mold during filling or heating of the kneaded sand and behavior information of the kneaded sand during filling are obtained by performing a behavior analysis of the kneaded sand in the mold using CAE, and when the mold temperature of a portion of the mold in contact with the kneaded sand during filling or heating deviates from a predetermined temperature range, or when a flow rate or energy in the behavior information deviates from a predetermined range, it is inferred that a defect has occurred in the core; and If it is inferred based on the inference result that a defect has occurred in the molded core, the molding conditions during the kneading of the kneaded sand, including the amount of water input, the amount of core sand input, the amount of surfactant input, and the kneading time, and the molding conditions during the filling of the kneaded sand into the mold, including the mold temperature of the mold, are controlled.

Images