Program, information processing information, and information processing device

Abstract

The state of the mold or the state of the injection molded product using the mold is appropriately estimated. [Solution] The program of this embodiment causes a computer to execute the following process: inputting time-series first heat-related information via an input layer to a heat reservoir layer having multiple virtual nodes that hold time-series second heat-related information observable at each of multiple positions as time-series first heat-related information input to the molding equipment during molding using the mold is propagated to multiple positions of the molding equipment, based on characteristics including heat-related characteristics of the molding equipment that uses the mold; outputting output values ​​via the output layer that are obtained by linearly combining the time-series second heat-related information held by the multiple virtual nodes in the heat reservoir layer using coupling coefficients between the multiple virtual nodes based on the first heat-related information; and evaluating the state of the molded product molded by the molding equipment based on the output values.

Description

[Technical Field]

[0001] The present invention relates to a program, information processing information, and an information processing device. [Background technology]

[0002] Conventionally, molds are used to repeatedly manufacture many products. For these molds, a predictive simulation system is known that performs advance optimization of molds and molding conditions, focusing on filling, pressure holding, cooling, etc. (See, for example, Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] International Publication No. 2019 / 113166 Summary of the Invention [Problem to be solved by the invention]

[0004] However, the flow phenomena that occur when pouring polymers or other molding materials into a mold are complex, and creating the optimal flow path to create the optimal mold is an artisanal process, with no systematic knowledge accumulated. Furthermore, using a mold multiple times results in repeated temperature fluctuations, and various unforeseen factors can lead to the production of many defective products and failures of the injection molding equipment for high-temperature molds. Currently, countermeasures for failures are based on past experience and involve shutting down the equipment early. Furthermore, for defective products, measures are taken to shut down the equipment when defects are actually detected through visual inspection, etc., but both failures and defects result in losses because the injection molding equipment cannot be used until the appropriate time.

[0005] The present invention has been made in consideration of the above circumstances, and aims to provide a program, an information processing method, and an information processing device that can appropriately estimate the state of a mold or the state of a molded product made using the mold. [Means for solving the problem]

[0006] The program of this embodiment inputs time-series first heat-related information during molding using a mold, which is input to the molding equipment based on characteristics including heat-related characteristics of the molding equipment, to multiple positions on the molding equipment, and propagates the information to multiple positions on the molding equipment.The program then inputs the time-series first heat-related information via an input layer to a thermal reservoir layer having multiple virtual nodes that hold time-series second heat-related information observable at each of the multiple positions.Based on the input time-series first heat-related information, the program linearly combines the time-series second heat-related information held by the multiple virtual nodes in the thermal reservoir layer using coupling coefficients between the multiple virtual nodes, and outputs output values ​​via an output layer.The program then executes a process of evaluating the state of the mold used in the molding equipment or the state of the molded product molded by the molding equipment based on the output value. [Effects of the Invention]

[0007] According to the present invention, it is possible to appropriately estimate the state of a mold or the state of a molded product made using a mold. [Brief explanation of the drawings]

[0008] [Figure 1A] FIG. 10 is a diagram showing an example of using TRC in a thermal simulation at the mold design stage in this embodiment. [Figure 1B] 10A and 10B are diagrams showing an example of using a TRC in a molding machine at the mass production stage using a mold according to this embodiment. [Figure 2] FIG. 1 is a diagram illustrating an example of a configuration of an information processing apparatus according to an embodiment of the present invention. [Figure 3] FIG. 2 is a diagram illustrating an example of the configuration of a thermal reservoir model section. [Figure 4] FIG. 10 is a diagram showing an example of training data used for learning a thermal reservoir model section. [Figure 5A] 10 is a diagram showing an example of a learning process of a thermal reservoir model unit 60. FIG. [Figure 5B] 10 is a diagram showing an example of a learning process of a thermal reservoir model unit 60. FIG. [Figure 6A] FIG. 10 is a diagram illustrating an example of evaluation of an evaluation target based on the output of an output layer. [Figure 6B] FIG. 10 is a diagram illustrating an example of evaluation of an evaluation target based on the output of an output layer. [Figure 7] FIG. 10 is a diagram illustrating an example of a learning process performed by an information processing device. [Figure 8] FIG. 10 is a diagram illustrating an example of an inference process performed by an information processing device. DETAILED DESCRIPTION OF THE INVENTION

[0009] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described below with reference to the drawings showing embodiments thereof. First, an example of implementing thermal reservoir computing (TRC) of the disclosed technology will be described with reference to Figures 1A and 1B.

[0010] Fig. 1A shows an example of using TRC in a simulation at the mold design stage in this embodiment. In the example shown in Fig. 1A, the mold material, the temperature of the mold manufacturing equipment during mold manufacturing, the mold material, the temperature of the material (e.g., polymer) flowing into the mold, the flow rate of the material, the mold temperature, etc. are set, and a simulation including a flow simulation of the material flowing into the mold (referred to as "thermal simulation" in the example shown in Fig. 1) is performed.

[0011] A thermal simulation is performed, a mold with good simulation results is prototyped, and a manufacturing test is performed using the prototype mold. If the molded product produced (e.g., injection molded) using the mold in this manufacturing test satisfies certain success conditions (e.g., meets certain standards, the manufacturing error of multiple molded products is less than a certain value), it is judged to be a success, but if it is judged to be a failure, the parameters of the thermal simulation are changed, and the thermal simulation, mold prototype, manufacturing test, and judgment are repeated again.

[0012] In this embodiment, in the above-described iterative process of the design stage, TRC analysis is performed on the thermal simulation. For example, in TRC analysis, the mold and molding material are regarded as a single reservoir element, and the state of the mold is determined from temperature changes at multiple points. In this case, parameters from several successful and unsuccessful thermal simulations are used as training data, and the TRC analysis is trained to determine the quality of the mold prototype before it is actually produced.

[0013] As a result, according to this embodiment shown in FIG. 1A, it is possible to reduce the number of mold prototypes that are actually manufactured, thereby reducing the cost of mold prototypes and shortening the design lead time.

[0014] FIG. 1B shows an example of using a TRC in a mass-production molding machine using a mold according to this embodiment. In the example shown in FIG. 1B, when material is poured into a mold to produce a molded product, the heat of the material affects the mold and molding machine. Therefore, by performing edge computing using the nonlinear heat conduction phenomenon of materials, including the mold and its internal materials, it is possible to estimate the state of the molding machine using the mold itself as a reservoir computing element.

[0015] For example, as heat from the material is continuously applied to the mold, the state of the mold may be evaluated based on temperature changes at multiple points in the molding equipment, including the mold, and ultimately, whether or not there is an abnormality in the molded product produced by this mold may be determined. This can prevent product defects when molding products using a mold, reducing the reject rate and disposal costs. It can also detect invisible problems and predict failures of mold molding equipment.

[0016] For example, in this embodiment, by measuring the temperature of a temperature sensor attached to the mold or detecting the temperature of the mold surface using external thermography, edge computing can be performed using the nonlinear heat conduction phenomenon of materials including the mold and internal materials, making it possible to estimate the state of the mold molding equipment using the mold itself as a reservoir computing element.

[0017] 2 is a diagram showing an example of the configuration of an information processing device 50 according to this embodiment. The information processing device 50 includes a control unit 51 that controls the entire device, an input / output unit 52, a memory 53, a learning processing unit 54, an evaluation unit 55, a storage unit 56, and a thermal reservoir model unit 60. The information processing device 50 may be configured with multiple devices by distributing the processing functions.

[0018] The storage unit 56 may be configured with, for example, a hard disk or a semiconductor memory, and stores a computer program 57 (program product) and required information. The computer program 57 may be downloaded from an external device and stored in the storage unit 56. Alternatively, the computer program 57 recorded on a recording medium (for example, an optically readable disk storage medium such as a CD-ROM) may be read by a recording medium reading unit and stored in the storage unit 56.

[0019] The control unit 51 is configured by incorporating a required number of CPUs (Central Processing Units), MPUs (Micro-Processing Units), GPUs (Graphics Processing Units), etc. The control unit 51 can execute processing defined by a computer program 57. In other words, the processing by the control unit 51 is also processing by the computer program 57. By executing the computer program 57, the control unit 51 can perform the functions of the learning processing unit 54 and the evaluation unit 55.

[0020] The input / output unit 52 acquires data necessary for processing by the information processing device 50 and outputs data obtained as a result of processing by the information processing device 50. The input / output unit 52 may also include a user interface such as a keyboard, a mouse, a display panel, or a touch panel. The input / output unit 52 can acquire the first heat-related information in time series input to the equipment (mold manufacturing equipment or mold molding equipment) 200 as the target object.

[0021] The memory 53 can be configured with semiconductor memory such as SRAM (Static Random Access Memory), DRAM (Dynamic Random Access Memory), flash memory, etc. A computer program 57 can be loaded into the memory 53, and the control unit 51 can execute the computer program 57.

[0022] The learning processing unit 54 performs learning processing for the thermal reservoir model unit 60. Details of the learning processing will be described later.

[0023] The thermal reservoir model unit 60 is a reservoir computing model and includes an input layer 61, a thermal reservoir layer 62, a coupling layer 63, and an output layer 64. The thermal reservoir model unit 60 can be implemented by hardware such as an FPGA (Field Programmable Gate Array). Under the control of the control unit 51, the input layer 61 receives first heat-related information (collectively referred to as "input data") in time series input to the device 200, acquired via the input / output unit 52. The location of the device 200 where the first heat-related information is input (also referred to as "input point") is not particularly limited, and may be a position where heat is applied to the device itself during the design stage, or an opening into which a heat-generating material is poured during mass production. The first heat-related information is nonlinearly transformed by the thermal reservoir layer 62 and mapped to a high-dimensional feature space. The mapping to the feature space is represented by the states of multiple virtual nodes in the thermal reservoir layer 62. The coupling layer 63 linearly combines the state values ​​of the multiple virtual nodes in the thermal reservoir layer 62 using coupling coefficients (weighting coefficients). The output layer 64 outputs the linearly combined output value. The output value output by the thermal reservoir model unit 60 can be used to evaluate the temperature-related state of the device 200. Details of the thermal reservoir model unit 60 will be described later.

[0024] The evaluation unit 55 evaluates the temperature-related condition of the equipment 200 or the mold based on the output value output by the thermal reservoir model unit 60. The evaluation unit 55 will be described in detail later.

[0025] The information processing device 50 includes thermal reservoir computing, which uses the device 200 or a material as a thermal reservoir to achieve reservoir computing. In this specification, a solid, liquid, or gas capable of transmitting heat is referred to as a thermal reservoir. Below, the thermal reservoir will be explained using a solid as an example. Note that the information processing device 50 and the device 200 may be configured as an integrated unit.

[0026] The device 200 according to this embodiment includes, for example, a heating section for heating the material, a material supply section for supplying the material to be poured into a mold, a mold, a support section for supporting the mold, and a receiving section for receiving the molded object. The entire device is configured as a thermal reservoir in thermal reservoir computing. In this configuration, the device 200 functions as a thermal reservoir, and a nonlinear mapping between thermal parameters (e.g., material temperature) obtained through thermal simulation, or the temperature of the material poured into the mold molding device, and the temperature response is realized through reservoir computing. Here, the thermal reservoir refers to the entire object through which heat can propagate. It has the function of storing and propagating input time-series heat-related information based on the object's physical properties, such as thermal conductivity, thermal diffusivity, and shape. In other words, the device 200 maintains the time-series changes in the heat distribution generated within the device as its internal state. At this time, temperature information observed at multiple locations within the device or physical quantities that change depending on temperature changes function as a group of virtual nodes, forming the internal state space of reservoir computing.

[0027] In the design stage, the entire equipment, including the mold material and mold manufacturing equipment used in thermal simulation, is treated as a thermal reservoir. Each of these has different material properties, and the temperature distribution that occurs depending on the temperature of the mold material and the mold manufacturing equipment responds nonlinearly. This temperature response functions as the internal state of the reservoir in thermal reservoir computing.

[0028] On the other hand, in the mass production stage, the entire equipment, including the mold itself, the material poured into the mold, and the molding equipment, constitutes a thermal reservoir. The temperature distribution that occurs depending on the temperature of the material poured into the mold and the molding equipment responds nonlinearly. This temperature response functions as the internal state of the reservoir in thermal reservoir computing.

[0029] The device 200 includes at least one temperature sensor 201. The temperature sensor 201 detects the temperature at a predetermined position in the device 200, and the detected value is associated as a virtual node value. That is, each temperature sensor 201 corresponds to a virtual node and functions as an interface for externally observing the internal temperature state of the device 200. Preferably, the temperature sensors 201 are provided at multiple positions on the device 200, and the temperatures at the multiple positions are detected.

[0030] In this embodiment, the evaluation unit 55 evaluates the temperature-related conditions inside the device 200 based on the output values ​​obtained by thermal reservoir computing. The evaluation unit 55 analyzes the temperature distribution and time series changes of the thermal reservoir (i.e., the entire device 200) and can detect, for example, a steepening of the temperature gradient, a local temperature rise, or thermal non-uniformity that affects the quality of the mold. When the evaluation target exceeds a predetermined threshold, it is determined to be abnormal, and the condition of the mold is quantitatively evaluated based on the error between the output value and the correct value.

[0031] Even when the environmental conditions (such as outside temperature and humidity) in which the device 200 is installed are different, the device 200 can manufacture molds or perform molding using molds that are optimized for each environment. In this embodiment, in order to address variations in thermal response due to differences in environmental conditions, multiple models may be generated for each environmental information. That is, multiple thermal reservoir computing models may be switched depending on environmental information such as region and climate. This makes the device less susceptible to environmental changes and enables it to predict near-future temperature distribution and proactively control the occurrence of mold-related defects.

[0032] 3 is a diagram showing an example of the configuration of the thermal reservoir model unit 60. When first heat-related information in time series is input, the input layer 61 inputs the first heat-related information in time series to the thermal reservoir layer 62.

[0033] The thermal reservoir layer 62 includes multiple virtual nodes. The thermal reservoir layer 62 is realized as a physical phenomenon of heat propagation in a thermal reservoir. That is, the thermal reservoir layer 62 includes multiple virtual nodes that hold time-series second heat-related information (e.g., including temperature data or physical quantities that change depending on temperature changes) that can be observed at each of multiple positions (observation points) of the thermal reservoir by propagating time-series first heat-related information input to the thermal reservoir based on the thermal conductivity or thermal diffusivity (heat-related properties) of the thermal reservoir capable of propagating heat, or the shape (properties) of the thermal reservoir. The construction and design of the thermal reservoir layer 62 can be performed, for example, by determining physical parameters (such as the dimensions, shape, thermal conductivity (or thermal diffusivity), and material of the thermal reservoir) according to the learning object or learning task based on simulation experiments.

[0034] The coupling layer 63 linearly combines the time-series second heat-related information held by the multiple virtual nodes in the thermal reservoir layer 62 using coupling coefficients between the multiple virtual nodes. The coupling layer 63 and the output layer 64 are also collectively referred to as a readout layer.

[0035] The output layer 64 outputs an output value linearly combined in the combination layer 63. The number of dimensions of the output layer 64 can be set arbitrarily.

[0036] Next, thermal reservoir computing, that is, a learning method for the thermal reservoir model unit 60, will be described.

[0037] FIG. 4 is a diagram showing an example of training data used for learning of the thermal reservoir model unit 60. The first heat-related information of various states in time series according to the device 200 shown in FIG. 4 is input. In the example shown in FIG. 4, for convenience, states are designated as 1, 2, ..., m. For example, state 1 corresponds to time point 1 (time t1=t0), state 2 corresponds to time point 2 (time t2=t0+dt), ... state m corresponds to time point m (time t m=t0+m·dt). The first heat-related information may include, for example, environmental information. The first heat-related information may also include temperature data or heat data. Learning input data collected corresponding to states 1, 2, ..., m of the device 200 are represented by vectors G1, G2, Gm. When the state of the device 200 is state 1, 2, ..., m, the correct values ​​(true values) of the output values ​​representing the state of the device 200 are represented by vectors D1, D2, ..., Dm. The vectors D1, D2, ..., Dm can be training data.

[0038] Furthermore, vector S1 represents values ​​indicating the states of the multiple virtual nodes in the thermal reservoir layer 62 when learning input data G1 is input to the thermal reservoir layer 62. Vector S1 corresponds to the second heat-related information in time series held by the multiple virtual nodes, and includes, for example, temperature data related to the mold or a physical quantity that changes depending on temperature changes. Y1 represents a vector of output values ​​output by the coupling layer 63 when vector S1 is input to the coupling layer 63. Furthermore, vector S2 represents values ​​indicating the states of the multiple virtual nodes in the thermal reservoir layer 62 when learning input data G2 is input to the thermal reservoir layer 62, and Y2 represents a vector of output values ​​output by the coupling layer 63 when vector S2 is input to the coupling layer 63. Similarly, below, the values ​​indicating the state of multiple virtual nodes in the thermal reservoir layer 62 when learning input data Gm is input to the thermal reservoir layer 62 are represented by vector Sm, and the vector of output values ​​output by the coupling layer 63 when vector Sm is input to the coupling layer 63 is represented by Ym.

[0039] In this embodiment, the correct value is defined based on the temperature of each location when an appropriate mold is manufactured in the design stage, or when the mold stably molds an appropriate molded product in the mass production stage. For example, when the device 200 is manufacturing an appropriate mold or molding an appropriate molded product using the mold, if temperature information observed at multiple locations corresponding to the internal structure does not change before and after the next manufacturing or molding using the mold, the temperature distribution at that time corresponds to the correct value.

[0040] More specifically, let's consider a case where a molded product is correctly molded using a mold at a certain time t1, and a new molded product is molded at the next time t2. At this time, if temperatures T1 and T2 at each observation point are equal, i.e., if the temperature change dT = T2 - T1 = 0, it is determined that no temperature disturbance has occurred. In this state, it is possible to mold a stable molded product using a mold, and this temperature distribution represents the ideal correct state.

[0041] Note that actual temperature information is composed of multiple temperature values ​​observed by multiple temperature sensors 201 installed inside or on the surface of the device 200, and is therefore treated as a vector quantity. Therefore, the correct value can be expressed as any linear combination of temperature vectors at each observation point. That is, not only when the temperature is completely constant at all observation points, but also when a linear combination of multiple temperature distributions allows a mold or molded product to be properly manufactured or molded, the temperature distribution is treated as the correct value. Thus, the correct value in this embodiment refers to a thermal equilibrium state corresponding to the manufacturing of the mold in the design stage and the stability of molding using the mold in the mass production stage, and is used as a criterion for minimizing the error with the output value vector in the learning stage.

[0042] Figure 5 illustrates an example of the learning process of the thermal reservoir model unit 60. A typical recurrent neural network is a complex, nonlinear, and high-dimensional model, requiring large amounts of data and long processing times for learning. In thermal reservoir computing (TRC), the thermal reservoir layer 62 is implemented as the entire device, i.e., a thermal reservoir capable of heat propagation (see Figure 3). This allows the nonlinear temperature response caused by heat input (applied) to the device to be realized as a high-dimensional, nonlinear model, independent of the neural network. The thermal reservoir layer 62 randomly sets the states of multiple virtual nodes according to the thermal reservoir's thermal properties (e.g., thermal conductivity, thermal diffusivity), eliminating the need to adjust parameters within the thermal reservoir layer 62 through learning. In other words, the output of the output layer 64 is given by the coupling layer 63 as a linear combination of the state of the thermal reservoir layer 62 (time-series temperature changes at observation points inside the device). The coupling coefficients (weighting coefficients) of the coupling layer 63 may be learned using a simple algorithm such as linear regression so that the output of the output layer 64 is a value that is the same as or close to the correct value.

[0043] As shown in FIG. 5A, the virtual nodes of the thermal reservoir layer 62 are denoted as s1, s2, s3, ..., sj. Here, q is the number of observation points, N is the number of discrete values, and j = q·N. The virtual node s corresponds to the second heat-related information (such as temperature data) in time series observed at multiple positions (observation points) of the device 200. The output values ​​(output nodes) of the output layer 64 are denoted as y1, y2, ..., yr. The coupling coefficients from the virtual node s1 to the output values ​​y1, y2, ..., yr are denoted as w11, w12, ..., w1r. Similarly, the coupling coefficients from the virtual node sj to the output values ​​y1, y2, ..., yr are denoted as wj1, wj2, ..., wjr.

[0044] As shown in Figure 5B, if the output value vector is represented by Y, the virtual node value vector corresponding to the temperature sensed at each location on the display or projector is represented by S, and the coupling coefficient matrix is ​​represented by W, then in the formula Y = S W, the coupling coefficients of the coupling coefficient matrix W can be calculated so that the output value vector Y approaches the correct value vector D.

[0045] As described above, the learning processing unit 54 functions as a determination unit, and acquires training data including time-series first heat-related information (heat data or temperature data) caused by heat applied to the device 200 and correct values ​​for the output value. Based on the acquired training data, when the time-series first heat-related information is input to the thermal reservoir layer 62, the learning processing unit 54 determines the coupling coefficients between the multiple virtual nodes so that the output value obtained by linearly combining the time-series second heat-related information held by the multiple virtual nodes in the thermal reservoir layer 62 using the coupling coefficients between the multiple virtual nodes approaches the correct value, thereby learning the thermal reservoir model unit 60. Note that the correct value in this case is expressed as a linear combination of temperature vectors at multiple observation points and means a thermal equilibrium state in which no defective products are produced.

[0046] As described above, learning can be performed using a simple configuration such as the coupling layer 63 and a simple algorithm called linear regression, so a huge amount of data is not required, the processing time is relatively short, operation is possible with low power consumption, and learning in real time is also possible. This makes the information processing device 50 of this embodiment applicable to edge computing such as IoT devices that predict temperature changes.

[0047] The output value y and the output value vector Y are not physical quantities with specific units, but are numerical results of a calculation. It is necessary to determine how this calculation result relates to the state of the thermal reservoir, particularly temperature-related conditions, such as the stability of the device 200 and thermal non-uniformities associated with mold manufacturing or molding.

[0048] FIG. 6 illustrates an example of an evaluation of an evaluation target based on the output of the output layer 64. The evaluation unit 55 evaluates the temperature-related condition of the mold used in the device 200, which is the evaluation target, based on the output values ​​(including a prediction of near-future temperature distribution) output by the output layer 64. The condition of the evaluation target includes a steepening of the temperature gradient, a local temperature rise, or the degree of thermal non-uniformity affecting the mold. FIG. 6A illustrates a case in which the output layer 64 outputs multiple output values. Multiple evaluations 1, 2, ..., r are set corresponding to each output value. If the output value is equal to or greater than a predetermined threshold, the evaluation corresponding to that output value can be determined as the evaluation result of the mold of the device 200, which is the target. The evaluations 1, 2, ..., r may be appropriately determined. Each of the evaluations 1, 2, ..., r may be an evaluation of a different condition (e.g., the temperature of the mold, the quality of the molded product, etc.) or may be a prediction of the degree of deterioration of a specific condition over time (e.g., a prediction of the degree of deterioration of the mold).

[0049] FIG. 6B shows a case where the output layer 64 outputs one output value y. For example, if the output value y is equal to or greater than a threshold, the mold of the device 200 is determined to be normal (in a state of thermal equilibrium), and if the output value y is less than the threshold, the evaluation object can be determined to be abnormal (possibility of temperature disturbance occurring). Based on this evaluation result, the evaluation unit 55 can reduce the defective product rate by stopping mold manufacturing or molding using the mold in advance in accordance with the predicted change in temperature distribution. Note that the evaluation method is not limited to the example of FIG. 6. Here, Example 1 in the mass production stage of molds and Example 2 in the design stage will be described.

[0050] <Example 1> The control unit 51 inputs the first heat-related information in time series to a heat reservoir layer 62 having a plurality of virtual nodes that hold second heat-related information in time series that can be observed at each of a plurality of positions by propagating the first heat-related information in time series during molding using a mold that is input to the molding equipment based on characteristics including heat-related characteristics of the molding equipment (equipment 200) that uses a mold to a plurality of positions of the molding equipment.

[0051] For example, the control unit 51 inputs first heat-related information, which is heat data or temperature data of the material flowing into the mold in a time series, to the input layer 61, and inputs it to the heat reservoir layer 62 via the input layer 61.

[0052] Based on the input time-series first heat-related information, the control unit 51 linearly combines the time-series second heat-related information held by the multiple virtual nodes in the heat reservoir layer 62 using coupling coefficients between the multiple virtual nodes, and outputs the output value via the output layer 64.

[0053] For example, the control unit 51 linearly combines the time-series second heat-related information held by virtual nodes corresponding to multiple observation points of the molding equipment using a coupling coefficient in the coupling layer 63, and outputs the linearly combined output value via the output layer 64.

[0054] The evaluation unit 55 evaluates the state of the molded product molded by the molding equipment or the state of the mold used to mold the molded product based on the output values ​​output by the output layer 64. For example, by learning data for normal and abnormal output values ​​in advance using a learning model, it is possible to set thresholds for normal and abnormal output values.

[0055] The above process makes it possible to properly estimate the condition of the mold or the condition of the molded product. As a result, it is possible to reduce the rate of defective products and the disposal cost. It is also possible to detect failures of the mold molding device in advance by simulation, etc.

[0056] The learning processing unit 54 determines the coupling coefficients so that when the time-series first heat-related information is input to the heat reservoir layer based on training data including the time-series first heat-related information and the correct value of the output value during molding using a mold, the output value obtained by linearly combining the time-series second heat-related information held by the multiple virtual nodes using the coupling coefficients between the multiple virtual nodes approaches the correct value.

[0057] For example, the learning processing unit 54 uses the first heat-related information when manufacturing defective products and the first heat-related information when manufacturing normal products as training data. Furthermore, the learning processing unit 54 performs thermal reservoir computing using this training data, thereby enabling the coupling coefficient of the coupling layer 63 to be appropriately set.

[0058] According to the above processing, the thermal reservoir model can be trained using training data, and the output values ​​output from this thermal reservoir model can be used to more appropriately estimate the state of the mold or the state of the molded product using the mold.

[0059] The input / output unit 52 acquires an image of the molded product formed by the mold. For example, an imaging device is provided at a position where an injection-molded product injection-molded using the mold is output, and the input / output unit 52 acquires an image of the molded product captured by this imaging device.

[0060] The evaluation unit 55 may evaluate the state of the molded product based on the image of the molded product and the output value output by the output layer 64. For example, the accuracy of determining whether or not an abnormality has occurred in the mold may be improved by having the evaluation unit learn normal and abnormal images of the molded product.

[0061] The evaluation unit 55 may also evaluate whether the molded product is defective. For example, the evaluation unit 55 may determine whether the molded product is defective based on the output value, or the output value and an image of the molded product. The output value may be determined based on a threshold value, or the image of the molded product may be determined based on a learning model that has learned normal and abnormal images. This makes it possible to determine whether the molded product is defective by focusing on the molded product.

[0062] <Example 2> As described above, thermal reservoir computing can also be applied in the mold design stage. Therefore, in the mold design stage, the control unit 51 may use information on the mold temperature and / or the mold material temperature input in a simulation when prototyping the mold as the other first heat-related information. Note that in Example 2, the configuration of the thermal reservoir model unit 60 is the same as in Example 1, but the input thermal data / temperature data and reservoir elements are different from those in Example 1, so the term "other" is used.

[0063] For example, the control unit 51 may use, as other first heat-related information, time-series data of the temperature of the mold manufacturing equipment during mold manufacturing, which is set by simulation including the flow simulation described above, and / or the temperature of the material (e.g., polymer) flowing into the mold, and / or the temperature of the mold, which are set during the mold design stage.

[0064] In addition, the control unit 51 may define the other thermal reservoir layer as a thermal reservoir layer having a plurality of other virtual nodes that hold other second heat-related information in a time series that can be observed at each of a plurality of positions as a result of heat propagating to a plurality of positions based on characteristics including heat-related characteristics of the equipment that manufactures the mold.

[0065] The control unit 51 may output, via another output layer, an output value obtained by linearly combining, based on other time-series first heat-related information input to another input layer, other time-series second heat-related information held by multiple other virtual nodes in another heat reservoir layer using coupling coefficients between multiple other virtual nodes in another coupling layer.

[0066] The evaluation unit 55 may evaluate the state of the mold used to mold the molded product based on the output value. For example, the evaluation unit 55 may evaluate whether any abnormality has occurred in the state of the mold due to heat based on other second heat-related information from multiple observation points on the mold.

[0067] The learning processing unit 54 may determine the coupling coefficients based on training data including the time-series other first heat-related information and the correct value of the output value in the simulation so that, when the other first heat-related information is input to the other heat reservoir layer, the output value obtained by linearly combining the time-series other second heat-related information held by the multiple virtual nodes using the coupling coefficients between the multiple other virtual nodes approaches the correct value. The correct value is, for example, a value that represents a state in which the heat applied to the mold is stable and a normal molded product is being molded using the mold.

[0068] The evaluation unit 55 may also include a function for evaluating whether or not the mold will mold a defective product. This makes it possible to predict when a defective product will be produced using the mold, and to perform maintenance of the molding equipment at an appropriate time.

[0069] Although the above-described example has been described in which the thermal reservoir computing is applied in the mold design stage and the mass production stage, it does not have to be applied in both stages and may be applied in either one stage. That is, the above-described thermal reservoir computing may be applied only in the design stage or only in the mass production stage.

[0070] FIG. 7 is a diagram illustrating an example of a learning process by the information processing device 50. For convenience, the following description will be given assuming that the control unit 51 is the subject of the process. The device 200 applies heat to a material or to the device itself. The control unit 51 acquires training data including input data (time-series first heat-related information) and correct values ​​for output values ​​of the thermal reservoir resulting from the application of heat to the device 200 (S11). The time-series first heat-related information may include, for example, heat data and temperature data. Then, the control unit 51 (or the learning processing unit 54) inputs input data (thermal data for learning near-future temperature responses) to the thermal reservoir layer 62 via the input layer 61 based on the acquired training data (S12).

[0071] The control unit 51 (or the learning processing unit 54) adjusts the coupling coefficients (weighting coefficients) of the coupling layer 63 so that the output value output by the output layer 64 approaches the correct value (S13). Here, the correct value is the temperature at which a mold can be stably manufactured or at which a normal molded product can be molded using the mold, for example, a value corresponding to a thermal equilibrium state. Since thermal reservoir computing uses a simple algorithm such as linear regression for learning, the learning cost is low and real-time learning is possible. The control unit 51 determines whether the difference between the output value and the correct value (the error in the predicted temperature disturbance) is within an allowable range (S14). If it is not within the allowable range (NO in S14), the control unit 51 continues processing from step S12 onwards.

[0072] If the difference between the output value and the correct value is within the allowable range (YES in S14), the control unit 51 (or the learning processing unit 54) stores the adjusted coupling coefficient (weight of the model for predicting the temperature in the near future) in the storage unit 56 (S15), and ends the processing. This coupling coefficient is used in the subsequent inference processing (FIG. 8) to predict the temperature distribution in the near future from the input heat data / temperature data, and to determine the appropriate timing for maintenance of the device 200, mold replacement, etc.

[0073] 8 is a diagram showing an example of inference processing by the information processing device 50. The control unit 51 acquires first heat-related information in a time series based on a second time point different from the first time point (S21). The first heat-related information at the second time point is, for example, the temperature of a material being poured into a mold. Then, the control unit 51 inputs the first heat-related information in a time series related to the second time point acquired via the input layer 61 to the heat reservoir layer 62 (S22).

[0074] The control unit 51 (or evaluation unit 55) acquires the output value (including a prediction of the temperature distribution in the near future) output by the output layer 64 when the first heat-related information at the second time point is input to the equipment 200 (S23), and evaluates the temperature-related state (steepening of the temperature gradient, thermal non-uniformity, etc.) of the equipment 200 being evaluated based on the acquired output value (S24).

[0075] As described above, the control unit 51 inputs time-series first heat-related information to a thermal reservoir layer having multiple virtual nodes that hold time-series second heat-related information observable at each of multiple positions by propagating the time-series first heat-related information input to the thermal reservoir to multiple positions in the thermal reservoir based on characteristics including heat-related characteristics of the thermal reservoir (the entire device) through which heat can be propagated, and reads out an output value (predicted thermal state) that is output by linearly combining the time-series second heat-related information held by the multiple virtual nodes using coupling coefficients between the multiple virtual nodes. This makes it possible to appropriately estimate the state of the mold or the state of the molded product using the mold.

[0076] (Appendix 1) On the computer, inputting the time-series first heat-related information into a thermal reservoir layer having a plurality of virtual nodes that hold time-series second heat-related information observable at each of a plurality of positions of the molding equipment, the time-series first heat-related information being input into the molding equipment based on characteristics including heat-related characteristics of the molding equipment that uses the mold, via an input layer; linearly combining the second heat-related information in time series held by the plurality of virtual nodes in the thermal reservoir layer using coupling coefficients between the plurality of virtual nodes based on the input first heat-related information in time series, and outputting the output value via an output layer; Evaluating the state of a molded product molded by the molding machine based on the output value. A program that executes a process. (Appendix 2) determining a coupling coefficient between the plurality of virtual nodes so that, when the time-series first heat-related information is input to the thermal reservoir layer based on training data including the time-series first heat-related information and a correct value of the output value during molding using a mold, the coupling coefficients between the plurality of virtual nodes are linearly combined to output an output value that approaches the correct value; 2. The program according to claim 1, which causes the computer to execute the process. (Appendix 3) Acquire an image of the molded product; evaluating a state of the molded product based on the image of the molded product and the output value output by the output layer; 3. The program according to claim 1 or 2, which causes the computer to execute the process. (Appendix 4) The evaluating step includes evaluating whether the molded product is a defective product. 4. The program of claim 1. (Appendix 5) In the step of designing the mold, The other first heat-related information is information about the temperature of the mold and / or the temperature of the mold material input in a simulation when prototyping the mold, a thermal reservoir layer including a plurality of other virtual nodes that hold other second heat-related information in a time series observable at each of the plurality of positions as heat propagates to the plurality of positions based on characteristics including heat-related characteristics of equipment for manufacturing the mold; and based on the input time-series other first heat-related information, linearly combine the time-series other second heat-related information held by a plurality of other virtual nodes in the other thermal reservoir layer using coupling coefficients between the plurality of other virtual nodes, and output the output value via another output layer; Evaluating the state of the mold to be manufactured based on the output value. 2. The program according to claim 1, which causes the computer to execute the process. (Appendix 6) determining a coupling coefficient between the plurality of other virtual nodes so that, when the other first heat-related information is input to the other thermal reservoir layer based on training data including the other first heat-related information in time series and a correct value of the output value in the simulation, the coupling coefficient is determined so that the output value obtained by linearly combining the other second heat-related information in time series held by the plurality of virtual nodes using the coupling coefficients between the plurality of other virtual nodes approaches the correct value; 6. The program according to claim 5, which causes the computer to execute the process. (Appendix 7) Evaluating the condition of the mold includes: evaluating whether the mold produces a defective molded product; The program described in Appendix 6. (Appendix 8) The computer inputting the time-series first heat-related information into a thermal reservoir layer having a plurality of virtual nodes that hold time-series second heat-related information observable at each of a plurality of positions of the molding equipment, the time-series first heat-related information being input into the molding equipment based on characteristics including heat-related characteristics of the molding equipment that uses the mold, via an input layer; linearly combining the second heat-related information in time series held by the plurality of virtual nodes in the thermal reservoir layer using coupling coefficients between the plurality of virtual nodes based on the input first heat-related information in time series, and outputting the output value via an output layer; Evaluating the state of a molded product molded by the molding machine based on the output value. An information processing method that performs processing. (Appendix 9) an input layer for inputting first heat-related information in time series during molding using a mold; a thermal reservoir layer to which first heat-related information in time series is input from the input layer, the thermal reservoir layer including a plurality of virtual nodes that hold second heat-related information in time series observable at each of a plurality of positions of the molding equipment by propagation of the first heat-related information to the plurality of positions of the molding equipment based on characteristics including heat-related characteristics of the molding equipment; a coupling layer that linearly couples the second heat-related information in time series held by a plurality of the virtual nodes in the thermal reservoir layer based on the first heat-related information in time series input to the thermal reservoir layer, using coupling coefficients between the plurality of virtual nodes; an output layer that outputs the second heat-related information linearly combined in the combination layer as an output value; and an evaluation unit that evaluates the state of a molded product molded by the molding machine based on the output value. Information processing device.

[0077] The matters described in each embodiment can be combined with each other. In addition, the independent claims and dependent claims described in the claims can be combined with each other in any combination, regardless of the reference format. Furthermore, although the claims use a format in which a claim references two or more other claims (multiple claim format), this is not limited to this format. A multiple claim (multi-multi claim) that references at least one other multiple claim may also be used. [Explanation of symbols]

[0078] 50...information processing device, 51...controller, 52...input / output unit, 53...memory, 54...learning processing unit, 55...evaluator, 56...storage unit, 57...computer program, 60...thermal reservoir model unit, 61...input layer, 62...thermal reservoir layer, 63...connection layer, 64...output layer, 200...device

Claims

1. On the computer, a thermal reservoir layer including a plurality of virtual nodes that hold time-series second heat-related information observable at each of a plurality of positions, the time-series first heat-related information being input to the molding equipment based on characteristics including heat-related characteristics of the molding equipment that uses the mold, and propagating the time-series first heat-related information to the molding equipment to a plurality of positions of the molding equipment; based on the input time-series first heat-related information, linearly combine the time-series second heat-related information held by the plurality of virtual nodes in the thermal reservoir layer using coupling coefficients between the plurality of virtual nodes, and output the output value via an output layer; Evaluating the state of the mold used in the molding machine or the state of the molded product molded by the molding machine based on the output value. A program that executes a process.

2. determining a coupling coefficient between the plurality of virtual nodes so that, when the time-series first heat-related information is input to the thermal reservoir layer based on training data including the time-series first heat-related information and a correct value of the output value during molding using a mold, the coupling coefficients between the plurality of virtual nodes are linearly combined to output an output value that approaches the correct value; The program according to claim 1 , which causes the computer to execute a process.

3. Acquire an image of the molded product; evaluating a state of the molded product based on the image of the molded product and the output value output by the output layer; The program according to claim 1 , which causes the computer to execute a process.

4. The evaluating step includes evaluating whether the molded product is a defective product. The program according to any one of claims 1 to 3.

5. In the step of designing the mold, information about the temperature of the mold and / or the temperature of the mold material input in a simulation when prototyping the mold, as the other first heat-related information; a thermal reservoir layer including a plurality of other virtual nodes that hold other second heat-related information in a time series observable at each of a plurality of positions as a result of heat propagating to the plurality of positions based on characteristics including heat-related characteristics of equipment for manufacturing the mold; and based on the input time-series other first heat-related information, linearly combine the time-series other second heat-related information held by a plurality of other virtual nodes in the other thermal reservoir layer using coupling coefficients between the plurality of other virtual nodes, and output the output value via another output layer; Evaluating the state of the mold to be manufactured based on the output value. The program according to claim 1 , which causes the computer to execute a process.

6. determining a coupling coefficient between the plurality of other virtual nodes so that, when the other first heat-related information is input to the other thermal reservoir layer based on training data including the other first heat-related information in time series and a correct value of the output value in the simulation, the coupling coefficient is determined so that the output value obtained by linearly combining the other second heat-related information in time series held by the plurality of virtual nodes using the coupling coefficients between the plurality of other virtual nodes approaches the correct value; The program according to claim 5, which causes the computer to execute processing.

7. Evaluating the condition of the mold includes: evaluating whether the mold produces a defective molded product; The program according to claim 6.

8. The computer a thermal reservoir layer including a plurality of virtual nodes that hold time-series second heat-related information observable at each of a plurality of positions, the time-series first heat-related information being input to the molding equipment based on characteristics including heat-related characteristics of the molding equipment that uses the mold, and propagating the time-series first heat-related information to the molding equipment to a plurality of positions of the molding equipment; based on the input time-series first heat-related information, linearly combine the time-series second heat-related information held by the plurality of virtual nodes in the thermal reservoir layer using coupling coefficients between the plurality of virtual nodes, and output the output value via an output layer; Evaluating the state of the mold used in the molding machine or the state of the molded product molded by the molding machine based on the output value. An information processing method that performs processing.

9. an input layer for inputting first heat-related information in time series during molding using a mold; a thermal reservoir layer to which first heat-related information in time series is input from the input layer, the thermal reservoir layer including a plurality of virtual nodes that hold second heat-related information in time series observable at each of a plurality of positions of the molding equipment by propagation of the first heat-related information to the plurality of positions of the molding equipment based on characteristics including heat-related characteristics of the molding equipment using the mold; a coupling layer that linearly couples the second heat-related information in time series held by the plurality of virtual nodes in the thermal reservoir layer based on the first heat-related information in time series input to the thermal reservoir layer, using coupling coefficients between the plurality of virtual nodes; an output layer that outputs the second heat-related information linearly combined in the combination layer as an output value; and an evaluation unit that evaluates the state of the mold used in the molding machine or the state of a molded product molded by the molding machine based on the output value. Information processing device.

Images