Prediction method and program

Abstract

This provides a simple way to predict the heat generated by a heating element. [Solution] The prediction method acquires a temperature distribution image of a substrate including a heat-generating element, extracts an evaluation surface from the temperature distribution image from an analysis model that includes specification information including the shape and layout of the substrate, sets the amount of heat generated by the heat-generating element contained in the substrate, calculates the temperature distribution of the extracted evaluation surface based on the set amount of heat generated and the specification information, and predicts the amount of heat generated by the heat-generating element contained in the substrate based on the comparison result between the temperature distribution image and the calculated temperature distribution.

Description

[Technical Field]

[0001] The present disclosure relates to a prediction method and a program. [Background Art]

[0002] Patent Document 1 discloses a design method for an electronic component mounting substrate including a substrate and a plurality of semiconductor elements integrated and mounted on the substrate. In this design method, design parameters include substrate temperature distribution, substrate in-plane thermal conductivity, substrate thickness-direction thermal conductivity, substrate size, substrate thickness, cooled body size, heat input amount, and overall heat transfer coefficient. Then, the design method includes inputting values to the design parameters excluding one unknown parameter selected from the design parameters, and calculating the temperature distribution of the substrate. [Prior Art Literature] [Patent Literature]

[0003] [Patent Document 1] Japanese Patent No. 6539862 [Summary of the Invention] [Problem to be Solved by the Invention]

[0004] In recent years, there has been demand for higher processing speed and higher functionality in electronic devices. However, electronic devices include many heating elements that generate heat when power is supplied thereto, and there is a risk that the processing speed may decrease due to heat from the heating elements. Therefore, at the time of design in the development stage of electronic devices, there is a need to appropriately evaluate, for example, what kind of heat generation will occur in a substrate according to the arrangement of electronic components scheduled to be mounted on the substrate. The temperature distribution of a substrate including heating elements can be predicted, for example, using simulation technology. In order to predict the temperature distribution of a substrate, it is required to grasp the amount of heat generated by the heating elements included in the substrate. For example, measuring the amount of heat generated by a manual measurement method using a probe or the like may take time or effort.

[0005] This disclosure was devised in light of the conventional circumstances described above, and aims to easily predict the amount of heat generated by a heating element. [Means for solving the problem]

[0006] This disclosure provides a prediction method that acquires a temperature distribution image of a substrate including a heating element, extracts an evaluation surface from the temperature distribution image from an analysis model including specification information including the shape and layout of the substrate, sets the amount of heat generated by the heating element contained in the substrate, calculates the temperature distribution of the extracted evaluation surface based on the set amount of heat generated and the specification information, and predicts the amount of heat generated by the heating element contained in the substrate based on a comparison result between the temperature distribution image and the calculated temperature distribution.

[0007] Furthermore, this disclosure provides a program for causing a computing device to acquire a temperature distribution image of a substrate including a heating element, to extract an evaluation surface from the temperature distribution image from an analysis model including specification information including the shape and layout of the substrate, to set the amount of heat generated by the heating element contained in the substrate, to calculate the temperature distribution of the extracted evaluation surface based on the set amount of heat generated and the specification information, and to predict the amount of heat generated by the heating element contained in the substrate based on a comparison result between the temperature distribution image and the calculated temperature distribution.

[0008] Furthermore, any combination of the above components, as well as any conversion of the expressions of this disclosure between methods, apparatus, systems, storage media, computer programs, etc., are also valid as aspects of this disclosure. [Effects of the Invention]

[0009] According to this disclosure, the amount of heat generated by a heating element can be easily predicted. [Brief explanation of the drawing]

[0010] [Figure 1] Conceptual diagram for designing a circuit board by predicting heat generation from a temperature distribution image. [Figure 2] Schematic diagram illustrating the temperature evaluation aspect of the analytical model. [Figure 3] Block diagram showing an example of the hardware configuration of the arithmetic unit. [Figure 4] Schematic diagram showing input / output data for each model. [Figure 5] A schematic diagram illustrating an example of temperature distribution prediction using a forward model. [Figure 6] A schematic diagram illustrating an example of predicting heat generation using an inverse model. [Figure 7] A schematic diagram illustrating an example of temperature distribution prediction using a forward model. [Figure 8] Flowchart to explain the prediction phase in a steady state [Figure 9] Flowchart to explain the prediction phase in a non-steady state [Figure 10] A flowchart illustrating an example of predicting heat generation in a transient state. [Figure 11] Flowchart to explain the design phase [Modes for carrying out the invention]

[0011] The following describes in detail embodiments of the prediction method and program disclosed herein, with reference to the drawings as appropriate. However, unnecessary details may be omitted. For example, detailed explanations of already well-known matters or redundant explanations of substantially identical configurations may be omitted. This is to avoid the following explanation becoming unnecessarily verbose and to facilitate understanding by those skilled in the art. The accompanying drawings and the following explanation are provided to enable those skilled in the art to fully understand this disclosure and are not intended to limit the subject matter described in the claims. Furthermore, the terms "first," "second," etc., in this specification are used merely to distinguish components for the sake of explanation and are not intended to be interpreted as limiting to specific components. Therefore, these expressions should be understood to be appropriately reinterpreted depending on the configuration to which the invention disclosed herein applies.

[0012] First, the prediction phase and design phase according to the present embodiment will be described with reference to FIG. 1. FIG. 1 is a conceptual diagram of designing a substrate CB1 by predicting a calorific value from a temperature distribution image TP1.

[0013] Conventionally, in circuit board design, the calorific value of a heating element such as a semiconductor element included in a substrate is measured by, for example, a manual measurement method using a probe or the like. However, such a manual measurement method takes time and labor, and it is difficult to perform measurement of the calorific value at the substrate design stage and prediction of the temperature distribution of the substrate based on the measurement results in real time.

[0014] In the present embodiment, for substrate design, Artificial Intelligence (hereinafter referred to as "AI") technology is used to predict the calorific value of heating elements included in the substrate. This makes it possible to easily predict the calorific value of heating elements included in the substrate, and further easily predict the temperature distribution of the substrate based on the predicted calorific value. That is, in the substrate design stage, real-time prediction of the calorific value and temperature distribution, and reflection of the prediction results in substrate design are realized.

[0015] First, the prediction phase in the present embodiment will be briefly described with reference to FIG. 1. In the present embodiment, as a premise, an analysis model M1 is generated from a substrate image P1 generated by imaging the substrate CB1. The analysis model M1 may be generated using a simulation technique such as Computational Fluid Dynamic (hereinafter referred to as "CFD") and a tool such as Computer Aided Design (hereinafter referred to as "CAD"). The analysis model M1 is a three-dimensional model imitating the substrate CB1, and includes heating elements (for example, a heating element HE1, a heating element HE2, etc.) included in the substrate CB1. For predicting the calorific value of the heating elements included in the substrate CB1, a temperature evaluation surface EA1 of the analysis model M1 is extracted. Details will be described later with reference to FIG. 2.

[0016] In the prediction phase, the heat generation amount of the heating element included in the analysis model M1 is predicted based on the temperature distribution image TP1 of the substrate CB1 generated by, for example, a thermography camera or the like, and the analysis model M1 (more precisely, the temperature evaluation surface EA1 of the analysis model M1). That is, the heat generation amount of the heating element included in the substrate CB1 is predicted. For predicting the heat generation amount, a pre-generated, in other words, trained AI model is used. Input data and output data to the AI model will be described later with reference to FIG. 4. Further, the AI models used in the present embodiment are broadly classified into two types: forward models and inverse models. The forward model and the inverse model will be described later with reference to FIGS. 4 to 7. Further, the detailed flow of processing in the prediction phase will be described later with reference to FIGS. 8 to 10.

[0017] The heat generation amount of the heating element included in the substrate CB1 predicted in the prediction phase is used for substrate design or design modification in the design phase. Specifically, for example, the temperature distribution T1 of the substrate CB1 is predicted based on the predicted heat generation amount and the layout of the substrate CB1. The layout of the substrate CB1 indicates the positions of the heating elements on the substrate CB1. This allows a user such as a developer of the substrate CB1 to confirm what kind of heat generation will occur when the substrate CB1 has the designed layout. The user can further change the layout of the substrate CB1 and obtain the temperature distribution T2. As described above, in the design phase, temperature distribution prediction using the heat generation amount predicted in the prediction phase and design of the substrate CB1 are performed. As in the case of heat generation amount prediction, a pre-generated AI model is used for temperature distribution prediction. The detailed flow of processing in the design phase will be described later with reference to FIG. 11.

[0018] FIG. 2 is a schematic diagram for explaining the temperature evaluation surface EA1 of the analysis model M1. In the description of FIG. 2, for convenience of explanation, a three-dimensional coordinate system composed of an x-axis, a y-axis, and a z-axis is used. Further, for each of the x-axis, y-axis, and z-axis, the direction of the arrow is defined as the positive direction, and the direction opposite to the positive direction is defined as the negative direction.

[0019] In the example shown in Figure 2, the analysis model M1 is assumed to include at least three heating elements: HE3, HE4, and HE5. The analysis model M1 is generated as a three-dimensional model. Figure 2 shows the analysis model M1 viewed from the positive z-axis direction (top view) and from the negative y-axis direction (side view). From the side view of the analysis model M1, it can be seen that heating element HE4 is stacked on top of heating element HE5.

[0020] The analysis model M1 includes specification information, including the shape and layout of the substrate CB1. For example, the analysis model M1 may include the size of the substrate CB1, the size of each heating element, and their position on the substrate CB1 (i.e., the layout of the substrate CB1) as specification information. The specification information for the substrate CB1 is added by referring to Figure 4.

[0021] In this embodiment, the substrate CB1 is imaged using a thermographic camera or the like, and a temperature distribution image of the substrate CB1 is generated. The temperature evaluation surface of the analysis model M1 is extracted to correspond to the generated temperature distribution image of the substrate CB1. At this time, the surface (top surface) of the analysis model M1 is extracted as the temperature evaluation surface. For example, as shown in the example in Figure 2, the temperature evaluation surface EA1 may be extracted from the analysis model M1. The temperature evaluation surface corresponding to the temperature distribution image may be, for example, a temperature evaluation surface in which the position of the heat-generating element included in the analysis model M1 and the position where a temperature rise is observed in the temperature distribution image substantially coincide. In other words, it is preferable that the temperature evaluation surface is extracted in such a way that the prediction result and the temperature distribution image can be compared when the amount of heat generated by the heat-generating element or the temperature distribution is predicted using the extracted temperature evaluation surface. The extraction of the temperature evaluation surface may be performed, for example, by image recognition using AI technology.

[0022] Figure 3 is a block diagram showing an example of the hardware configuration of the computing device 10. In this embodiment, the prediction of the heat generation amount of the heat source using an AI model may be performed by the computing device 10 or on the cloud 21. When the prediction of heat generation amount is performed on the cloud 21, the computing device 10 may be used as an interface for the user to use the cloud 21.

[0023] The computing unit 10 may be a general-purpose computer device such as a Personal Computer (hereinafter referred to as "PC") or a server computer, or it may be a mobile device such as a tablet terminal or a smartphone. Note that the configuration shown in Figure 3 is just one example, and one part may be divided into multiple parts, or multiple parts may be combined into one.

[0024] The computing unit 10 comprises a processor 11, memory 12, communication device 13, input device 14, external interface device 15, display device 16, and image acquisition unit 17. Each component is configured to communicate via an internal interface device 18.

[0025] The processor 11 may be configured using, for example, a Central Processing Unit (hereinafter referred to as "CPU"), a Graphical Processing Unit (hereinafter referred to as "GPU"), a Micro Processing Unit (hereinafter referred to as "MPU"), a Digital Signal Processor (hereinafter referred to as "DSP"), or a Field Programmable Gate Array (hereinafter referred to as "FPGA"). The processor 11 realizes various functions by reading and executing various data and programs held in the memory 12.

[0026] Memory 12 is a storage unit for storing various data and programs. Memory 12 may consist of volatile / non-volatile storage devices such as Random Access Memory (hereinafter referred to as "RAM"), Read Only Memory (hereinafter referred to as "ROM"), Hard Disk Drive (hereinafter referred to as "HDD"), and Solid State Drive (hereinafter referred to as "SSD").

[0027] The communication device 13 is an interface for communicating with external devices via the network 20. The communication standards that the communication device 13 can support are not particularly limited and may support either wired or wireless communication standards. Furthermore, the communication device 13 may support multiple communication standards. Therefore, the network 20 used by the communication device 13 may be composed of a combination of networks using multiple communication standards.

[0028] The input device 14 receives operations and instructions from a user, for example, who is selling goods in an automated warehouse. The input device 14 may consist of a mouse, keyboard, touch panel display, etc.

[0029] The external interface device 15 is an interface for sending and receiving data with an external device.

[0030] The display device 16 displays various user interfaces to the user. The display device 16 may consist of a liquid crystal display, a touch panel display, or the like.

[0031] The image acquisition unit 17 is an interface for acquiring images from the camera 19. The camera 19 includes a camera capable of generating temperature distribution images, such as a thermographic camera. The computing unit 10 may also include the camera 19.

[0032] In this embodiment, the AI ​​model may be stored and held in the cloud 21, memory 12, or an external device (not shown) that can communicate with the computing device 10. Similarly, the analysis model M1, pre-generated based on the substrate image P1 of the substrate CB1, may also be stored and held in the cloud 21, memory 12, or an external device (not shown) that can communicate with the computing device 10. The computing device 10 may then use the temperature distribution image acquired from the camera 19, the analysis model M1, and the AI ​​model to predict the amount of heat generated or the temperature distribution. A function capable of predicting the amount of heat generated or the temperature distribution may be provided as a tool on the cloud 21, for example, and this tool may use the temperature distribution image, the analysis model M1, and the AI ​​model to predict the amount of heat generated or the temperature distribution.

[0033] Next, we will explain the input and output data of the AI ​​models, referring to Figure 4. Figure 4 is a schematic diagram showing the input and output data of each model.

[0034] The AI ​​model receives input from various data points, not all of them, but a selection of data from the substrate information, heat source information, and substrate temperature distribution information. Specific input and output examples will be described later.

[0035] The substrate information includes at least data D1 indicating the size, or in other words, the shape, of substrate CB1. The substrate information may further include data D2 indicating the thermal conductivity distribution of substrate CB1 and data D3 indicating the emissivity distribution of substrate CB1. Note that the substrate information may also include other data.

[0036] The heat-generating element information includes at least data D4 indicating the position and size of the various heat-generating elements contained in substrate CB1, in other words, the layout of substrate CB1. The heat-generating element information may further include data D5 indicating the thermal conductivity of the various heat-generating elements contained in substrate CB1, and data D6 indicating the amount of heat generated by the various heat-generating elements. The heat-generating element information may also include other data.

[0037] The substrate temperature distribution information includes a temperature distribution image of substrate CB1 as data D7.

[0038] The specification information included in the analysis model M1 includes at least the shape and layout of the substrate CB1. That is, the specification information includes at least data D1 and data D4. However, the specification information may include further data. For example, the specification information may further include data D2, data D3, data D5, or data D6.

[0039] The AI ​​model outputs at least one of the various data included in the substrate information, heat source information, and substrate temperature distribution information, depending on the input data. The explanations for data D8, D9, D10, D11, D12, D13, and D14 are omitted because they overlap with the explanations for data D1, D2, D3, D4, D5, D6, and D7.

[0040] For example, the AI ​​model may be generated to output data that was not input as input data from the various data included in the substrate information, heat-generating element information, and substrate temperature distribution information, respectively. For example, the AI ​​model may be generated to output data D9, D10, D11, D12, and D13 when data D1 and data D7 are input. In this case, the AI ​​model can predict and output not only the amount of heat generated by the heat-generating element, but also the physical properties of the substrate CB1 and the heat-generating element. These physical properties include, for example, the thermal conductivity distribution (data D9) and emissivity distribution (data D10) of the substrate CB1, and the thermal conductivity (data D12) of the heat-generating element.

[0041] Alternatively, for example, the AI ​​model may be generated to output data D13 when data D1, data D2, data D3, data D4, data D5, and data D7 are input. In this case, the AI ​​model can predict and output the amount of heat generated by the heat source (data D13).

[0042] Alternatively, for example, the AI ​​model may be generated to output data D2 when data D1, data D3, data D4, data D5, data D6, and data D7 are input. In this case, the AI ​​model can predict and output the physical properties (data D2) of substrate CB1.

[0043] Thus, the AI ​​model may be generated not only to predict and output the amount of heat generated, but also to predict and output the amount of heat generated and physical properties, or only the physical properties.

[0044] In this embodiment, the AI ​​model that outputs data D14, i.e., the temperature distribution image, is referred to as the forward model, and the AI ​​model that outputs data D13, i.e., the heat generated by the heat source, is referred to as the inverse model. For the sake of explanation, in this embodiment, the AI ​​model is not generated to output both data D13 and data D14.

[0045] The learning algorithm for generating forward models is not particularly limited, but machine learning techniques including neural network technologies such as Convolutional Neural Networks (hereinafter referred to as "CNNs") and U-Nets, as well as deep learning techniques, may be used. The learning algorithm for generating inverse models is not particularly limited, but machine learning techniques including neural network technologies such as Region Based CNNs (hereinafter referred to as "RCNNs"), Generative Adversarial Networks (hereinafter referred to as "GANs"), Conditional GANs (hereinafter referred to as "CGANs"), Diffusion Models, and Conditional Diffusion Models, as well as deep learning techniques, may be used.

[0046] Next, we will explain specific examples of the inputs and outputs of the forward and inverse models with reference to Figures 5, 6, and 7. Figure 5 is a schematic diagram illustrating an example of temperature distribution prediction using the forward model 30.

[0047] In the example shown in Figure 5, the input data includes the heat generated by the heating element (data D6), the thermal conductivity of the heating element (data D5), the thermal conductivity distribution of the substrate CB1 (data D2), and the emissivity distribution of the substrate CB1 (data D3). Data D5 or D6 may include data D1, i.e., the size of the substrate CB1, and data D4, i.e., the position and size of the heating element. Data D2 or D3 may include data D1. These data may be included in the temperature evaluation surface EA1 of the analysis model M1. For example, the heat generated by the heating element may be a value set for the temperature evaluation surface EA1 of the analysis model M1.

[0048] The forward model 30 outputs a temperature distribution image (data D14) in response to the input data. The output temperature distribution image shows the temperature distribution of the temperature evaluation surface EA1 of the analysis model M1. Now, consider the case where, for example, the temperature distribution image output from the forward model 30 is compared with the temperature distribution image of the substrate CB1 generated by a thermographic camera or the like. If the comparison result, that is, the difference between the temperature distribution images, is within a predetermined tolerance range, then, for example, data D6, that is, the amount of heat generated by the heat source set for the temperature evaluation surface EA1 of the analysis model M1, is likely to be reasonable. In this way, it becomes possible to predict the amount of heat generated by the heat source using the forward model 30.

[0049] Figure 6 is a schematic diagram illustrating an example of heat generation prediction using the inverse model 31. In the example in Figure 6, the input data includes a temperature distribution image (data D7), the thermal conductivity of the heat-generating element (data D5), the thermal conductivity distribution of the substrate CB1 (data D2), and the emissivity distribution of the substrate CB1 (data D3). Data D5 may include data D1, i.e., the size of the substrate CB1, and data D4, i.e., the position and size of the heat-generating element. Data D2 or data D3 may include data D1. Data D7 may include data D1.

[0050] The inverse model 31 outputs the heat generated by the heat source (data D13) in response to the input data. The output heat generated by the heat source represents the heat generated by the heat source included in the analysis model M1. By using the data D13 output from the inverse model 31 as input data for the forward model, for example, the forward model 30), the forward model 30 can output a temperature distribution image with higher accuracy.

[0051] Furthermore, for example, the input data may not include data D2, and the inverse model 31 may be generated to output not only data D13 but also data D9.

[0052] During the design or development of substrate CB1, power is supplied to substrate CB1. As a result, the heating elements on substrate CB1 generate heat, and the temperature of substrate CB1 rises. In other words, the temperature distribution of substrate CB1 changes over time. The temperature rise of substrate CB1 becomes more gradual over time, and eventually, it becomes a nearly constant temperature distribution. Thus, the temperature state of substrate CB1 includes two states: a steady state and a transient state. In the examples in Figures 5 and 6, it is assumed that substrate CB1 is in a steady state. In the following Figure 7, an example of input and output in a forward model when substrate CB1 is in a transient state is explained.

[0053] Figure 7 is a schematic diagram illustrating an example of temperature distribution prediction using forward model 32. In the example in Figure 7, the input data includes the heat generated by the heat source (data D6), the thermal conductivity of the heat source (data D5), the thermal conductivity distribution of the substrate CB1 (data D2), and the emissivity distribution of the substrate CB1 (data D3). Data D5 or data D6 may include data D1, i.e., the size of the substrate CB1, and data D4, i.e., the position and size of the heat source. Data D2 or data D3 may include data D1. These data may be included in the temperature evaluation surface EA1 of the analysis model M1. For example, the heat generated by the heat source may be a value set for the temperature evaluation surface EA1 of the analysis model M1.

[0054] Furthermore, Figure 7 includes data D15 as input data, which represents the estimated time of the temperature distribution. Data D15 indicates that the estimated time of the temperature distribution is time_g.

[0055] The forward model 32 outputs a temperature distribution image (data D14) in response to the input data. The output temperature distribution image shows the temperature distribution of the temperature evaluation surface EA1 of the analysis model M1 at time time_g.

[0056] The forward model 32 may be generated, for example, to output a temperature distribution image at a time after a predetermined time has elapsed from a predetermined time when a predetermined time is input. Furthermore, the predetermined time, or the estimated time of the temperature distribution, does not have to be prepared individually as input data, but may be included in other input data (in the example in Figure 7, data D1, data D2, data D3, data D4, data D5, or data D6).

[0057] Next, with reference to Figure 8, the heat generation prediction process in the prediction phase when substrate CB1 is in a steady state will be explained. Figure 8 is a flowchart for explaining the prediction phase in a steady state. Note that each process in the flowchart shown in Figure 8 will be explained assuming that it is performed by the processor 11 of the arithmetic unit 10. Also, at the start of the flowchart shown in Figure 8, substrate CB1 is assumed to be in a steady state. Furthermore, it is assumed that an analysis model M1 that mimics substrate CB1 has been generated, and a temperature distribution image TP1 of substrate CB1 has been generated by the camera 19.

[0058] The processor 11 acquires a temperature distribution image TP1 of the substrate CB1 from the camera 19 via the image acquisition unit 17 (step S100).

[0059] The processor 11 extracts the temperature evaluation surface EA1 of the analysis model M1 based on the temperature distribution image TP1 acquired in step S100 (step S101).

[0060] Here, the processor 11 may accept a selection, for example, by user input, regarding whether to use a forward model or an inverse model in the prediction phase. Alternatively, the user may have already selected whether to use a forward model or an inverse model in the prediction phase.

[0061] The processor 11 determines whether the forward model or the reverse model has been selected (step S102).

[0062] If the processor 11 determines that a forward model has been selected (step S102: forward model), it sets the amount of heat generated by the heating element on the temperature evaluation surface EA1 extracted in step S101 (step S103). The method of setting the amount of heat generated in step S103 is not particularly limited; it may be set randomly or within a range predetermined by the user. Also, as shown in the flowchart in Figure 8, the processing in step S103 and the subsequent step S104 may be executed repeatedly. Therefore, the amount of heat generated may be set based on the results of the processing in previous steps S103 and S104. The same applies to the method of setting the amount of heat generated in the flowcharts shown in Figures 9 and 10, respectively, which will be described later.

[0063] The processor 11 inputs the set amount of heat generated into the forward model and calculates the temperature distribution of the temperature evaluation surface EA1 (step S104). At this time, the forward model is also input with specification information including at least the shape and layout of the substrate CB1. The forward model may output the temperature distribution of the temperature evaluation surface EA1 as a temperature distribution image. Then, the processor 11 proceeds to step S108.

[0064] Returning to the process in step S102, we will now explain the case where the inverse model is selected. If the processor 11 determines that the inverse model has been selected (step S102: inverse model), it sets the temperature distribution of the temperature evaluation surface EA1 extracted in step S101 as the target value (step S105). The method of setting the temperature distribution in step S105 is not particularly limited; it may be set randomly or within a range predetermined by the user. Also, as shown in the flowchart in Figure 8, the processes in step S105 and the subsequent steps S106 and S107 may be executed repeatedly. Therefore, the temperature distribution may be set based on the results of the processes in previous steps S105, S106 and S107.

[0065] The processor 11 inputs the temperature distribution set in step S105 as an image into the inverse model and calculates the amount of heat generated by the heat-generating element on the temperature evaluation surface EA1 (step S106). At this time, the inverse model is also input specification information, including at least the shape and layout of the substrate CB1.

[0066] The processor 11 inputs the amount of heat generated in step S106 into the forward model and calculates the temperature distribution of the temperature evaluation surface EA1 (step S107). Then, the processor 11 proceeds to step S108.

[0067] The processor 11 compares the temperature distribution calculated by the forward model with the temperature distribution image TP1 acquired in step S100 (step S108). At this time, the processor 11 may calculate, for example, either the mean squared error or the mean absolute error, or both, between the temperature distribution image TP1 and the calculated temperature distribution. For illustrative purposes, both the mean squared error and the mean absolute error are calculated. In addition, the region used for comparison between the temperature distribution image TP1 and the calculated temperature distribution may be arbitrarily set. This is because if there is a region with no temperature rise or a region with low reliability of the predicted value, the reliability of the temperature comparison result in such a region may also be low.

[0068] The processor 11 determines whether the comparison result in step S108 is within an acceptable range (step S109). For example, if both the mean squared error and the mean absolute error calculated in step S108 are lower than a predetermined judgment criterion temperature, the processor 11 may determine that the comparison result is within an acceptable range.

[0069] Furthermore, the comparison between the temperature distribution image and the calculated temperature distribution, and the determination of whether the comparison result falls within an acceptable range, as described later in the flowcharts of Figures 9, 10, and 11, are also carried out in the same manner as described above.

[0070] The processor 11 determines that the comparison result in step S108 is not within an acceptable range, and if the forward model is selected (step S109: NO, step S102: forward model), it returns to step S103 and repeats the process.

[0071] If the processor 11 determines that the comparison result in step S108 is not within an acceptable range, and if the inverse model is selected (step S109: NO, step S102: inverse model), it returns to step S105 and repeats the process.

[0072] If the processor 11 determines that the comparison result in step S108 is within an acceptable range (step S109: YES), it saves the amount of heat generated by the heat-generating element on the temperature evaluation surface EA1, which was set in step S103 or calculated in step S106 (step S110). Then, the processor 11 terminates this processing flow. The processor 11 may, for example, save the amount of heat generated in memory 12. The saved amount of heat generated by the heat-generating element is used, for example, in the design of the substrate CB1 during the design phase.

[0073] Thus, when designing or modifying a substrate CB1, for example, by using a temperature distribution image generated by a camera 19 at the site (e.g., a research institute, factory, or other work site) and a pre-generated AI model, the amount of heat generated by the heat-generating elements contained in the substrate CB1 can be predicted quickly and accurately, assuming that the substrate CB1 is in a steady state.

[0074] Although the processes in the flowchart shown in Figure 8 were described as being performed by the processor 11 of the arithmetic unit 10, they may also be performed on the cloud 21. In this case, the processor 11 of the arithmetic unit 10 may upload the temperature distribution image acquired from the camera 19 to the cloud 21. Similarly, the processes in the flowcharts shown in Figures 9, 10, and 11, which will be described later, may also be performed on the cloud 21.

[0075] Hereinafter, the forward model that calculates the temperature distribution of a predetermined temperature evaluation surface in the analysis model M1 based on the input of the heat generation amount of a heat source set on that surface and specification information including the shape and layout of the substrate CB1 may be referred to as the first AI model. Furthermore, the inverse model that calculates the heat generation amount of a heat source on a predetermined temperature evaluation surface based on the input of the temperature distribution set on that surface in the analysis model and the specification information may be referred to as the second AI model.

[0076] Next, with reference to Figure 9, the heat generation prediction process in the prediction phase when substrate CB1 is in a transient state will be explained. Figure 9 is a flowchart for explaining the prediction phase in a transient state. Note that each process in the flowchart shown in Figure 9 will be explained assuming that it is performed by the processor 11 of the arithmetic unit 10. Also, at the start of the flowchart shown in Figure 9, substrate CB1 is assumed to be in a transient state. Furthermore, it is assumed that an analysis model M1 that mimics substrate CB1 has been generated, and that temperature distribution images of substrate CB1 at different times have been generated by the camera 19. In other words, it is assumed that multiple temperature distribution images along the course of time have been generated.

[0077] The processor 11 acquires N temperature distribution images of the substrate CB1 from the camera 19 via the image acquisition unit 17 (step S200). Here, N is a natural number greater than or equal to 2.

[0078] The processor 11 extracts the temperature evaluation surface EA1 of the analysis model M1 based on the N temperature distribution images acquired in step S201 (step S201).

[0079] The processor 11 sets the value i, which is used to determine the end of the heat generation prediction process that starts in step S203, to 1 (step S202).

[0080] The processor 11 starts the heat generation prediction process and executes the processes from step S204 to step S211 while i is less than or equal to N (step S203).

[0081] The processor 11 selects the i-th image from the N temperature distribution images acquired in step S200.

[0082] The processor 11 sets the amount of heat generated by the heating element on the temperature evaluation surface EA1 extracted in step S201 (step S205).

[0083] The processor 11 sets the estimated time for the temperature distribution of the temperature evaluation surface EA1 (step S206). Preferably, the set estimated time (and the amount of heat generated in step S205) is such that the temperature distribution calculated from the forward model in the subsequent step S207 and the i-th temperature distribution image are at the same time. This is because if the calculated temperature distribution and the temperature distribution image are at the same time, they can be compared. However, even if both are at the same time, there may be unacceptable errors between them. Therefore, the process in step S206 may be repeated to adjust the set estimated time.

[0084] The processor 11 inputs the set heat generation amount and estimated time into the forward model and calculates the temperature distribution of the temperature evaluation surface EA1 at the estimated time (step S207). At this time, the forward model is also input specification information, including at least the shape and layout of the substrate CB1.

[0085] The processor 11 compares the temperature distribution calculated in step S207 with the temperature distribution image selected in step S204 (step S208). As described above in the explanation of Figure 8, the processor 11 may calculate either the mean squared error or the mean absolute error or both of the temperature distribution image and the calculated temperature distribution.

[0086] The processor 11 determines whether the comparison result in step S208 is within an acceptable range (step S209).

[0087] If the processor 11 determines that the comparison result in step S208 is not within an acceptable range (step S209: NO), it returns to step S205 or step S206 and repeats the process. Whether to return to step S205 or step S206 may be predetermined by the user, for example, or it may be set to be determined based on the comparison result in step S208. For example, if the difference between the mean squared error or mean absolute error calculated in step S208 and a predetermined judgment criterion temperature is greater than or equal to a predetermined threshold, the processor 11 may return to step S205.

[0088] If the processor 11 determines that the comparison result in step S208 is within an acceptable range (step S209: YES), it saves the amount of heat generated by the heat-generating element on the temperature evaluation surface EA1, which was set in step S205 (step S210). The processor 11 may, for example, save the amount of heat generated in memory 12.

[0089] The processor 11 adds 1 to the value of i (step S211). For example, if the value of i is set to 1, step S211 makes the value of i 2.

[0090] The processor 11 repeatedly executes the processes from step S204 to step S211 as long as i is less than or equal to N.

[0091] This process compares, for example, the temperature distribution image TPi1 with the temperature distribution Ti1 calculated by the forward model, and the heat generation amount is repeatedly predicted based on the comparison results and finally saved. Then, in the next heat generation amount prediction process, the temperature distribution image TPi2 with the temperature distribution Ti2 calculated by the forward model is compared, and the heat generation amount is repeatedly predicted based on the comparison results and finally saved. And again, in the next heat generation amount prediction process, the temperature distribution image TPi3 with the temperature distribution Ti3 calculated by the forward model is compared, and the heat generation amount is repeatedly predicted based on the comparison results and finally saved. In this way, the heat generation amount prediction process is repeated, and N heat generation amounts are predicted and saved.

[0092] The processor 11 calculates the average value of the N stored heat generation values ​​(step S212). The calculated average value of the heat generation values ​​may be stored in memory 12, for example. The stored average value of the heat generation values ​​is then used as the heat generation value of the heat-generating element in the design of the board CB1 during the design phase.

[0093] Thus, when designing or modifying a substrate CB1, for example, by using a temperature distribution image generated by a camera 19 at the site (e.g., a research institute, factory, or other work site) and a pre-generated AI model, the amount of heat generated by the heat-generating elements contained in the substrate CB1 can be predicted quickly and accurately, even when the substrate CB1 is in a non-steady state.

[0094] Hereinafter, the sequential model that calculates the temperature distribution of a predetermined temperature evaluation surface at a predetermined time, based on inputs of the amount of heat generated by a heat source set on a predetermined temperature evaluation surface of analysis model M1, a predetermined estimated time for the temperature distribution of the predetermined temperature evaluation surface, and specification information including the shape and layout of the substrate CB1, may be referred to as the third AI model.

[0095] Next, referring to Figure 10, we will explain another pattern of heat generation prediction processing in the prediction phase when the substrate CB1 is in a transient state. Figure 10 is a flowchart illustrating an example of heat generation prediction in a transient state. Note that at the start of the flowchart shown in Figure 10, the processing up to step S201 of the flowchart shown in Figure 9 has been completed.

[0096] The processor 11 selects the first temperature distribution image from the N temperature distribution images acquired in step S201 (step S300).

[0097] The processor 11 selects a second temperature distribution image from the N temperature distribution images acquired in step S201 (step S301). The second temperature distribution image is an image acquired at a predetermined time after the time the first temperature distribution image was acquired.

[0098] In the example shown in Figure 10, the processor 11 selects temperature distribution image TP4 and temperature distribution image TP5 as the first and second temperature distribution images, respectively.

[0099] The processor 11 calculates the temperature difference between the first temperature distribution image selected in step S300 and the second temperature distribution image selected in step S301 (step S302). In the example in Figure 10, the processor 11 calculates the temperature difference Td1.

[0100] The processor 11 sets the amount of heat generated by the heat source on the temperature evaluation surface EA1 extracted in step S201 (step S303).

[0101] The processor 11 sets a first estimated time and a second estimated time for the temperature distribution of the temperature evaluation surface EA1 (step S304). These estimated times are set such that, for example, the difference between the first estimated time and the second estimated time is the difference in time between the first temperature distribution image and the second temperature distribution image. It is preferable that the first estimated time (and the amount of heat generated in step S303) is set so that the temperature distribution calculated from the forward model in the subsequent step S305 and the temperature distribution image of the first temperature distribution are at the same time. Similarly, it is preferable that the second estimated time (and the amount of heat generated in step S303) is set so that the temperature distribution calculated from the forward model in the subsequent step S306 and the temperature distribution image of the second temperature distribution are at the same time. This is because if the calculated temperature distribution and the temperature distribution image are at the same time, they can be compared. However, even if both are at the same time, there may be unacceptable errors between them. Therefore, the process in step S304 may be repeated to adjust the set estimated times.

[0102] The processor 11 inputs the set heat generation amount and the first estimated time into the forward model and calculates the temperature distribution of the temperature evaluation surface EA1 at the first estimated time (step S305). At this time, specification information including at least the shape and layout of the substrate CB1 is also input into the forward model. As a result, the temperature distribution T4 is calculated.

[0103] The processor 11 inputs the set heat generation amount and the second estimated time into the forward model and calculates the temperature distribution of the temperature evaluation surface EA1 at the second estimated time (step S306). At this time, specification information including at least the shape and layout of the substrate CB1 is also input into the forward model. As a result, the temperature distribution T5 is calculated.

[0104] The processor 11 calculates the temperature difference between the temperature distribution calculated in step S305 and the temperature distribution calculated in step S306 (step S307). In the example in Figure 10, the processor 11 calculates the temperature difference Td2.

[0105] The processor 11 compares the temperature difference calculated in step S302 with the temperature difference calculated in step S307 (step S308). In the example in Figure 10, the processor 11 compares the temperature difference Td1 with the temperature difference Td2.

[0106] The processor 11 determines whether the comparison result in step S308 is within an acceptable range (step S309). The explanations of the comparison in step S308 and the determination in step S309 are the same as the explanations of steps S208 and S209 in Figure 8, so they are omitted here.

[0107] If the processor 11 determines that the comparison result in step S308 is not within an acceptable range (step S309: NO), it returns to step S303 or step S304 and repeats the process. Whether to return to step S303 or step S304 may be predetermined by the user, for example, or it may be set to be determined based on the comparison result in step S308.

[0108] If the processor 11 determines that the comparison result in step S208 is within an acceptable range (step S309: YES), it saves the amount of heat generated by the heat-generating element on the temperature evaluation surface EA1, which was set in step S303 (step S310). The processor 11 may, for example, save the set amount of heat generated in memory 12. The saved amount of heat generated is used as the amount of heat generated by the heat-generating element in the design of the board CB1 during the design phase.

[0109] As explained with reference to Figure 9, not only can the amount of heat generated be predicted at multiple time points and the prediction results be averaged, but the amount of heat generated can also be predicted based on the change in temperature distribution, as shown in the example in Figure 10.

[0110] Next, the design phase process will be explained with reference to Figure 11. Figure 11 is a flowchart for explaining the design phase. Each process in the flowchart shown in Figure 11 will be explained assuming that it is performed by the processor 11 of the arithmetic unit 10. Furthermore, it is assumed that at the start of the flowchart shown in Figure 11, the series of processes in Figure 8, Figure 9, or Figure 10 have been completed, the amount of heat generated by the heat-generating elements contained in the substrate CB1 has been predicted and stored in the memory 12, etc.

[0111] The processor 11 sets the temperature evaluation surface of the analysis model M1 for the design of the substrate CB1 (step S400).

[0112] The processor 11 may accept a selection, for example, by user input, regarding whether the forward model or the reverse model is used during the design phase. Alternatively, the user may have already selected whether the forward model or the reverse model is used during the design phase.

[0113] The processor 11 determines whether the forward model or the reverse model has been selected (step S401).

[0114] If the processor 11 determines that the forward model has been selected (step S401: forward model), it sets specification information such as the shape or layout of the temperature evaluation surface set in step S400 for the substrate CB1 (step S402).

[0115] The processor 11 inputs the heat generation amount of the heat source saved in the prediction phase and the specification information set in step S402 into the forward model and calculates the temperature distribution of the temperature evaluation surface set in step S400 (step S403). The forward model used in step S403 may be, for example, the first AI model. Then, the processor 11 proceeds to step S407.

[0116] Returning to the process of step S401, we will now explain the case where the inverse model is selected. If the processor 11 determines that the inverse model has been selected (step S401: inverse model), it sets the temperature distribution of the temperature evaluation surface set in step S400 as the target value (step S404). The method for setting the temperature distribution in step S404 was described above in the explanation for step S105 of the flowchart shown in Figure 8, so the explanation is omitted here.

[0117] The processor 11 inputs the temperature distribution set in step S405 as an image into the inverse model and calculates the shape or layout of the temperature evaluation surface of the substrate CB1, more precisely, the shape or layout of the temperature evaluation surface set in step S400, i.e., specification information (step S405).

[0118] The processor 11 inputs the heat generation amount of the heat source saved in the prediction phase and the specification information calculated in step S405 into the forward model and calculates the temperature distribution of the temperature evaluation surface set in step S400 (step S406). The forward model used in step S406 may be, for example, the first AI model. Then, the processor 11 proceeds to step S407.

[0119] The processor 11 determines whether the temperature distribution calculated in step S403 or step S406 is within an acceptable range (step S407). For example, an upper limit temperature for the substrate CB1 may be set during the design phase, and the determination in step S407 may be made by comparing the temperature distribution calculated in step S403 or step S406 with the upper limit temperature. For example, if the temperature distribution calculated in step S403 or step S406 is less than the upper limit temperature, the processor 11 may determine that the temperature distribution is within an acceptable range.

[0120] If the processor 11 determines that the temperature distribution is not within the acceptable range, and if the forward model is selected (step S407: NO, step S401: forward model), it returns to step S402 and repeats the process.

[0121] If the processor 11 determines that the temperature distribution is not within the acceptable range, and if the inverse model is selected (step S407: NO, step S401: inverse model), it returns to step S404 and repeats the process.

[0122] If the processor 11 determines that the temperature distribution is within an acceptable range (step S407: YES), it saves the design details of the board CB1 (step S408). Then, the processor 11 terminates this processing flow. The design details of the board CB1 refer to the specification information set in step S402 or calculated in step S405.

[0123] In this way, using AI models in the design phase makes it possible to support the design of the substrate CB1.

[0124] Hereinafter, the inverse model that calculates specification information, including the shape and layout of a predetermined temperature evaluation surface, in response to the input of a temperature distribution set on a predetermined temperature evaluation surface of the analysis model M1 may be referred to as the fourth AI model.

[0125] (Summary of this embodiment) The following technologies are disclosed based on the above description of this embodiment. Note that the components and other elements in parentheses are examples of those corresponding to the above embodiment, but are not limited to these.

[0126] <Technology 1> The prediction method involves acquiring a temperature distribution image (e.g., temperature distribution image TP1) of a substrate (e.g., substrate CB1) containing a heat-generating element (e.g., heat-generating element HE1), extracting an evaluation surface (e.g., temperature evaluation surface EA1) from an analysis model (e.g., analysis model M1) that includes specification information including the substrate shape (e.g., data D1) and layout (e.g., data D4), setting the heat generation amount of the heat-generating element contained in the substrate, calculating the temperature distribution of the extracted evaluation surface based on the set heat generation amount and the specification information, and predicting the heat generation amount of the heat-generating element contained in the substrate based on the comparison result between the temperature distribution image and the calculated temperature distribution.

[0127] This allows the prediction method to predict the amount of heat generated by heat-generating elements in the substrate based on the substrate's temperature distribution image and an analysis model of the substrate. Therefore, it is possible to predict the amount of heat generated more easily compared to measuring the amount of heat generated using manual measurement methods, such as probes.

[0128] <Technology 2> The prediction method described in Technology 1 calculates the temperature distribution of an extracted evaluation surface by inputting the set heat output and specification information into a first AI model that calculates the temperature distribution of a predetermined evaluation surface in response to inputs of the heat output and specification information of a heat source set on a predetermined evaluation surface of the analysis model.

[0129] This allows the prediction method to calculate the temperature distribution of the evaluation surface of the analysis model using an AI model. The prediction method can then compare the calculated temperature distribution with a temperature distribution image. By using an AI model, the prediction method can predict heat generation simply, quickly, and with high accuracy.

[0130] <Technology 3> The prediction method described in Technology 2 involves setting the temperature distribution of the extracted evaluation surface, and inputting the set temperature distribution and specification information into a second AI model that calculates the amount of heat generated by heat-generating elements on a predetermined evaluation surface in response to the input of the temperature distribution and specification information set on a predetermined evaluation surface of the analysis model. This calculates the amount of heat generated by heat-generating elements on the extracted evaluation surface, and the calculated amount of heat generated by heat-generating elements on the extracted evaluation surface is set as the amount of heat to be input into the first AI model.

[0131] This allows the prediction method to calculate the heat generated by the heat source in the evaluation area of ​​the analysis model using an AI model. The prediction method can then input the calculated heat generated into an AI model for calculating the temperature distribution.

[0132] <Technology 4> The prediction method described in Technology 1 involves acquiring multiple temperature distribution images at different times, and for each of the multiple temperature distribution images, it performs the following: setting the amount of heat generated by the heat source on the extracted evaluation surface, setting the estimated time for the temperature distribution of the extracted evaluation surface, inputting the set amount of heat generated by the heat source on the predetermined evaluation surface of the analysis model, a predetermined estimated time for the temperature distribution of the predetermined evaluation surface, and specification information into a third AI model that calculates the temperature distribution of a predetermined evaluation surface at a predetermined estimated time, thereby calculating the temperature distribution of the extracted evaluation surface at a predetermined estimated time, predicting the amount of heat generated by the heat source on the substrate based on the comparison result between the temperature distribution image and the calculated temperature distribution, and calculating the average value of the multiple predicted heat amounts.

[0133] This allows the prediction method to predict the amount of heat generated by heat-generating elements in the substrate, even when the substrate temperature is in a non-steady state, by using multiple temperature distribution images.

[0134] <Technology 5> The prediction method described in Technology 1 involves acquiring multiple temperature distribution images at different times, calculating the temperature difference between a first temperature distribution image acquired at a first time and a second temperature distribution image acquired at a second time, a predetermined time after the first time, setting the amount of heat generated by the heat-generating elements on the extracted evaluation surface, setting a first estimated time for the temperature distribution of the extracted evaluation surface, and inputting the set amount of heat generated, the first estimated time, and the specification information into a third AI model that calculates the temperature distribution of a predetermined evaluation surface at a predetermined estimated time based on inputs of the amount of heat generated by the heat-generating elements set on a predetermined evaluation surface in the analysis model, a predetermined estimated time for the temperature distribution of the predetermined evaluation surface, and specification information. This calculates the first temperature distribution of the extracted evaluation surface at the set first estimated time, and inputting the set amount of heat generated, the second estimated time, a predetermined time after the first estimated time, and specification information into the third AI model. This calculates the second temperature distribution of the extracted evaluation surface at the second estimated time, and based on the temperature difference and the comparison result of the temperature difference between the first and second temperature distributions, the amount of heat generated by the heat-generating elements contained in the substrate is predicted.

[0135] As a result, the prediction method can predict the amount of heat generated by heat-generating elements in the substrate by comparing the temperature difference between temperature distribution images with the temperature difference between calculated temperature distributions, even when the substrate temperature is in a non-steady state.

[0136] <Technology 6> The prediction method described in Technology 2 or 3 sets an evaluation surface for the analysis model for substrate design, accepts the setting of specification information for the evaluation surface set for design, and calculates the temperature distribution of the evaluation surface set for design by inputting the predicted heat generation amount of the heat-generating elements contained in the substrate and the set specification information into the first AI model.

[0137] This allows the prediction method to use the predicted heat generation of the heat source and an AI model to design the circuit board.

[0138] <Technology 7> The prediction method described in Technology 6 involves setting an evaluation surface for the analysis model for substrate design, setting a target temperature distribution for the evaluation surface set for design, inputting the target temperature distribution into a fourth AI model that calculates specification information for a predetermined evaluation surface in response to the input of the temperature distribution set on the predetermined evaluation surface of the analysis model to calculate the specification information for the evaluation surface set for design, and inputting the predicted heat generation amount of the heat-generating elements contained in the substrate and the calculated specification information for the evaluation surface set for design into the first AI model.

[0139] This allows the prediction method to use the predicted heat generation of the heat source and an AI model to design the circuit board.

[0140] <Technology 8> In the prediction method described in any one of Techniques 1 to 7, the specification information further includes the thermal conductivity of the substrate and the thermal conductivity of the heating element.

[0141] As a result, the specification information includes the thermal conductivity of the substrate and the thermal conductivity of the heating element.

[0142] <Technology 9> The program causes the computing unit to acquire a temperature distribution image of a substrate containing heat-generating elements, extract evaluation surfaces from the temperature distribution image from an analysis model that includes specification information including the shape and layout of the substrate, set the amount of heat generated by the heat-generating elements contained in the substrate, calculate the temperature distribution of the extracted evaluation surfaces based on the set amount of heat generated and the specification information, and predict the amount of heat generated by the heat-generating elements contained in the substrate based on the comparison result between the temperature distribution image and the calculated temperature distribution.

[0143] This allows the program to achieve the same effect as Technique 1.

[0144] The functions of the above-described embodiment can also be realized by supplying programs and applications for realizing the functions of the above-described embodiment to a system or device using a network or storage medium, and by having one or more processors in the computer of that system or device read and execute the programs.

[0145] Furthermore, the functions of the above-described embodiment may be implemented by a circuit that implements one or more functions (for example, an Application Specific Integrated Circuit (hereinafter referred to as "ASIC") or an FPGA).

[0146] While embodiments have been described above with reference to the attached drawings, this disclosure is not limited to such examples. It is clear to those skilled in the art that various modifications, alterations, substitutions, additions, deletions, and equivalents can be conceived within the scope of the claims, and these are also understood to fall within the technical scope of this disclosure. Furthermore, the components of the embodiments described above can be combined in any way without departing from the spirit of the invention. [Industrial applicability]

[0147] The technology disclosed herein is useful as a prediction method and program. [Explanation of symbols]

[0148] 10 Arithmetic unit 11 processors 12 memory 13. Communication equipment 14 Input devices 15 External Interface Device 16 Display device 17 Image acquisition unit 18 Internal Interface Device 19 Cameras 20 Networks 21 Cloud

Claims

1. We acquire a temperature distribution image of the substrate including the heating element. From the analysis model, which includes specification information including the shape and layout of the substrate, the evaluation surface of the temperature distribution image is extracted. The amount of heat generated by the heating element contained in the substrate is set, Based on the set heat generation amount and the specification information, the temperature distribution of the extracted evaluation surface is calculated. Based on the comparison result between the temperature distribution image and the calculated temperature distribution, the amount of heat generated by the heating element contained in the substrate is predicted. Prediction method.

2. The first AI model calculates the temperature distribution of a predetermined evaluation surface in accordance with the input of the heat generation amount of a heat source set on the predetermined evaluation surface of the analysis model and the specification information, by inputting the set heat generation amount and the specification information, the extracted temperature distribution of the evaluation surface is calculated. The prediction method according to claim 1.

3. The temperature distribution of the extracted evaluation surface is set, A second AI model calculates the amount of heat generated by a heat source on a predetermined evaluation surface in response to inputting the temperature distribution set on the predetermined evaluation surface of the analysis model and the specification information, by inputting the set temperature distribution and the specification information, the amount of heat generated by the heat source on the extracted evaluation surface is calculated. The calculated heat generation amount of the heat-generating element on the extracted evaluation surface is set as the heat generation amount to be input to the first AI model. The prediction method according to claim 2.

4. Multiple temperature distribution images are acquired at different time points. For each of the multiple temperature distribution images, Setting the amount of heat generated by the heat source on the extracted evaluation surface, Setting the estimated time for the temperature distribution of the extracted evaluation surface, A third AI model calculates the temperature distribution of a predetermined evaluation surface at a predetermined estimated time in response to inputs of the heat generation amount of a heat source set on a predetermined evaluation surface of the analysis model, a predetermined estimated time for the temperature distribution of the predetermined evaluation surface, and the specification information, by inputting the set heat generation amount, the estimated time, and the specification information, the calculation of the extracted temperature distribution of the evaluation surface at the set estimated time, Based on the comparison result between the temperature distribution image and the calculated temperature distribution, the amount of heat generated by the heating element contained in the substrate is predicted and performed. The average value of the multiple predicted heat generation amounts is calculated. The prediction method according to claim 1.

5. Multiple temperature distribution images are acquired at different time points. From among the multiple temperature distribution images, the temperature difference between the first temperature distribution image acquired at a first time and the second temperature distribution image acquired at a second time, a predetermined time after the first time, is calculated. The amount of heat generated by the heating element on the extracted evaluation surface is set, The first estimated time for the temperature distribution of the extracted evaluation surface is set, A third AI model calculates the temperature distribution of a predetermined evaluation surface at a predetermined estimated time in response to inputs of the heat generation amount of a heating element set on a predetermined evaluation surface of the analysis model, a predetermined estimated time for the temperature distribution of the predetermined evaluation surface, and the specification information. By inputting the set heat generation amount, the first estimated time, and the specification information into the third AI model, the first temperature distribution of the extracted evaluation surface at the set first estimated time is calculated. By inputting the set heat generation amount, the second estimated time (after a predetermined time interval from the first estimated time), and the specification information into the third AI model, the second temperature distribution of the extracted evaluation surface at the second estimated time is calculated. Based on the temperature difference and the comparison result of the temperature difference between the first temperature distribution and the second temperature distribution, the amount of heat generated by the heating element contained in the substrate is predicted. The prediction method according to claim 1.

6. The evaluation surface of the analysis model for the design of the substrate is set, The system accepts the setting of the specification information for the evaluation surface set for the design, By inputting the predicted heat generation amount of the heating element contained in the substrate and the set specification information into the first AI model, the temperature distribution of the evaluation surface set for the design is calculated. The prediction method according to claim 2.

7. The evaluation surface of the analysis model for the design of the substrate is set, The target temperature distribution of the evaluation surface set for the design is set, The fourth AI model calculates the specification information for the predetermined evaluation surface in response to the input of the temperature distribution set on the predetermined evaluation surface of the analysis model, by inputting the target temperature distribution, the specification information for the evaluation surface set for the design is calculated. The predicted amount of heat generated by the heating element contained in the substrate and the calculated specification information of the evaluation surface set for the design are input into the first AI model. The prediction method according to claim 6.

8. The aforementioned specification information further includes the thermal conductivity of the substrate and the thermal conductivity of the heating element. The prediction method according to claim 1.

9. In the computing unit, Obtain a temperature distribution image of the substrate including the heating element. From the analysis model, which includes specification information including the shape and layout of the substrate, the evaluation surface of the temperature distribution image is extracted. The amount of heat generated by the heating element included in the substrate is set. Based on the heat generation amount set and the specification information, the temperature distribution of the extracted evaluation surface is calculated. Based on the comparison result between the temperature distribution image and the calculated temperature distribution, the amount of heat generated by the heating element contained in the substrate is to be predicted. program.

Images