Information processing method, computer program, and information processing device

Abstract

The present invention provides an information processing method, a computer program, and an information processing device for which high-speed processing such as simulation using a large-scale model can be expected.　According to the present embodiment, provided is an information processing method in which an information processing device performs simulation using a large-scale model including a plurality of small-scale models. The information processing device stores, in a storage unit in advance, correspondence between input data and output data of a prescribed small-scale model included in the large-scale model. When data is input to the prescribed small-scale model in the simulation, a proxy model, which represents the prescribed small-scale model in a local range including the input data, is generated on the basis of the data stored in advance in the storage unit, and output data corresponding to the input data for the prescribed small-scale model is generated by using the generated proxy model.

Description

Information processing method, computer program, and information processing device

[0001] The present disclosure relates to an information processing method, a computer program, and an information processing device.

[0002] Patent Document 1 proposes a prediction method that includes an operation for obtaining a machine learning model that predicts performance metrics for the operation of a semiconductor manufacturing tool and an operation for receiving a process definition for manufacturing a product in the semiconductor manufacturing tool, and that uses one or more machine learning models to estimate the performance of the process definition used in the semiconductor manufacturing tool and presents the estimated results of the product manufacturing performance on a display.

[0003] Special Publication No. 2023-511122

[0004] The present disclosure provides an information processing method, a computer program, and an information processing device that are expected to speed up verification using large-scale models.

[0005] An information processing method according to one embodiment is an information processing method in which an information processing device performs a simulation using a large-scale model that includes multiple small-scale models, wherein the information processing device pre-stores in a memory unit the correspondence between input data and output data for a specified small-scale model included in the large-scale model, and when data is input to the specified small-scale model in a simulation, generates a proxy model that represents the specified small-scale model in a local range that includes the input data based on the data pre-stored in the memory unit, and uses the generated proxy model to generate output data that corresponds to the input data to the specified small-scale model.

[0006] According to the present disclosure, it is expected that verification using large-scale models can be performed at high speed.

[0007] FIG. 1 is a schematic diagram for explaining an overview of an information processing system according to the present embodiment. FIG. 2 is a schematic diagram for explaining an overview of an information processing system according to the present embodiment. FIG. 3 is a block diagram showing an example of a configuration of an information processing device according to the present embodiment. FIG. 4 is a schematic diagram for explaining a proxy model. FIG. 5 is a flowchart showing an example of the procedure of a proxy model generation process performed by the information processing device according to the present embodiment. FIG. 6 is a flowchart showing an example of the procedure of an abnormality determination process performed by the information processing device according to the present embodiment. FIG. 7 is a flowchart showing an example of the procedure of an optimization process performed by the information processing device according to the present embodiment.

[0008] Specific examples of information processing systems according to embodiments of the present disclosure will be described below with reference to the drawings. Note that the present disclosure is not limited to these examples, but is defined by the claims, and is intended to include all modifications within the meaning and scope equivalent to the claims.

[0009] 1 and 2 are schematic diagrams for explaining an overview of an information processing system according to this embodiment. The information processing system according to this embodiment is a system that reproduces a physical system existing in real space (physical space) in a virtual space (digital space or cyberspace) and manages and controls the physical system. This is a technology that can be called a digital twin. In this embodiment, the information processing system handles a digital twin in which a substrate processing apparatus 3 that performs substrate processing, such as manufacturing semiconductor wafers, is reproduced in a virtual space as an example of a physical system. However, the physical system handled by the information processing system according to this embodiment is not limited to the substrate processing apparatus 3. The information processing system can be applied to any physical system that can be virtualized, such as a digital twin.

[0010] 1 , a large-scale model 100 of the substrate processing apparatus 3 is created based on various data collected by the substrate processing apparatus 3. In this embodiment, the large-scale model includes a power supply model 101, a heater model 102, a chiller model 103, and the like, which are small-scale models that individually model a plurality of components that make up the substrate processing apparatus 3, such as a power supply, heater, and chiller. In this example, the large-scale model 100 also includes various other small-scale models, such as an RF (Radio Frequency) model 104, a surface temperature model 105, and a surface reaction model 106.

[0011] The large-scale model 100 according to this embodiment simulates the operation or processing of the corresponding components of these multiple small-scale models, and data on the simulation results is exchanged between the small-scale models, thereby enabling a simulation of the overall operation or processing of the substrate processing apparatus 3. In this example, the large-scale model 100 is configured to simulate an etching process performed in response to input data on the set voltage and set temperature of the substrate processing apparatus 3, and to output a predicted etching rate in the etching process as output data as a simulation result. Note that, although the present embodiment will be described using an etching process as an example, the process simulated using the large-scale model 100 is not limited to an etching process and may be various other processes.

[0012] In this example, power supply model 101 included in large-scale model 100 models the power supply of substrate processing apparatus 3, and provides output data of the results of simulating the operation of the power supply to heater model 102, chiller model 103, and RF model 104. Heater model 102 models the heater of substrate processing apparatus 3, simulates the operation of the heater based on input data from power supply model 101, and provides output data of the simulation results to surface temperature model 105. Chiller model 103 models the chiller of substrate processing apparatus 3, simulates the operation of the chiller based on input data from power supply model 101, and provides output data of the simulation results to surface temperature model 105.

[0013] The RF model 104 models a high-frequency circuit (not shown) provided in the substrate processing apparatus 3, simulates the operation of the high-frequency circuit based on input data from the power supply model 101, and provides output data of the simulation result to the surface temperature model 105. The surface temperature model 105 models the surface temperature characteristics of a substrate such as a wafer processed in the substrate processing apparatus 3, simulates changes in the surface temperature of the substrate based on input data provided by the heater model 102, the chiller model 103, and the RF model 104, and provides output data of the simulation result to the surface reaction model 106. In this example, an etching process is assumed as the substrate processing performed by the substrate processing apparatus 3. The surface reaction model 106 models the characteristics of the reaction on the substrate surface during the etching process, simulates the reaction on the substrate surface due to the etching process based on input data provided by the surface temperature model 105, and outputs etching rate data as the simulation result.

[0014] By simulating the operation or processing of the substrate processing apparatus 3 using such a large-scale model 100, it is possible to predict the results of substrate processing by the substrate processing apparatus 3, detect abnormalities in the substrate processing apparatus 3, or search for optimal settings for the substrate processing apparatus 3. For example, the etch rate obtained when the substrate processing apparatus 3 is operated at a certain set voltage and set temperature can be predicted by simulating the large-scale model 100 in accordance with input data for the set voltage and set temperature. Furthermore, for example, a predicted etch rate output by the large-scale model 100 in accordance with the input data for the set voltage and set temperature can be compared with a measured etch rate obtained as a result of operating the substrate processing apparatus 3 at the same set voltage and set temperature. If the difference between the predicted and measured values ​​exceeds a threshold, it can be determined that an abnormality has occurred in the substrate processing apparatus 3. Furthermore, for example, by calculating the error between the predicted etch rate output by the large-scale model 100 in accordance with the input data for the set voltage and set temperature and the target value of the etching rate desired by the user, and repeating the simulation by increasing or decreasing the set voltage and set temperature so as to reduce this error, it is possible to obtain set voltages and set temperatures that can bring the etching rate to or close to the target value.

[0015] These small-scale models included in the large-scale model 100 are created in advance by extracting input data to and output data from the components from data collected about the substrate processing apparatus 3, and adjusting the parameters of the model to reproduce the correspondence between input and output based on the extracted data. The small-scale model may be expressed by an arithmetic formula based on the physical characteristics of the components, or may be a machine learning model such as a neural network, or may be a model with a configuration other than these.

[0016] The multiple small-scale models included in the large-scale model 100 vary in the amount of calculation required for simulating each small-scale model, the amount of data handled in the processing, and the time required for processing. The amount of data required to generate each small-scale model and the time required to generate each small-scale model also vary. In this example, the power supply model 101, heater model 102, chiller model 103, RF model 104, and surface temperature model 105 included in the large-scale model 100 are models (light models) that require relatively little calculation, processing time, and generation time, while the surface reaction model 106 is a model (heavy model) that requires relatively much calculation, processing time, and generation time. Such heavy models require a large amount of data to generate a highly accurate model, and therefore require a significant amount of time to generate.

[0017] Therefore, in the information processing system according to this embodiment, as shown in FIG. 2 , an information processing device 1 that performs a simulation using a large-scale model 100 generates a surrogate model 120 that performs processing of a surface reaction model 106, which is a heavy model, on behalf of the large-scale model. The surrogate model 120 does not perform processing for the entire range of possible input data values ​​for the surface reaction model 106, but rather performs processing for the surface reaction model 106 only within a portion of the entire range of possible input data values. The surrogate model 120 is a model that reduces the amount of calculation, processing time, generation time, etc. by limiting the range of corresponding input data to a local range. Note that this embodiment describes an example in which a surrogate model 120 is generated for a heavy model, but this is not limited thereto, and the information processing device 1 may also generate a surrogate model 120 for a light model. The information processing device 1 may also generate surrogate models 120 for multiple small-scale models included in the large-scale model 100.

[0018] The information processing device 1 according to this embodiment performs a simulation using, for example, a large-scale model 100, and generates a surrogate model 120 of the surface reaction model 106 when processing of the surface reaction model 106 becomes necessary during the simulation. Therefore, at the start of the simulation, the power supply model 101, heater model 102, chiller model 103, RF model 104, and surface temperature model 105 of the large-scale model 100 must be generated in advance, but the surface temperature model 105 does not have to be generated in advance. In order to generate the surrogate model 120, the information processing device 1 acquires in advance the correspondence between input data and output data of the surface reaction model 106 and stores it in an input / output DB (database) 20. The input data and output data stored in the input / output DB 20 are, for example, data of a range and amount capable of generating a surface reaction model 106 corresponding to the entire range of input data.

[0019] The information processing device 1 starts a simulation using the large-scale model 100, and when the data output by the surface temperature model 105 and input to the surface reaction model 106 is calculated, the information processing device 1 reads out input data of a local range including the value of this input data and the corresponding output data from the input / output DB 20. The information processing device 1 generates a surrogate model 120 based on the input data and output data of the local range read out from the input / output DB 20. The information processing device 1 inputs the output data of the surface temperature model 105 into the generated surrogate model 120 and acquires the output data of the surrogate model 120 (predicted etch rate data in this example), thereby simulating the large-scale model 100.

[0020] In a simulation using the large-scale model 100, the information processing device 1 repeatedly performs calculations on each small-scale model to repeatedly calculate a predicted value of the etch rate. At this time, data from the surface temperature model 105 is repeatedly input to the surface reaction model 106, and the value of the input data changes each time. The information processing device 1 monitors changes in the value of the input data from the surface temperature model 105 to the surface reaction model 106, and when a value that exceeds the local range that the generated proxy model 120 can handle is input, the information processing device 1 updates the proxy model 120 by regenerating and replacing it. By updating the proxy model 120 as needed, the information processing device 1 can maintain the simulation using the large-scale model 100 regardless of how the value of the input data to the surface reaction model 106 changes.

[0021] <Device Configuration> Fig. 3 is a block diagram showing an example of the configuration of an information processing device 1 according to this embodiment. The information processing device 1 according to this embodiment is a device that predicts substrate processing results and detects abnormalities through simulations or the like using a large-scale model 100 that models, for example, a substrate processing device 3 as a target device. The information processing device 1 according to this embodiment is connected to the substrate processing device 3 via, for example, a communication cable, and can acquire data from the substrate processing device 3 and control the substrate processing device 3 in accordance with the results of predictions, abnormality detection, etc. However, the information processing device 1 does not need to be communicatively connected to the substrate processing device 3, and may only perform processing such as simulations using the large-scale model 100 without controlling the substrate processing device 3.

[0022] The information processing device 1 can be realized by installing a computer program according to this embodiment in a general-purpose information processing device such as a personal computer or a server computer. The information processing device 1 according to this embodiment is configured to include a processing unit 11, a storage unit 12, a communication unit 13, a display unit 14, and an operation unit 15. Note that although this embodiment will be described assuming that processing is performed by a single information processing device 1, the processing of the information processing device 1 may be distributed among a plurality of devices.

[0023] The processing unit 11 is configured using an arithmetic processing device such as a CPU (Central Processing Unit), an MPU (Micro-Processing Unit), a GPU (Graphics Processing Unit) or a quantum processor, a ROM (Read Only Memory), a RAM (Random Access Memory), etc. The processing unit 11 reads and executes a program 12a stored in the storage unit 12 to perform various processes, such as a simulation using a large-scale model 100 of the substrate processing apparatus 3 and a process of generating, as necessary, a substitute model of one or more small-scale models included in the large-scale model 100.

[0024] The storage unit 12 is configured using a large-capacity storage device such as a hard disk. The storage unit 12 stores various programs executed by the processing unit 11 and various data required for the processing of the processing unit 11. In this embodiment, the storage unit 12 stores a program 12a executed by the processing unit 11. The storage unit 12 also includes a model information storage unit 12b that stores information related to the large-scale model 100 of the substrate processing apparatus 3, and an input / output DB 20 that stores data used to generate the model.

[0025] In this embodiment, the program (computer program, program product) 12a is provided in a form recorded on a recording medium 99 such as a memory card or an optical disc, and the information processing device 1 reads the program 12a from the recording medium 99 and stores it in the storage unit 12. However, the program 12a may also be written to the storage unit 12, for example, during the manufacturing stage of the information processing device 1. Alternatively, the program 12a may be distributed by a remote server device or the like and acquired by the information processing device 1 via communication. For example, the program 12a may be read from the recording medium 99 by a writing device and written to the storage unit 12 of the information processing device 1. The program 12a may be provided in a form distributed via a network or in a form recorded on the recording medium 99.

[0026] The model information storage unit 12b of the storage unit 12 stores information about a large-scale model 100 of the substrate processing apparatus 3 that has been generated in advance, and one or more small-scale models included therein. The information about the large-scale model 100 may include, for example, information about what small-scale models are included in the large-scale model 100 and how the multiple small-scale models are connected. The information about the small-scale models may also include, for example, information indicating the configuration of the model and information about internal parameters of the model determined by machine learning or the like. The information processing apparatus 1 can configure the large-scale model 100 by reading out this information previously stored in the model information storage unit 12b, and use the large-scale model 100 for processing such as simulation of the substrate processing apparatus 3.

[0027] 1 and 2, for example, information such as the configuration and internal parameter values ​​of the power supply model 101, heater model 102, chiller model 103, RF model 104, and surface temperature model 105 is stored in advance in the model information storage unit 12b. Furthermore, the model information storage unit 12b may store information regarding the surface reaction model 106, such as the input data and output data for this model, but does not store the internal parameter values ​​in advance. However, the internal parameter values ​​of the proxy model 120 of the surface reaction model 106 generated in the proxy model generation process described below may be stored in the model information storage unit 12b.

[0028] Note that the method for generating the large-scale model 100 of the substrate processing apparatus 3 and the multiple small-scale models included therein is an existing technology, and therefore a detailed description thereof will be omitted in this embodiment. The large-scale model 100 and the small-scale models can be generated, for example, by machine learning processing using various data obtained from the substrate processing apparatus 3. These models may be generated by the information processing apparatus 1, or the information processing apparatus 1 may acquire information about models generated by an apparatus other than the information processing apparatus 1 and store it in the model information storage unit 12b. In any case, in the information processing system according to this embodiment, the large-scale model 100 of the substrate processing apparatus 3 as shown in FIG. 1 is generated in advance by an appropriate method. Information about the generated large-scale model 100 and the small-scale models is stored in advance in the model information storage unit 12b so that the information processing apparatus 1 can use the information when performing a simulation of the substrate processing apparatus 3, etc.

[0029] The input / output DB 20 is a database that stores data necessary for generating the proxy model 120. For example, data associating input data for the proxy model 120 with output data output by the proxy model 120 when this input data is input is stored in advance in the input / output DB 20. In the case of the information processing system shown in FIGS. 1 and 2 , for example, temperature distribution data as input data from the surface temperature model 105 to the surface reaction model 106 and etching rate data when an etching process is performed with respect to the input temperature distribution as output data of the surface reaction model 106 are associated and stored in the input / output DB 20. These data are collected in advance, for example, by a designer of the system, for example, by measuring the temperature distribution and etching rate when an etching process is performed in the substrate processing apparatus 3, and are stored in the input / output DB 20.

[0030] Furthermore, the input / output DB 20 of the information processing device 1 according to this embodiment associates input data and output data into a set, and stores the plurality of sets of data in groups based on the values ​​of each input data. For example, the information processing device 1 can group the sets of input data and output data by classifying the input data into a plurality of clusters using a clustering technique based on unsupervised learning. Furthermore, for example, the information processing device 1 may perform grouping based on the range of values ​​contained in the input data, or may sort and store the data in ascending or descending order of any value contained in the input data. The grouping of sets of input data and output data may be performed based on the values ​​of the output data rather than the values ​​of the input data. Data stored in the input / output DB 20 may be added and deleted at appropriate times. When adding data, the information processing device 1 classifies the added data into appropriate groups according to predetermined rules and stores them.

[0031] The communication unit 13 is connected to the substrate processing apparatus 3 via a cable such as a communication line or a signal line, and transmits and receives data to and from the substrate processing apparatus 3 via this cable. In this embodiment, the information processing apparatus 1 can, for example, acquire data obtained when the substrate processing apparatus 3 performs substrate processing, and perform processing such as predicting the results of the substrate processing by a simulation using the large-scale model 100. Furthermore, for example, the information processing apparatus 1 can determine setting values ​​for the substrate processing apparatus 3 by a simulation using the large-scale model 100, and control the operation of the substrate processing apparatus 3 by transmitting the determined setting values ​​to the substrate processing apparatus 3. The communication unit 13 transmits data provided by the processing unit 11 to the substrate processing apparatus 3, and also receives data transmitted from the substrate processing apparatus 3 and provides the received data to the processing unit 11.

[0032] The display unit 14 is configured using a liquid crystal display or the like, and displays various images, characters, etc. based on the processing of the processing unit 11. In the present embodiment, the display unit 14 displays, for example, information regarding the results of a simulation using the large-scale model 100, and information regarding the operating status of the substrate processing apparatus 3, etc. The operation unit 15 accepts user operations and notifies the processing unit 11 of the accepted operations. For example, the operation unit 15 accepts user operations via an input device such as a mechanical button or a touch panel provided on the surface of the display unit 14. Furthermore, for example, the operation unit 15 may be an input device such as a mouse and a keyboard, and these input devices may be configured to be detachable from the information processing apparatus 1.

[0033] The storage unit 12 may be an external storage device connected to the information processing device 1. The information processing device 1 may be a multi-computer including multiple computers, or may be a virtual machine virtually constructed by software. The information processing device 1 is not limited to the above configuration, and may not include, for example, the display unit 14 and the operation unit 15.

[0034] In addition, in the information processing device 1 according to this embodiment, the processing unit 11 reads and executes the program 12a stored in the memory unit 12, whereby the simulation processing unit 11a, the proxy model generation unit 11b, the update determination unit 11c, the abnormality determination unit 11d, the display processing unit 11e, etc. are realized in the processing unit 11 as software functional units.

[0035] The simulation processing unit 11a performs a simulation using the large-scale model 100 of the substrate processing apparatus 3 based on the information stored in the model information storage unit 12b. The simulation processing unit 11a constructs multiple necessary small-scale models based on the information stored in the model information storage unit 12b and connects these multiple small-scale models to construct the large-scale model 100. The simulation processing unit 11a acquires input data for the large-scale model 100 based on, for example, simulation conditions set by a user, inputs the acquired input data to the corresponding small-scale model, and acquires output data output from the small-scale model. The simulation processing unit 11a inputs output data from one small-scale model to the next small-scale model and acquires output data from the next small-scale model, repeating this process according to the connection relationships of the multiple small-scale models included in the large-scale model 100, and sets the data output by the final small-scale model as output data for the large-scale model 100, and this output data is used as a simulation result. The simulation processing unit 11a can, for example, sequentially input input data that changes over time to the large-scale model 100 and acquire output data that changes over time from the large-scale model 100 as a simulation result.

[0036] When the simulation processing unit 11a performs a simulation using the large-scale model 100, the surrogate model generation unit 11b performs processing to generate a surrogate model 120 for a specific small-scale model included in the large-scale model 100. In this embodiment, the surrogate model 120 is a model that guarantees operation only in a local range relative to the entire range of input data corresponding to the small-scale model. When input data is provided to a specific small-scale model in a simulation of the large-scale model 100, the surrogate model generation unit 11b acquires sets of input data and output data for a local range including the input data from the input / output DB 20, and generates a surrogate model using the acquired multiple sets of data. The surrogate model generation unit 11b can use various methods to generate the surrogate model 120, such as DMD (Dynamic Mode Decomposition), SINDy (Sparse Identification of Nonlinear Dynamical Systems), the least squares method, spline interpolation, machine learning, or a genetic algorithm.

[0037] The update determination unit 11c performs processing to determine whether or not the proxy model 120 generated by the proxy model generation unit 11b needs to be updated. When input data is given to a specific small-scale model in a simulation of the large-scale model 100, the update determination unit 11c determines whether or not the input data is within the range of input data guaranteed by the generated proxy model 120. If the input data is within the guaranteed range, the update determination unit 11c determines that the proxy model 120 does not need to be updated, and if the input data is outside the guaranteed range, the update determination unit 11c determines that an update is necessary. If an update is necessary, the update determination unit 11c causes the proxy model generation unit 11b to generate a proxy model 120 corresponding to the new input data.

[0038] The abnormality determination unit 11d communicates with the substrate processing apparatus 3 via the communication unit 13, for example, to acquire measured values ​​related to substrate processing measured by sensors or the like of the substrate processing apparatus 3, and compares the measured values ​​with predicted values ​​predicted by the simulation processing unit 11a through a simulation using the large-scale model 100. The abnormality determination unit 11d calculates the difference between the measured values ​​and the predicted values, and if the difference value exceeds a predetermined threshold, determines that there is a possibility that an abnormality has occurred in the substrate processing apparatus 3.

[0039] The display processing unit 11e performs processing to display various characters and images on the display unit 14. In the present embodiment, the display processing unit 11e displays various information, such as information relating to the results of a simulation performed by the simulation processing unit 11a or information relating to the results of an abnormality determination of the substrate processing apparatus 3 performed by the abnormality determination unit 11d.

[0040] <Surrogate Model Generation Process> In the information processing system according to this embodiment, data is collected using various sensors or measuring devices, for example, for the substrate processing apparatus 3 that is the target of the simulation, and a large-scale model 100 that models the substrate processing apparatus 3 is generated based on the collected data. At this time, in the information processing system according to this embodiment, for example, for the large-scale model 100 shown in Figures 1 and 2, small-scale models other than the surface reaction model 106 are generated based on the collected data. Generation of these small-scale models may be performed by the information processing apparatus 1 or by a device different from the information processing apparatus 1. Furthermore, these small-scale models may have any configuration and may be generated by various methods, such as DMD, SINDy, the least squares method, spline interpolation, machine learning, or a genetic algorithm.

[0041] 1 and 2, the information processing apparatus 1 according to this embodiment stores in advance in the model information storage unit 12b small-scale models other than the surface reaction model 106 that have been generated in advance by an appropriate method. The information processing apparatus 1 also stores in advance in the input / output DB 20 a large amount of data for generating a proxy model 120 for the surface reaction model 106 included in the large-scale model 100. The data for generating the proxy model 120 is data obtained by extracting input data and output data for the surface reaction model 106 from a large amount of data collected regarding the substrate processing apparatus 3 in the above-mentioned data collection, and associating the extracted input data with the output data.

[0042] By performing a simulation using the large-scale model 100, the information processing apparatus 1 can predict an etching rate as a result of an etching process performed by the substrate processing apparatus 3 when, for example, a set temperature and a set voltage are determined. In this simulation, the information processing apparatus 1 individually performs simulations for multiple small-scale models included in the large-scale model 100, and transfers the simulation results between the small-scale models to ultimately calculate a predicted value of the etch rate. For example, in the case of the large-scale model 100 shown in FIGS. 1 and 2 , the information processing apparatus 1 first performs a simulation of the power supply model 101 and provides the results to the heater model 102, the chiller model 103, and the RF model 104. The information processing apparatus 1 provides the simulation results of the heater model 102, the chiller model 103, and the RF model 104 to the surface temperature model 105. The information processing apparatus 1 performs a simulation of the surface temperature model 105 based on the simulation results of the heater model 102, the chiller model 103, and the RF model 104, calculates predicted values ​​such as a surface temperature distribution of the substrate, and inputs the simulation results to the surface reaction model 106.

[0043] As described above, the surface reaction model 106 is not generated in advance. When data of the simulation results of the surface temperature model 105 is input to the surface reaction model 106, the information processing device 1 according to this embodiment determines whether or not it is necessary to generate a proxy model 120 of the surface reaction model 106 (whether or not an update is necessary). If it is determined that it is necessary to generate the proxy model 120, the information processing device 1 acquires necessary data from the input / output DB 20 and generates the proxy model 120 of the surface reaction model 106. The proxy model 120 may be a model of any configuration, and can be generated by various methods such as DMD, SINDy, the least squares method, spline interpolation, machine learning, or a genetic algorithm.

[0044] 4 is a schematic diagram for explaining the surrogate model 120. Fig. 4 shows an example of the correspondence between the input data and output data of the surface reaction model 106 in a two-dimensional graph, with the horizontal axis (x-axis) of the graph representing the temperature, which is the input data, and the vertical axis (y-axis) representing the etch rate, which is the output data. Note that in this example, for the sake of simplicity, both the input data and the output data of the surface reaction model 106 are assumed to be one-dimensional (i.e., one numerical value), but this is not limiting, and the input data and the output data may be multidimensional vectors.

[0045] The solid curve in Fig. 4 indicates the correspondence between input and output of the surface reaction model 106. This curve corresponds to a graph of the relationship between input data and output data for the surface reaction model 106, which has been measured in advance in the substrate processing apparatus 3. The input / output DB 20 stores at least an amount of data that can reproduce this graph. As shown in Fig. 4, the correspondence between input and output of the surface reaction model 106 is complex, and if one were to attempt to model this graph using some kind of mathematical formula, it is estimated that a complex, high-dimensional formula would need to be used.

[0046] The surrogate model 120 generated by the information processing device 1 according to this embodiment is a model that reproduces only a portion of the input / output correspondence of the surface reaction model 106. The curve shown by the dashed line in Fig. 4 indicates the input / output correspondence of the surrogate model 120 that has been generated to represent the surface reaction model 106 for the local temperature range of x1 to x2. The guaranteed range of input data by the surrogate model 120 is narrowed to x1 to x2, but when modeling is performed only in this local range, for example, it is expected that modeling can be performed using low-dimensional mathematical expressions of approximately second to third order.

[0047] The information processing device 1 determines a local range including the values ​​of input data from the surface temperature model 105 to the surface reaction model 106, reads data within a range that can ensure the operation of the model within this local range from the input / output DB 20, and generates a surrogate model 120. For example, the information processing device 1 determines a local range x1 to x2 for input data x0 to the surface reaction model 106, and generates the surrogate model 120 by reading data within a range x3 to x4, for example, as data necessary for generating the surrogate model 120 that can ensure the operation within this local range x1 to x2 (where x3≦x1≦x0≦x2≦x4). The information processing device 1 inputs the input data from the surface temperature model 105 into the generated surrogate model 120 and obtains data on the predicted value of the etch rate output by the surrogate model 120. The information processing device 1 stores information about the generated surrogate model 120 in the storage unit 12, allowing it to be reused in subsequent processing.

[0048] If the input data x5 from the surface temperature model 105 to the proxy model 120 is subsequently within the guaranteed range x1 to x2 (x1≦x5≦x2), the information processing device 1 determines that it is not necessary to update the proxy model 120, and calculates the predicted value of the etch rate by reusing the already generated proxy model 120. On the other hand, if the input data x5 from the surface temperature model 105 to the proxy model 120 is subsequently outside the guaranteed range x1 to x2 (x5<x1 or x5>x2), the information processing device 1 determines that it is necessary to update the proxy model 120, and generates a proxy model 120 in a local range that includes the new input data x5.

[0049] Here, there are several possible methods for determining the range of data used to generate the surrogate model 120.

[0050] (Data Range Determination Method 1) In the first method, for example, the range of x±r is set as the guaranteed range for input data x to the surface reaction model 106, the range of x±rN is set as the range of data to be read from the input / output DB 20, and a user such as a designer or administrator of this system sets the values ​​of r and N in advance. The information processing device 1 accepts and stores input of setting values ​​of r and N from the user, reads data in the range of x±rN from the input / output DB 20 in accordance with the stored setting values ​​when generating the proxy model 120, and generates the proxy model 120 based on the read data.

[0051] (Data Range Determination Method 2) The second method is a method of determining the data range or the number of pieces of data based on a predetermined formula for calculating an error and a value of an allowable error. For example, the order of the surrogate model 120 to be generated is n, and the range of data used to generate the surrogate model 120 is D=x max -X min Let the number of data be M, and the error in the value output by the surrogate model 120 be Error(D, M, n)=D a ×M b ×n c Furthermore, if the allowable error is Acc, then the inequality Error(D, M, n)<Acc is obtained.

[0052] Here, for example, when the order n of the surrogate model 120 is set to 2 and the number M of data used for generation is also set, the only unknown quantity in the calculation formula for the error Error is the range D, so it is possible to calculate the range D in which the error Error is smaller than the allowable error Acc. After calculating the range D, the information processing device 1 can generate the surrogate model 120 using data in the range D that includes the input data x to the surface reaction model 106, for example, data in the range from x-D / 2 to x+D / 2.

[0053] Furthermore, based on the above inequality, it is also possible to determine the order n or the number of pieces of data M, rather than the data range D. For example, if the order n of the surrogate model 120 is set to 2 and the data range D is determined by an appropriate method (for example, based on the above-mentioned "data range determination method 1"), the number of pieces of data M can be calculated based on the above inequality. The same applies to the order n.

[0054] In this example, the formula for calculating the error is Error(D, M, n)=D a ×M b ×n c However, the present invention is not limited to this, and any calculation formula may be used. This calculation formula is determined in advance by, for example, a designer or administrator of the information processing system, and is stored in advance in the storage unit 12 of the information processing device 1.

[0055] Regardless of which of the above methods is adopted, if all of the data in the determined range is read from the input / output DB 20 to generate the proxy model 120, there is a risk that the generation time will increase if the amount of data is large. In such a case, the information processing device 1 may generate the proxy model 120 by appropriately thinning out the data rather than using all of the data in the determined range. The information processing device 1 may thin out the data, for example, by randomly selecting a predetermined number of data from the data in the determined range, or by selecting data at equal intervals in the storage order of the data stored in the input / output DB 20, or may thin out the data by other methods.

[0056] Whether or not the information processing device 1 thins out data can be set in advance by, for example, a user. For example, the user sets a maximum value for the number or amount of data to be used when generating the proxy model 120, and if the target data stored in the input / output DB 20 exceeds the set number or amount, the information processing device 1 thins out the data to generate the proxy model 120. Also, for example, the user can set an upper limit for the time required for the process of generating the proxy model 120, and the information processing device 1 can estimate the time required for generation from the configuration of the proxy model 120, the amount of parameters, etc., and thin out the data so as not to exceed the set upper limit.

[0057] The information processing device 1, which has generated the proxy model 120 using the data stored in the input / output DB 20, inputs input data for the surface reaction model 106 to the proxy model 120, obtains output data output by the proxy model 120 in response to this, and provides this output data to a subsequent small-scale model or the like as output data for the surface reaction model 106. The information processing device 1 stores information about the generated proxy model 120 in the storage unit 12, and can read out and use the generated proxy model 120 from the storage unit 12 if data subsequently input to the surface reaction model 106 is within a range guaranteed by the stored proxy model 120. It is preferable that the information processing device 1 stores, as the information about the proxy model 120, for example, information about the structure and internal parameters of the proxy model 120, and also stores information about the range of input data values ​​guaranteed by the proxy model 120.

[0058] Furthermore, the information processing device 1 needs to generate and update a surrogate model 120 when data input to the surface reaction model 106 is not within the range guaranteed by the stored surrogate model 120. At this time, the previous surrogate model 120 may be saved without being deleted from the storage unit 12. As a result, if the input to the surface reaction model 106 is within the range guaranteed by the saved surrogate model 120, the information processing device 1 can reuse the saved surrogate model 120 without generating a new surrogate model 120. The information processing device 1 may generate multiple surrogate models 120 with different guarantee ranges and store them in the storage unit 12. By generating and storing surrogate models 120 in this manner, the information processing device 1 has stored enough surrogate models 120 to cover the entire range of possible input data to the surface reaction model 106. Thereafter, it is sufficient to appropriately select and use the generated surrogate models 120, and the information processing device 1 does not need to generate a surrogate model 120.

[0059] 5 is a flowchart showing an example of the procedure of a proxy model generation process performed by the information processing apparatus 1 according to this embodiment. The simulation processing unit 11a of the processing unit 11 of the information processing apparatus 1 according to this embodiment performs a simulation of the substrate processing apparatus 3 using the large-scale model 100. In this simulation, the processing unit 11 determines whether or not data has been input from another small-scale model to a proxy target small-scale model included in the large-scale model 100, for example, the surface reaction model 106 (step S1). If there is no input to the surface reaction model 106 (S1: NO), the processing unit 11 continues to perform the simulation using the large-scale model 100 until data is input to the surface reaction model 106.

[0060] When data is input to the surface reaction model 106 (S1: YES), the update determination unit 11c of the processing unit 11 determines whether or not the proxy model 120 of the surface reaction model 106 needs to be updated based on the data input to the surface reaction model 106 (step S2). When the proxy model 120 needs to be updated (S2: YES), the proxy model generation unit 11b of the processing unit 11 reads data of a local range corresponding to the data input to the surface reaction model 106 from the data stored in the input / output DB 20 (step S3). The proxy model generation unit 11b generates the proxy model 120 of the surface reaction model 106 by appropriately using an existing model generation algorithm based on the data that associates the input data and output data read from the input / output DB 20 (step S4). The proxy model generation unit 11b stores information about the generated proxy model 120, such as information indicating the structure of the proxy model 120 and internal parameters, in the storage unit 12 (step S5), and proceeds to step S7. If the proxy model 120 does not need to be updated (S2: NO), the processing unit 11 reads out information about the proxy model 120 stored in the storage unit 12 (step S6), and proceeds to step S7.

[0061] After generating the surrogate model 120 in steps S3 to S5, or after reading out the surrogate model 120 in step S6, the processing unit 11 inputs the input data input to the surface reaction model 106 in step S1 to the surrogate model 120 (step S7). The processing unit 11 acquires output data output by the surrogate model 120 in response to the data input in step S7 (step S8). The processing unit 11 provides the output data of the surrogate model 120 acquired in step S8 as output data of the surface reaction model 106, for example, to a small-scale model in the next stage (step S9), and returns the process to step S1.

[0062] (Modification) In the above example, the information processing device 1 determines whether or not the proxy model 120 needs to be updated depending on whether or not the value of the input data to the small-scale model that is the proxy target of the proxy model 120 is within the guaranteed range of the proxy model 120, but the conditions for determining whether or not an update is necessary are not limited to this. The information processing device 1 according to the modification determines whether or not an update is necessary depending on, for example, the data accuracy required for the output data of the small-scale model that is the proxy target of the proxy model 120. The information processing device 1 according to the modification accepts settings related to the accuracy of the simulation or the accuracy of the proxy model 120 from the user.

[0063] The proxy model generation unit 11b of the information processing device according to the modified example generates a proxy model 120 capable of outputting output data according to the accuracy set by the user. The proxy model 120 can change the accuracy of the output data of the proxy model 120 by changing the configuration of the proxy model 120, for example, the number of parameters included in the proxy model 120 or the number of terms in an arithmetic expression. Furthermore, for example, the proxy model 120 can change the accuracy of the output data of the proxy model 120 by increasing or decreasing the number of pairs of input data and output data read from the input / output DB 20 and used for generation. Furthermore, for example, the proxy model generation unit 11b can change the accuracy of the output data of the proxy model 120 by increasing or decreasing the range of input data values ​​guaranteed by the proxy model 120. Note that the proxy model generation unit 11b may change the accuracy of the output data of the proxy model 120 by any of these means, or by employing other means, or by using a combination of multiple means.

[0064] When generating the proxy model 120, the proxy model generating unit 11b can use information of the proxy model 120 that has already been created, for example, internal parameters. For example, in a situation where the proxy model 120 expressed by a linear function (y=bx+c) has already been generated, the proxy model generating unit 11b can use information of the proxy model 120 that has already been created, for example, internal parameters of the proxy model 120, for example, when the proxy model 120 expressed by a linear function (y=bx+c) has already been generated, the proxy model generating unit 11b can use information of the proxy model 120 expressed by a quadratic function (y=ax 2 +bx+c). In this case, the proxy model generation unit 11b uses the values ​​of parameters (coefficients) b and c of the generated proxy model 120 of the linear function as parameters b and c of the proxy model 120 of the quadratic function, and determines the value of a new parameter a of the quadratic function using the data in the input / output DB 20, thereby making it possible to generate a proxy model 120 of the quadratic function with improved accuracy.

[0065] In this case, the proxy model generation unit 11b may not use the parameters b and c as they are, but may use them as initial values ​​of the parameters b and c of a new proxy model 120 of a quadratic function, and may determine the final values ​​of the parameters b and c from these initial values. Furthermore, the reuse of parameters as described above is one example, and the present invention is not limited to this, and the proxy model generation unit 11b may generate a new proxy model 120 by using any information of the proxy model 120 that has already been generated.

[0066] The information processing device 1 according to the modified example stores in the storage unit 12 information on the configuration, parameters, etc. of the generated proxy model 120, information on the range of values ​​of input data guaranteed by this proxy model 120, and information on the accuracy output by this proxy model 120, in association with each other. The information processing device 1 according to the modified example acquires the accuracy setting made by the user, and can determine whether or not the proxy model 120 needs to be updated depending on whether or not the stored proxy model 120 matches this accuracy setting.

[0067] <Application Examples> Several application examples of the information processing system according to this embodiment will be described. The first application example is determining an abnormality in a substrate processing apparatus 3. The information processing apparatus 1 can determine an abnormality in an arm apparatus using a large-scale model 100 that models an arm apparatus that transports substrates as the substrate processing apparatus 3. The large-scale model 100 receives, for example, the torque of the arm apparatus as input data and outputs a predicted value of the arm movement speed as output data. The information processing apparatus 1 acquires input data of the torque to the arm apparatus, performs a simulation of the large-scale model 100, and predicts the movement speed of the arm of the arm apparatus according to the input torque. The information processing apparatus 1 acquires an actual measurement value of the movement speed of the arm of the arm apparatus according to the same input data, and determines that an abnormality has occurred in the arm apparatus if the difference between the predicted and actual movement speed values ​​exceeds a predetermined threshold.

[0068] 6 is a flowchart showing an example of the procedure of an abnormality determination process performed by the information processing device 1 according to this embodiment. The abnormality determination unit 11d of the processing unit 11 of the information processing device 1 according to this embodiment communicates with the substrate processing device 3, for example, via the communication unit 13, and acquires input data on torque for the arm device (step S21). The simulation processing unit 11a of the processing unit 11 inputs the acquired input data to the large-scale model 100 (step S22), and performs a simulation of the operation of the substrate processing device 3 using the large-scale model 100. Note that at this time, the information processing device 1 according to this embodiment can perform a simulation for one or more small-scale models included in the large-scale model 100 using the above-mentioned representative model 120.

[0069] The simulation processing unit 11a acquires output data output by the large-scale model 100 through simulation (step S23). The abnormality determination unit 11d communicates with the substrate processing apparatus 3, for example, via the communication unit 13, and acquires an actual measurement value of the movement speed of the arm apparatus relative to the input data acquired in step S21 (step S24). The abnormality determination unit 11d calculates the difference between the predicted value of the output data acquired in step S23 and the actual measurement value of the output data acquired in step S24 (step S25). The abnormality determination unit 11d determines whether the difference calculated in step S25 exceeds a predetermined threshold value (step S26). If the difference exceeds the threshold value (S26: YES), the abnormality determination unit 11d determines that an abnormality exists in the substrate processing apparatus 3 (step S27) and terminates processing. If the difference does not exceed the threshold value (S26: NO), the abnormality determination unit 11d determines that no abnormality exists in the substrate processing apparatus 3 (step S28) and terminates processing.

[0070] The second use case is optimization of the settings of the substrate processing apparatus 3. A user sets a target value for the result of substrate processing in the substrate processing apparatus 3, such as an etch rate, and the information processing apparatus 1 optimizes the settings of the substrate processing apparatus 3 so as to achieve this target value. FIG. 7 is a flowchart showing an example of the procedure of the optimization process performed by the information processing apparatus 1 according to this embodiment. The processing unit 11 of the information processing apparatus 1 according to this embodiment determines initial values ​​for the settings of the substrate processing apparatus 3 to be optimized, i.e., initial values ​​of the input data to the large-scale model 100 (step S41). Note that the initial values ​​may be determined by the user, in which case the information processing apparatus 1 determines the initial values ​​by accepting input from the user.

[0071] The simulation processing unit 11a of the processing unit 11 inputs the determined or updated input data to the large-scale model 100 that models the substrate processing apparatus 3 (step S42), and performs a simulation of the operation of the substrate processing apparatus 3 using the large-scale model 100. At this time, the information processing apparatus 1 according to the present embodiment can perform a simulation using the above-mentioned proxy model 120 for one or more small-scale models included in the large-scale model 100. The simulation processing unit 11a acquires output data output by the large-scale model 100 through the simulation (step S43).

[0072] The processing unit 11 calculates the error between a preset target value and the value of the output data acquired in step S43 (step S44). Next, the processing unit 11 determines whether an optimization termination condition, such as the error calculated in step S44 being smaller than a predetermined threshold, is satisfied (step S45). Note that termination conditions may include not only those based on the error but also various other conditions, such as the processing time reaching an upper limit or the number of iterations reaching an upper limit.

[0073] If the termination condition is not satisfied (S45: NO), the processing unit 11 updates the input data to the large-scale model 100 based on an existing algorithm such as the steepest descent method or Newton's method so that the output data approaches the target value (step S46), and returns the process to step S42. The processing unit 11 repeats the processes of steps S42 to S46 until the termination condition is satisfied, thereby optimizing the input data to the large-scale model 100, i.e., optimizing the setting values ​​to be input to the substrate processing apparatus 3. If the termination condition is satisfied (S45: YES), the processing unit 11 stores the value of the input data at this time in the storage unit 12 as an optimal value (step S47), and terminates the process.

[0074] <Summary> In the information processing system according to the present embodiment, configured as described above, the information processing device 1 performs a simulation using a large-scale model 100 including a plurality of small-scale models. The information processing device 1 stores in advance in the input / output DB 20 of the storage unit 12 the correspondence between input data and output data for a predetermined small-scale model (e.g., the surface reaction model 106) included in the large-scale model 100. While other small-scale models included in the large-scale model 100 are generated in advance and information such as parameters is stored in the model information storage unit 12b, the predetermined small-scale model is not generated in advance, and the large-scale model 100 includes information regarding the predetermined small-scale model, such as the format of input / output data. When data is input to the predetermined small-scale model in a simulation using the large-scale model 100, the information processing device 1 generates, based on the data stored in the input / output DB 20, a proxy model 120 that represents the small-scale model within a local range including the input data. The information processing device 1 uses the generated proxy model 120 to input input data for a specified small-scale model to the proxy model 120 and obtains output data output by the proxy model 120, thereby generating output data corresponding to the input data for the specified small-scale model.

[0075] As a result, the information processing system according to this embodiment does not need to generate in advance small-scale models that are difficult to model or require a long time to model, for example, the entire range of inputs and outputs, and therefore can quickly perform a simulation using the large-scale model 100. Furthermore, the local proxy model 120 can be generated in a short time, and a simulation using the proxy model 120 can also be performed at high speed. Therefore, compared to a case in which all small-scale models of the large-scale model 100 are generated in advance and a simulation is performed, the information processing system according to this embodiment is expected to shorten the time required from modeling the target device to completing the simulation, and to speed up verification by simulation using a large-scale model, etc.

[0076] Furthermore, in the information processing system according to this embodiment, the information processing device 1 determines whether or not an update is required for the generated proxy model 120, depending on the data input to a predetermined small-scale model in a simulation using the large-scale model 100. If the information processing device 1 determines that an update is required, the information processing device 1 generates the proxy model 120 according to the data input to the small-scale model, based on the data stored in the input / output DB 20. By generating the proxy model 120 only when it is determined that generation is required, rather than generating it every time, the information processing system according to this embodiment is expected to shorten the time required for a simulation using the large-scale model 100, compared to generating the proxy model 120 every time.

[0077] Furthermore, in the information processing system according to this embodiment, the information processing device 1 stores the generated proxy model 120 in the storage unit 12. When determining whether or not the proxy model 120 needs to be updated, the information processing device 1 determines that updating is not necessary if a proxy model 120 corresponding to input data input to a predetermined small-scale model in a simulation using the large-scale model 100 has already been stored. The information processing device 1 generates output data for the predetermined small-scale model corresponding to the input data, using the generated proxy model 120 stored in the storage unit 12. This allows the information processing system according to this embodiment to reuse a proxy model 120 that has been generated once, which is expected to reduce the frequency of generating proxy models 120 and speed up simulations.

[0078] Furthermore, in the information processing system according to this embodiment, the information processing device 1 determines that the proxy model 120 needs to be updated when data input to a predetermined small-scale model exceeds the local range guaranteed by the generated proxy model 120. Furthermore, in the information processing system according to this embodiment, the information processing device 1 acquires information regarding the accuracy required for the data output by the proxy model 120, and determines that the proxy model 120 needs to be updated when the generated proxy model 120 does not satisfy this accuracy. By determining whether an update is necessary based on these conditions, the information processing system according to this embodiment can be expected to appropriately generate the proxy model 120.

[0079] Furthermore, in the information processing system according to this embodiment, the information processing device 1 selects a set of input data and output data to be used for generating the proxy model 120 according to the input data to a predetermined small-scale model from the sets of input data and output data stored in the input / output DB 20, and uses the selected data to generate the proxy model 120. As a result, the information processing system according to this embodiment can be expected to select appropriate data from the large number of data stored in the input / output DB 20 in order to generate the proxy model 120 that can guarantee output data within a local range of the input data.

[0080] Furthermore, in the information processing system according to this embodiment, the information processing device 1 groups sets of input data and output data related to a predetermined small-scale model that have been collected in advance according to the values ​​of the input data or output data, and stores the groups in the input / output DB 20. As a result, the information processing system according to this embodiment is expected to facilitate the selection of sets of input data and output data to be used in generating the proxy model 120.

[0081] Furthermore, in the information processing system according to this embodiment, the information processing device 1 compares actual measurements obtained by measuring the operation of a target device such as the substrate processing device 3 with predicted values ​​obtained by simulation using a large-scale model 100 that models the target device, and detects an abnormality in the target device based on the comparison results. The information processing device 1 determines that an abnormality exists in the target device, for example, when the difference between the actual measurements and the predicted values ​​exceeds a threshold. As a result, the information processing system according to this embodiment can be expected to accurately predict the operation of the target device through simulation using the large-scale model 100, and accurately determine an abnormality in the target device.

[0082] Furthermore, in the information processing system according to this embodiment, the information processing device 1 predicts output data for the target device through a simulation using the large-scale model 100 based on input data for the target device, calculates the error between the predicted value of the output data and a target value, and updates the input data for the target device based on the calculated error. The information processing device 1 repeatedly performs simulation using the large-scale model 100 and updates the input data based on the error to determine input data for the target device that will achieve the target value. As a result, the information processing system according to this embodiment can be expected to accurately determine input data that will enable the target device to achieve the target value through a simulation using the large-scale model 100, and to achieve the target value by operating the target device using the determined input data.

[0083] The embodiments disclosed herein are to be considered as illustrative in all respects and not restrictive. The scope of the present disclosure is defined by the claims, not by the above meaning, and is intended to include all modifications within the meaning and scope of the claims.

[0084] The matters described in each embodiment can be combined with each other. Furthermore, the independent claims and dependent claims described in the claims can be combined with each other in any and all combinations, regardless of the reference format. Furthermore, the claims use a format in which a claim references two or more other claims (multiple claim format), but this is not limited to this. A multiple claim (multi-multi claim) that references at least one other multiple claim may also be used.

[0085] REFERENCE SIGNS LIST 1 Information processing device (computer) 3 Substrate processing device 11 Processing unit 11a Simulation processing unit 11b Substitute model generation unit 11c Update determination unit 11d Abnormality determination unit 11e Display processing unit 12 Storage unit 12a Program (computer program) 12b Model information storage unit 20 Input / output DB 99 Recording medium 100 Large-scale model 101 Power supply model (small-scale model) 102 Heater model (small-scale model) 103 Chiller model (small-scale model) 104 RF model (small-scale model) 105 Surface temperature model (small-scale model) 106 Surface reaction model (predetermined small-scale model) 120 Substitute model

Claims

1. An information processing method in which an information processing device performs a simulation using a large-scale model composed of multiple small-scale models, wherein the information processing device: pre-stores in a memory unit the correspondence between input data and output data for a specific small-scale model included in the large-scale model; when data is input to the specific small-scale model in a simulation, generates a proxy model that represents the specific small-scale model in a local area including the input data based on the data pre-stored in the memory unit; and uses the generated proxy model to generate output data corresponding to the input data to the specific small-scale model.

2. The information processing method according to claim 1, further comprising: determining whether the generated representative model needs to be updated based on input data to the specified small-scale model; and, if it is determined that an update is necessary, generating a representative model according to the input data based on data previously stored in the memory unit.

3. The information processing method according to claim 2, further comprising: storing the generated one or more proxy models in a memory unit; determining that updating of the proxy models is unnecessary if a proxy model corresponding to the input data is stored in the memory unit; and generating the output data using the proxy model stored in the memory unit.

4. The information processing method according to claim 2, further comprising determining that the representative model needs to be updated when input data to the predetermined small-scale model exceeds a local range guaranteed by the representative model.

5. The information processing method according to claim 2, further comprising: acquiring information regarding the accuracy required for the output data output by the proxy model; and determining that the proxy model needs to be updated if the output data of the proxy model does not satisfy the accuracy related to the acquired information.

6. The information processing method according to claim 1, further comprising the steps of: selecting input data and output data to be used in generating the representative model from the input data and output data stored in the memory unit according to input data to the small-scale model; and generating the representative model based on the selected input data and output data.

7. The information processing method according to claim 6, wherein the storage unit stores groups of sets of input data and output data according to values ​​of the input data or output data.

8. The information processing method according to claim 1, further comprising: comparing actual measured values ​​of the operation of a target device with predicted values ​​obtained by a simulation using the large-scale model that models the target device; and detecting an abnormality in the target device based on the comparison results.

9. The information processing method of claim 1, further comprising the steps of: predicting output data of a target device for input data based on the input data to the target device by simulating the large-scale model that models the target device; calculating an error between the predicted value of the output data and a target value; updating the input data based on the calculated error; and determining input data to the target device that achieves the target value by repeating the prediction of the output data, the calculation of the error, and the updating of the input data.

10. A computer program that causes a computer to perform a simulation using a large-scale model composed of multiple small-scale models, the computer program executing the following processes: store in advance in a memory unit the correspondence between input data and output data for a specific small-scale model included in the large-scale model; when data is input to the specific small-scale model in a simulation, generate a proxy model that represents the specific small-scale model in a local area including the input data based on the data previously stored in the memory unit; and use the generated proxy model to generate output data corresponding to the input data to the specific small-scale model.

11. An information processing device having a processing unit that performs a simulation using a large-scale model consisting of multiple small-scale models, wherein the processing unit pre-stores in a memory unit the correspondence between input data and output data for a specific small-scale model included in the large-scale model, and when data is input to the specific small-scale model in a simulation, generates a proxy model that represents the specific small-scale model in a local area including the input data based on the data pre-stored in the memory unit, and uses the generated proxy model to generate output data corresponding to the input data to the specific small-scale model.

Images