Operation verification method

Abstract

To provide a technique for facilitating the generation of a script for performing operational verification in an emulator.SOLUTION: An instruction for generating a script S for verifying whether or not a mechanism to be controlled operates correctly is inputted to a large-scale language model 60 to obtain the script S output from the large-scale language model 60. Thereafter, the obtained script S is inputted to an emulator 50. Then, an operation executed by the emulator 50 is verified. Thus, by using the large-scale language model 60, it is possible to easily generate the script S for operational verification. A workload imposed on operational verification using the emulator 50 can be thereby mitigated.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a technique for verifying whether a mechanism to be controlled in an emulator operates correctly. [Background technology]

[0002] Conventionally, when manufacturing industrial machinery such as semiconductor manufacturing equipment, each mechanism within the equipment is operated to verify whether it is being controlled correctly in order to debug the control program. However, this verification cannot be performed until the assembly of the equipment is complete. Furthermore, even after the assembly of the equipment is complete, if there are multiple mechanisms within the equipment, it is difficult to simultaneously verify those mechanisms using a single piece of equipment.

[0003] For this reason, in recent years, emulators such as SILS (Software In the Loop Simulation) that mimic the operation of the actual device have been built on computers, and the operation is verified using these emulators. Using emulators makes it possible to debug control programs before the assembly of the actual device is complete. Emulators also make it possible to verify multiple mechanisms simultaneously. This shortens the lead time for manufacturing the actual device.

[0004] A conventional operation verification technique using an emulator is described in, for example, Patent Document 1. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Special Publication No. 2005-500618 Summary of the Invention [Problem to be solved by the invention]

[0006] When verifying the operation of each mechanism in an emulator, a script that serves as a command for operating the mechanism is input into the emulator. However, creating a script requires careful consideration of whether it sufficiently covers the operations to be verified, placing a heavy burden on the person in charge. In particular, when a mechanism is complex, the script to be created becomes more complex.

[0007] Therefore, an object of the present invention is to provide a technique that enables easy generation of a script for verifying operation in an emulator. [Means for solving the problem]

[0008] In order to solve the above problems, the first invention of the present application is an operation verification method for verifying whether a mechanism to be controlled operates correctly in an emulator that imitates the operation of a real device on a computer, and includes the steps of: (a) inputting an instruction to create a script for verifying the operation of the mechanism into a large-scale language model; (b) obtaining a script output from the large-scale language model in response to the input of step (a); (c) inputting the script obtained by step (b) into the emulator; and (d) verifying the operation of the mechanism executed in the emulator in response to the input of step (c).

[0009] A second aspect of the present invention is the operation verification method of the first aspect, wherein in the step (a), a specific operation of the mechanism is designated and input to the large-scale language model.

[0010] A third invention of the present application is an operation verification method according to the first or second invention, further comprising, before step (a), a training step of inputting training data into the large-scale language model, wherein the training data includes multiple combinations of information identifying the mechanism and a script corresponding to the mechanism.

[0011] A fourth aspect of the present invention is the operation verification method of the third aspect, wherein the learning data further includes information indicating a mechanical configuration of the mechanism.

[0012] A fifth aspect of the present invention is an operation verification method according to the third aspect of the present invention, further comprising, after step (d), a step (e) of modifying the script and inputting the modified script into the large-scale language model as training data.

[0013] A sixth aspect of the present invention is an operation verification method according to the fifth aspect of the present invention, in which the steps (a) to (e) are repeated.

[0014] The seventh invention of the present application is an operation verification method of the first or second invention, which further includes, after step (d), the steps of (f) inputting the script output from the large-scale language model into the real device, and (g) verifying the operation of the mechanism executed in the real device based on the input of step (f). [Effects of the Invention]

[0015] According to the first to seventh aspects of the present invention, a script for verifying operation in an emulator can be easily generated by using a large-scale language model, thereby reducing the workload involved in verifying operation using an emulator.

[0016] In particular, according to the second aspect of the present invention, a script for verifying a specific operation can be generated with high accuracy from a large-scale language model.

[0017] In particular, according to the third aspect of the present invention, it is possible to improve the accuracy of the script output from the large-scale language model.

[0018] In particular, according to the fourth aspect of the present invention, it is possible to further improve the accuracy of the script output from the large-scale language model.

[0019] In particular, according to the fifth aspect of the present invention, it is possible to further improve the accuracy of the script output from the large-scale language model.

[0020] In particular, according to the sixth aspect of the present invention, it is possible to further improve the accuracy of the script output from the large-scale language model. [Brief explanation of the drawings]

[0021] [Figure 1] FIG. 1 illustrates a configuration of an operation verification system. [Figure 2] 10 is a flowchart showing a flow of operation verification. [Figure 3] FIG. 10 is a diagram illustrating a learning process. [Figure 4] FIG. 10 is a diagram showing how a script for operation verification is obtained using a large-scale language model. DETAILED DESCRIPTION OF THE INVENTION

[0022] Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

[0023] <1. Operation verification system> 1 is a diagram showing the configuration of an operation verification system 1 capable of executing an operation verification method according to one embodiment of the present invention. This operation verification system 1 is a system that verifies control operations on an emulator 50 in order to debug a control program P that controls the real device 100, such as when the real device 100 is started up. The real device 100 is, for example, an exposure device that exposes substrates such as semiconductor wafers and printed circuit boards. However, the real device 100 may also be other industrial machinery such as a cleaning device that cleans substrates or a digital printing machine.

[0024] 1, the behavior verification system 1 includes a computer 10, an input unit 20, a display unit 30, and a large-scale language model 60. The computer 10 is an information processing device having a processor such as a CPU, a memory such as a RAM, and a storage unit such as a hard disk drive.

[0025] The computer 10 has a control unit 40 that controls the operation of the real device 100 and the emulator 50. The control unit 40 is realized by the CPU of the computer 10 operating in accordance with a control program P installed in the computer 10. The control unit 40 outputs a script S, which serves as an operation command, to each mechanism of the real device 100 and the emulator 50. The script S is source code that describes the operation of the mechanism to be controlled in a predetermined programming language. Therefore, the script S is composed of text data.

[0026] The computer 10 also has an emulator 50 that imitates (simulates) the operation of the real device 100. The emulator 50 is, for example, a software in the loop simulation (SILS). The emulator 50 is constructed in a virtual space based on design information related to the control of multiple mechanisms included in the real device 100. The emulator 50 operates in accordance with a script S output from the control unit 40.

[0027] The display unit 30 is a device that displays the simulation by the emulator 50. For example, a liquid crystal display is used for the display unit 30. The display unit 30 is electrically connected to the computer 10. The emulator 50 operates a part to be controlled in a three-dimensional model corresponding to the actual device 100. The display unit 30 displays the movement of the three-dimensional model on a screen.

[0028] The input unit 20 is a device for inputting various information to the computer 10. For example, a keyboard or a mouse is used as the input unit 20. The input unit 20 is electrically connected to the computer 10. A user of the operation verification system 1 inputs information to the computer 10 by operating the input unit 20 while referring to the display unit 30. The display unit 30 and the input unit 20 may be realized by a single device such as a touch panel display.

[0029] The computer 10 is connected to large language models 60 via a network N. The large language models 60 are machine learning models that have been trained on a huge amount of text data. It is desirable that the large language models 60 are commercially available and capable of being retrained. In this operation verification system 1, a script S used for operation verification in an emulator 50 is generated using the large language models 60.

[0030] <2. Operation verification method> The following describes a method for verifying operation using the large-scale language model 60. Fig. 2 is a flowchart showing the flow of operation verification.

[0031] When performing operation verification, first, a script S is trained in the large-scale language model 60 (step S1). FIG. 3 is a diagram showing the learning process in step S1. As shown in FIG. 3, in step S1, training data TD, which is training data, is input to the large-scale language model 60. The training data TD includes multiple combinations of mechanism information D1 that identifies the mechanism that is the target of operation verification of the real device 100 and the emulator 50, and script information D2 that indicates the script S corresponding to that mechanism. The user inputs the training data TD to the large-scale language model 60 from the computer 10 via the network N, for example, by operating the input unit 20.

[0032] The mechanism information D1 is text data that identifies the mechanism to be verified. As shown in FIG. 3, the mechanism to be verified is, for example, a motor-driven mechanism. The training data TD may include information indicating the mechanical configuration of the drive mechanism. In the example of FIG. 3, the drive mechanism is composed of a motor, a positive limit sensor, a negative limit sensor, and an origin sensor. The information indicating the mechanical configuration may be input to the large-scale language model 60 as natural language, or may be input in a format capable of retaining structural information, such as yaml, json, or xml.

[0033] The script information D2 includes one or more scripts S for verifying the operation of the mechanism. In the example of Fig. 3, the script information D2 indicates four operations: "movement in the + direction by a predetermined amount," "movement in the - direction by a predetermined amount," "movement in the + direction until the + side limit sensor turns OFF," and "movement in the - direction until the - side limit sensor turns OFF." The script information D2 includes scripts S that describe these operations in a predetermined programming language.

[0034] When training data TD is input, the large-scale language model 60 learns the relationship between the mechanism included in the training data TD and the script S. As a result, the large-scale language model 60 is customized so that when text data indicating a mechanism to be verified is input, the large-scale language model 60 can output a script S for verifying the operation of the mechanism.

[0035] The mechanism information input during learning in step S1 may be the mechanism whose operation is verified in step S5 described below, or a similar mechanism. For example, when verifying the operation of a new model of real device 100, in step S1, information about the mechanism of an old model and a script S used during the operation verification of the mechanism of the old model may be input to large-scale language model 60 as training data TD.

[0036] Furthermore, in step S1, supplementary comments (text data) may be added to the training data TD and input to the large-scale language model 60. In this way, the large-scale language model 60 can also learn the content of the comments. Therefore, the large-scale language model 60 can also reflect the content of the comments in the output of the script S.

[0037] 4 is a diagram showing how a script S for operation verification is obtained using a large-scale language model 60 for which the learning process of step S1 has been completed. In order to verify the operation in the emulator 50, the user inputs an instruction I for creating the script S to the large-scale language model 60 as shown in FIG. 4 (step S2). The user operates the input unit 20, for example, to input the instruction I for creating the script S to the large-scale language model 60 from the computer 10 via the network N.

[0038] A user can input creation instructions I for a script S in natural language to the large-scale language model 60. The creation instructions I for a script S include at least text that specifies the mechanism that is the target of operation verification. The user may also specify a specific operation of the mechanism and input it to the large-scale language model 60. In other words, the creation instructions I for a script S may include text that represents a specific operation of the mechanism.

[0039] When an instruction I for creating a script S is input, the large-scale language model 60 outputs the script S in response to the input (step S3). Specifically, the large-scale language model 60 outputs a script S for operation verification corresponding to the mechanism specified in the instruction I for creating the script S. The output script S is transmitted from the large-scale language model 60 to the computer 10 and displayed on the display unit 30. This allows the user to easily obtain a script S suitable for verifying the operation of the mechanism.

[0040] Next, the user inputs the script S output from the large-scale language model 60 to the emulator 50 (step S4). Then, the relevant mechanism of the emulator 50 operates in accordance with the input script S. The operation of the emulator 50 is displayed on the display unit 30.

[0041] The user verifies whether the operation of the mechanism executed in the emulator 50 is correct by checking the display unit 30 (step S5). If the operation of the mechanism displayed on the display unit 30 is incorrect, it is possible that there is a defect in the control program P or the emulator 50. The user debugs the control program P or the emulator 50 based on the verification results.

[0042] If the script S output from the large-scale language model 60 in step S3 is inappropriate or insufficient, the large-scale language model 60 may be re-trained (step S6). For example, if the script S output from the large-scale language model 60 does not sufficiently cover the operations that require verification, re-training is performed. In this case, the user corrects the script S output in step S3. Then, the corrected script S is included in the training data TD and input to the large-scale language model 60. The user may also input text data indicating the corrections to the script S to the large-scale language model 60. This causes the large-scale language model 60 to learn the corrected script S. As a result, it becomes possible to output an improved script S from the large-scale language model 60 at the next and subsequent verification times.

[0043] Furthermore, the processes of steps S2 to S6 described above may be repeated multiple times, thereby further improving the accuracy of the script S output from the large-scale language model 60.

[0044] In this operation verification system 1, the operation of each mechanism can be verified in the emulator 50 before the assembly of the real device 100 is completed. This allows early debugging of the control program P and shortens the manufacturing lead time of the real device 100. In particular, in this operation verification system 1, the script S for verifying operation in the emulator 50 can be easily generated by using the large-scale language model 60. This reduces the workload involved in creating the script S. Furthermore, the time required for operation verification can be further shortened.

[0045] Once the assembly of the real device 100 is complete, it becomes possible to verify the operation of the real device 100. At this time, the script S output from the large-scale language model 60 in the above-mentioned step S3 may be input from the control unit 40 to the real device 100 (step S7). Then, the operation executed by the mechanism of the real device 100 may be verified by inputting the script S (step S8).

[0046] <3. Modifications> Although one embodiment of the present invention has been described above, the present invention is not limited to the above embodiment.

[0047] The script S output from the large-scale language model 60 may be usable as is, or may be used after the user adds some information. For example, the large-scale language model 60 may output a script S without parameters. In this case, the user may add parameters to the script S output from the large-scale language model 60 and input it to the emulator 50.

[0048] 3 in the above embodiment shows an example of verifying a drive mechanism using a motor. However, the mechanism to be verified is not limited to this drive mechanism. The mechanism to be verified may be, for example, a fluid control mechanism using a pump and a valve, a temperature control mechanism using a heater, or other mechanisms.

[0049] Furthermore, some of the elements appearing in the above-described embodiments and modifications may be omitted, or other elements may be added, without departing from the spirit of the present invention. [Explanation of symbols]

[0050] 1: Operation verification system 10: Computer 20: Input section 30:Display section 40: Control section 50: Emulator 60: Large-scale language models 100: Actual device N: Network P: Control program S: Script TD: training data D1:Mechanism information D2: Script information I: Script creation instructions

Claims

1. 1. An operation verification method for verifying whether a mechanism to be controlled operates correctly in an emulator that imitates the operation of a real device on a computer, comprising: (a) inputting instructions for creating a script for verifying the operation of the mechanism into a large-scale language model; (b) obtaining a script output from the large-scale language model in response to the input of step (a); (c) inputting the script obtained in step (b) into the emulator; (d) verifying the operation of the mechanism executed in the emulator in response to the input of step (c); The operation verification method includes:

2. 2. The operation verification method according to claim 1, In the step (a), a specific operation of the mechanism is specified and input to the large-scale language model.

3. 3. The operation verification method according to claim 1, further comprising: a learning step, prior to the step (a), of inputting training data into the large-scale language model; and The learning data is information identifying the mechanism; a script corresponding to said mechanism; A method for verifying operation that includes multiple combinations of the above.

4. 4. The operation verification method according to claim 3, The learning data is Information indicating the mechanical configuration of the mechanism The operation verification method further includes:

5. 4. The operation verification method according to claim 3, After the step (d), (e) modifying the script and inputting the modified script as training data into the large-scale language model. The operation verification method further comprises:

6. 6. The operation verification method according to claim 5, The operation verification method repeats the steps (a) to (e).

7. 3. The operation verification method according to claim 1, further comprising: After the step (d), (f) inputting the script output from the large-scale language model into the real device; (g) verifying the operation of the mechanism executed in the real device based on the input of (f); The operation verification method further comprises:

Images