Operation procedure verification device
The operation procedure verification device addresses the issue of unsafe AI-generated machine tool procedures by simulating and validating them against pre-registered abnormal states, ensuring safe and accurate execution.
Patent Information
- Application Number
- PCT/JP2024/000638
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2024-01-12
- Publication Date
- 2025-07-17
AI Technical Summary
Conventional AI systems, particularly generative AI, may provide inappropriate or unsafe operation procedures for machine tools due to hallucinations and lack of consideration for machine-specific settings, potentially leading to dangerous operations.
An operation procedure verification device that acquires machine information, simulates operations using a simulator, and validates the generated procedures against pre-registered abnormal states to ensure safety and accuracy.
Enhances the reliability and safety of operation procedures by verifying the validity of AI-generated instructions through simulation, preventing dangerous operations and ensuring accurate execution on actual machines.
Smart Images

Figure JP2024000638_17072025_PF_FP_ABST
Abstract
Description
Operation procedure verification device
[0001] An embodiment of the present invention relates to a technique for presenting an operating procedure for a machine tool (hereinafter also simply referred to as a machine).
[0002] Conventionally, when a user does not know how to use a machine or encounters a problem, an AI system is available that, when the user inputs the details, will suggest a solution based on pre-registered data.
[0003] JP 2023-71560 A
[0004] The use of generative AI, which can answer questions in various forms, such as natural sentences or programs, is being promoted in AI chat systems. However, the answers provided by generative AI are not always correct, and they may produce a seemingly plausible but inappropriate response known as hallucination. Furthermore, because generative AI is not a model based on the internal processing of machines and control devices, the behavior of the generative AI model may differ from the behavior of the actual machine. Furthermore, even for the same question, the appropriate solution may differ depending on the machine's settings, and generative AI may not be able to provide an appropriate answer because it is not guaranteed to take all conditions into account.
[0005] If the answer given by such a generated AI is implemented as is, the problem may not be solved, or in some cases the machine may operate dangerously. Therefore, a system that can present more appropriate operating procedures is desired.
[0006] In an embodiment, an operation procedure verification device acquires machine information including at least one of the configuration, settings, and status of a machine tool using a machine information acquisition unit, and acquires an operation procedure of the machine tool generated by AI using the operation procedure acquisition unit. At this time, the operation procedure verification device executes a simulation that simulates the operation of the machine tool based on the machine information and the operation procedure using a validity confirmation unit, and verifies the validity of the execution result of the simulation by comparing it with pre-registered abnormal states. Then, the operation procedure verification device outputs the confirmation result by the validity confirmation unit using an output unit.
[0007] FIG. 1 is a block diagram showing the functional configuration of an operation procedure verification device according to a first embodiment. FIG. 2 is a diagram illustrating input / output data of each functional unit in the operation procedure verification device according to the first embodiment. FIG. 3 is a block diagram showing the functional configuration of an operation procedure verification device according to a second embodiment. FIG. 4 is a block diagram showing the functional configuration of an operation procedure verification device according to a third embodiment. FIG. 5 is a flowchart illustrating a processing procedure of an operation procedure verification device according to the third embodiment. FIG. 6 is a first diagram illustrating input / output data of each functional unit in the operation procedure verification device according to the third embodiment. FIG. 7 is a second diagram illustrating input / output data of each functional unit in the operation procedure verification device according to the third embodiment. FIG. 8 is a diagram illustrating questions and answers to a generation AI according to a modified example. FIG. 9 is a first diagram illustrating input / output of a generation AI according to a modified example. FIG. 10 is a second diagram illustrating input / output of a generation AI according to a modified example. FIG. 11 is a diagram illustrating an example of an output from an operation procedure verification device according to a modified example.
[0008] The following describes an embodiment of the present invention. The operation procedure verification device of this embodiment uses a simulator that simulates machine operation based on machine information to verify the validity of answers provided by a generation AI regarding operation procedures, and also presents the results.
[0009] Here, generative AI is a machine learning model that has undergone additional learning (e.g., fine tuning or instruction tuning) to be able to generate machine operating procedures in response to questions about the machine based on data such as manuals, specifications, past inquiries and their responses regarding the target machine and its control device, or a model that can access the above data and generate answers.
[0010] Furthermore, the simulator in this embodiment operates based on the internal processing of the target machine and control device, and is equipped with an input / output interface that mimics the actual machine and control device. Note that simulations of operations related to physical laws operate based on an approximation model, so they do not operate exactly the same as the actual machine, but are closer to reality than the models possessed by the generative AI.
[0011] 1 is a block diagram showing the functional configuration of an operation procedure verification device 1 according to a first embodiment. The operation procedure verification device 1 is an information processing device (computer) equipped with a control unit 10 (processor), a storage unit 20 (memory), and various input / output interfaces. The operation procedure verification device 1 also includes a built-in generation AI or has an interface with an external generation AI.
[0012] The control unit 10 includes a question input unit 11, a machine information acquisition unit 12, an operation procedure acquisition unit 13, a validity confirmation unit 14, and an output unit 15, and these functional units operate by the control unit 10 reading and executing various software stored in the memory unit 20.
[0013] The question input unit 11 accepts questions from the user regarding the operation of the machine and inputs the question text to the generation AI. The questions are, for example, as follows. At this time, specific measurement data may be attached. - The execution time of program O0001 is long, so I would like to shorten it. - I would like to display a setting screen to correct the tool length, and I would like to move two axes synchronously, so please tell me how to set it up. - Alarm 100 has occurred, so I would like to cancel it. - The axes do not move even when the program is executed. - The path is not smooth, so I would like to make it smoother.
[0014] The machine information acquisition unit 12 acquires machine information including information related to at least one of the configuration, settings, and status of the target machine. Information related to the machine configuration includes, for example, the model, software version, number of axes, axis configuration, manufacturer, and manufacturing date. Information related to the machine settings includes, for example, setting parameters for each function, axis connection settings, machining programs, and programs for the Programmable Controller (PC). Information related to the machine status includes, for example, the display screen, alarm status, operation history, program execution history, and sensor information (temperature, pressure, and signals).
[0015] The operation procedure acquisition unit 13 acquires the operation procedure of the machine tool generated as an answer by the generation AI from the question content received by the question input unit 11 and the machine information acquired by the machine information acquisition unit 12.
[0016] Examples of operational procedures that the generation AI will respond to include: - Pressing a switch key. - Pressing a button displayed on the screen. - Entering numbers or characters (such as setting parameter 1 to 10.0). - Editing a program (such as inserting Gxx on line x). - Executing a program. - Restarting the machine. - Opening and closing the machine door. - Attaching and detaching a tool or workpiece.
[0017] The validity confirmation unit 14 executes a simulation using a simulator that simulates the operation of the machine based on the machine information and the operating procedure, and confirms the validity of the results of the simulation by comparing them with pre-registered abnormal conditions.
[0018] Abnormal conditions that will determine that the results of a simulation are invalid include, for example, the following, and various thresholds are set in advance: - The axis movement speed or spindle rotation speed exceeded the maximum speed or maximum acceleration. - The axis moved outside the operating range of the machine. - The tool collided with the workpiece or jig. - The axis position error became excessive. In other words, the machine was unable to follow a command for rapid movement. - The machine overloaded (overvoltage, overpressure, overheating, etc.). - An alarm occurred, indicating an invalid setting (for example, entering a negative value where a positive value should have been entered) or an invalid state (for example, a signal that should have been ON is OFF). - An attempt was made to change a setting that the user cannot change without purchasing an option. - The simulation terminated abnormally.
[0019] The output unit 15 outputs the operation procedure generated by the generation AI as an answer to the question, and the confirmation result by the validity confirmation unit 14.
[0020] 2 is a diagram illustrating input / output data of each functional unit in the operation procedure verification device 1 according to the first embodiment. To obtain an operation procedure as an answer, the generation AI prompt is set with a fixed phrase such as "Please create a machine operation procedure that answers the question based on the following machine information."
[0021] In response to this, the machine information acquisition unit 12 collects machine configuration and settings, program contents, program execution time, etc. as machine information and provides it to the generation AI. Then, when a question such as "The execution time of program O0001 is long, so I would like to shorten it" is input by the question input unit 11, the generation AI generates an operating procedure for changing function settings, such as adjusting the feed rate of an axis, from among multiple means based on the machine information and past data, and outputs it as a response to the question.
[0022] The operation procedure acquisition unit 13 acquires the operation procedure provided by the generation AI and provides it to the validity confirmation unit 14. The validity confirmation unit 14 executes a simulation based on this operation procedure and compares the execution result with the registered abnormal conditions described above to confirm whether or not an abnormal condition exists and obtains the details of the abnormal condition. The output unit 15 then outputs the confirmation result together with the operation procedure. For example, in the example of FIG. 2, the simulation result shows that the program is interrupted and an alarm is displayed.
[0023] According to the first embodiment, the operation procedure verification device 1 performs simulation to verify the validity of the machine operation procedure provided by the generating AI in response to a user's question. The verification result indicates the validity of the answer, for example, that if the answer provided by the generating AI is applied to the machine as is, the axis will move at maximum speed, resulting in a dangerous operation. Therefore, the user can determine in advance whether the answer will be safe even if applied to the machine.
[0024] Although the first embodiment is configured to acquire operation procedures as answers from the generation AI to questions, the present invention is not limited to this. For example, an operation procedure output by an external system equipped with the generation AI may be input to the operation procedure verification device 1, which may verify the validity of the operation procedure and present the validity to the user. In this case, the operation procedure verification device 1 may not include the question input unit 11. The generation AI may be replaced by various AI systems. For example, the operation procedure verification device 1 may verify the validity of operation procedures output by machine learning models such as rule-based AI, supervised learning, unsupervised learning, and reinforcement learning.
[0025] 3 is a block diagram showing the functional configuration of the operational procedure verification device 1 according to the second embodiment. In the second embodiment, the functions of the validity confirmation unit 14 and the output unit 15 are expanded compared to those in the first embodiment.
[0026] The validity confirmation unit 14 inputs the confirmation result of the validity of the operation procedure obtained from the generation AI to the generation AI, and asks the generation AI to evaluate whether the confirmation result is satisfactory as an answer to the original question. The output unit 15 outputs the operation procedure, the confirmation result of validity, and the evaluation of the confirmation result.
[0027] According to the second embodiment, the operating procedure verification device 1 can determine whether the confirmation result by the validity confirmation unit 14 is correct as an answer to the question, and therefore can provide a more appropriate answer to the user.
[0028] 4 is a block diagram showing the functional configuration of an operational procedure verification device 1 according to a third embodiment. In the third embodiment, the function of the validity checking unit 14 is expanded compared to the second embodiment.
[0029] The validity confirmation unit 14 repeats the process of asking the generation AI for another operation procedure to improve the evaluation by the generation AI. The number of repetitions may be until the generation AI determines that there is no more appropriate operation procedure, or until a predetermined number of repetitions is reached, or until at least one of these conditions is met. The output unit 15 outputs the optimal operation procedure obtained in the process of repetition, the validity confirmation result, and the evaluation of this confirmation result.
[0030] 5 is a flowchart illustrating the processing procedure of the operation procedure verification device 1 in the third embodiment. In step S1, the question input unit 11 accepts a question from a user and inputs it to the generation AI. In step S2, the machine information acquisition unit 12 inputs the acquired machine information to the generation AI. In step S3, the operation procedure acquisition unit 13 acquires an operation procedure as a response to the question from the generation AI.
[0031] In step S4, the validity confirmation unit 14 executes a simulation in accordance with the operation procedure acquired in step S3, and further inputs the validation result of the execution result to the generation AI. At this time, the validity confirmation unit 14 asks the generation AI for an evaluation of the validity confirmation result, and requests another operation procedure to improve the evaluation.
[0032] In step S5, the validity confirmation unit 14 determines whether a new operating procedure candidate has been presented by the generation AI. If the determination is YES, the process proceeds to step S6, and if the determination is NO, the process proceeds to step S7.
[0033] In step S6, the validity confirmation unit 14 determines whether the number of questions posed to the generated AI has reached a specified number. If the determination is YES, the process proceeds to step S7, and if the determination is NO, the process proceeds to step S4.
[0034] In step S7, the output unit 15 outputs the operation procedure, the validity confirmation result, and the evaluation by the generation AI.
[0035] FIG. 6 is a first diagram illustrating input / output data of each functional unit in the operation procedure verification device 1 of the third embodiment. The validity confirmation result for the operation procedure obtained in the first embodiment (FIG. 2) is further input to the prompt of the generation AI, and the confirmation result is evaluated by a question such as, "Please determine whether this result is a satisfactory result as an answer to the following question." Furthermore, a new operation procedure is obtained as a response by a question such as, "If the result is not satisfactory, please present another operation procedure based on this result." In this example, an alarm is generated and the operation is determined to be abnormal, and a revised operation procedure is generated by the generation AI by taking into account a new maximum speed condition.
[0036] 7 is a second diagram illustrating input / output data of each functional unit in the operation procedure verification device 1 according to the third embodiment. Regarding the operation procedure in which the axis feed rate is changed, the simulation result by the validity confirmation unit 14 shows that the program execution time is reduced from 50 seconds to 45 seconds. Therefore, in response to a prompt similar to that shown in FIG. 6, the generation AI evaluates the result as "satisfactory."
[0037] Furthermore, if there is a way to further reduce the time by using a prompt such as "Please let us know if you have a better operating procedure," we can expect to be able to present a different operating procedure that will be evaluated more highly by the generation AI.
[0038] The output unit 15 formats and presents the confirmation results and evaluation of the simulation along with the operation procedure generated by the generation AI.
[0039] According to the third embodiment, the operation procedure verification device 1 inputs the confirmation results of the simulation of the operation procedure obtained from the generation AI back into the generation AI, thereby increasing the amount of information taken into account by the generation AI and expecting more accurate and reliable output. As a result, the user can try out the operation procedure on the actual machine with peace of mind.
[0040] [Modifications] The first to third embodiments described above can be modified as appropriate, and the operating procedure verification device 1 may be configured, for example, as follows.
[0041] The question input unit 11 may accept input of images or a combination of images and text as the question content, rather than just text. This allows, for example, situations that are difficult to express in text to be input to the generation AI, and an appropriate answer can be obtained.
[0042] The question input unit 11 may input the question content to the generation AI, and may also ask the generation AI about necessary items of machine information that should be input. In this case, the machine information acquisition unit 12 acquires machine information corresponding to the necessary items answered by the generation AI, and inputs it to the generation AI. This allows the necessary and sufficient information to be efficiently input to the generation AI, and an appropriate answer can be obtained. Furthermore, since the necessary items are limited, the user may input the relevant information themselves.
[0043] 8 is a diagram illustrating an example of a question and answer to the generation AI in a modified example. By inputting "Please provide the equipment information necessary to create a machine operating procedure that answers the question" into the prompt, the generation AI answers the question "I would like to display a setting screen for correcting the tool length" with the necessary machine information items, such as the machine type and software version.
[0044] The question input unit 11 may request an answer in multiple language formats from the generation AI. In this case, the operation procedure acquisition unit 13 selects at least one of the multiple language formats and provides it to the validity confirmation unit 14 and the output unit 15, respectively. For example, when an answer is obtained in a natural language format and a programming language format, the answer can be presented to the user in natural language, and a simulation can be easily performed using the programming language. In this way, the operation procedure verification device 1 can prepare multiple outputs according to the application. Note that the generation AI pre-learns the correspondence between multiple language formats.
[0045] 9 is a first diagram illustrating the input and output of the generation AI in the modified example. Here, the machine information specified by the generation AI in FIG. 8 is input, and an operation procedure written in the form of a natural language and an operation procedure written in the form of a programming language are output as a pair.
[0046] It is also possible to ask the generation AI questions about items for confirming the results and have them output in advance, thereby enabling the validity confirmation unit 14 to grasp the items to be confirmed (for example, the displayed screen), i.e., the items to be input to the generation AI as simulation results.
[0047] The validity confirmation unit 14 may generate an image or video that mimics the machine's operation screen as a result of the simulation. The output unit 15 may further output an image or video of the operation screen, thereby obtaining a more accurate and easy-to-understand answer that matches the machine configuration and settings. Furthermore, by inputting this operation screen into the generation AI, an appropriate evaluation can be obtained as to whether the answer to the original question is satisfactory.
[0048] 10 is a second diagram illustrating the input and output of the generated AI in a modified example. Here, the displayed image is added to the prompt and an evaluation is output as a result of the simulation performed according to the operation procedure obtained from the generated AI in FIG. 9. However, since this evaluation is not a satisfactory answer to the original question, a new operation procedure is presented.
[0049] 11 is a diagram showing an example of an output from the operation procedure verification device 1 in the modified example. Here, for the new operation procedure shown in FIG. 10, the displayed image is added to the prompt of the generating AI as a result of the simulation, and the result is evaluated as "satisfactory result for the question."
[0050] The operation procedure obtained in the form of natural language is then combined with the operation screen obtained as a result of executing the simulation, and an evaluation is also output as an answer to the original question.
[0051] According to any of the above-described embodiments, the operation procedure verification device 1 can verify whether an operation procedure automatically generated by AI will cause an abnormal state in the machine by performing a simulation before applying the operation procedure to an actual machine. As a result, the reliability and safety of the obtained operation procedure are improved, and the user can apply the operation procedure to the actual machine with confidence.
[0052] Although the embodiments of the present invention have been described in detail above, the present invention is not limited to the individual embodiments described above. Various additions, substitutions, modifications, partial deletions, etc. are possible to these embodiments without departing from the gist of the invention or the idea and intent of the present invention derived from the content of the claims and their equivalents. For example, in the above-described embodiments, the order of each operation and the order of each process are shown as examples and are not limited to these. The same applies when numerical values or mathematical expressions are used in the description of the above-described embodiments.
[0053] The following supplementary notes are further disclosed regarding the above-described embodiment and modified examples. (Supplementary Note 1) An operation procedure verification device (1) comprising: a machine information acquisition unit (12) that acquires machine information including at least one of the configuration, settings, and state of a machine tool; an operation procedure acquisition unit (13) that acquires an operation procedure of the machine tool generated by AI; a validity confirmation unit (14) that executes a simulation to simulate the operation of the machine tool based on the machine information and the operation procedure and confirms the validity of the execution result of the simulation by comparing it with a pre-registered abnormal state; and an output unit (15) that outputs a confirmation result by the validity confirmation unit (14). (Supplementary Note 2) An operation procedure verification device (1) described in (Supplementary Note 1) comprising: a question input unit (11) that accepts input of a question regarding the operation of the machine tool; the operation procedure acquisition unit (12) acquires the operation procedure generated as a response by the generation AI from the question content and the machine information; and the output unit (15) outputs the operation procedure and the confirmation result by the validity confirmation unit (14). (Supplementary Note 3) The validity confirmation unit (14) inputs the confirmation result to the generation AI and asks for an evaluation of whether the confirmation result is satisfactory as an answer to the question, and the output unit (15) outputs the operation procedure, the confirmation result, and the evaluation. (Supplementary Note 4) The operation procedure verification device (1) described in (Supplementary Note 3) wherein the validity confirmation unit (14) repeats a process of asking the generation AI for another operation procedure to improve the evaluation. (Supplementary Note 5) The operation procedure verification device (1) described in any of (Supplementary Note 2) to (Supplementary Note 4) wherein the question input unit (11) accepts input of either text or an image, or a combination thereof, as the question. (Supplementary Note 6) The question input unit (11) asks the generation AI about items required as the machine information along with the content of the question, and the machine information acquisition unit (12) acquires the machine information corresponding to the items. The operation procedure verification device (1) described in any one of (Supplementary Note 2) to (Supplementary Note 5)(Supplementary Note 7) The operation procedure verification device (1) according to any one of (Supplementary Note 2) to (Supplementary Note 6), wherein the question input unit (11) requests an answer in a plurality of language formats from the generation AI, and the operation procedure acquisition unit (12) selects and provides one of the plurality of language formats to the validity confirmation unit (14) and the output unit (15), respectively. (Supplementary Note 8) The operation procedure verification device (1) according to any one of (Supplementary Note 1) to (Supplementary Note 7), wherein the validity confirmation unit (14) generates an image or video simulating an operation screen of the machine tool as a result of executing the simulation, and the output unit (15) further outputs the image or video.
[0054] REFERENCE SIGNS LIST 1 Operation procedure verification device 10 Control unit 11 Question input unit 12 Machine information acquisition unit 13 Operation procedure acquisition unit 14 Validity confirmation unit 15 Output unit 20 Storage unit
Claims
1. A machining procedure verification device comprising: a machine information acquisition unit that acquires machine information including at least any one of the configuration, settings, and status of a machine tool; an operation procedure acquisition unit that acquires the operation procedure of the machine tool generated by an AI; a validity confirmation unit that executes a simulation to simulate the operation of the machine tool based on the machine information and the operation procedure, and confirms the validity of the execution result of the simulation by comparing it with a pre-registered abnormal state; and an output unit that outputs the confirmation result by the validity confirmation unit.
2. The machining procedure verification device according to claim 1, further comprising a question input unit that receives an input of question content regarding the operation of the machine tool, wherein the operation procedure acquisition unit acquires the operation procedure generated by the AI as an answer from the question content and the machine information, and the output unit outputs the operation procedure and the confirmation result by the validity confirmation unit.
3. The machining procedure verification device according to claim 2, wherein the validity confirmation unit inputs the confirmation result to the generating AI and asks an evaluation as to whether the confirmation result can satisfy as an answer to the question content, and the output unit outputs the operation procedure, the confirmation result, and the evaluation.
4. The machining procedure verification device according to claim 3, wherein the validity confirmation unit repeats a process of asking the generating AI for another operation procedure for improving the evaluation.
5. The machining procedure verification device according to any one of claims 2 to 4, wherein the question input unit receives an input of either a character or an image, or a combination thereof, as the question content.
6. The machining procedure verification device according to any one of claims 2 to 5, wherein the question input unit asks the generating AI for items necessary as the machine information together with the question content, and the machine information acquisition unit acquires the machine information corresponding to the items.
7. The machining procedure verification device according to any one of claims 2 to 6, wherein the question input unit requests an answer in a plurality of language formats from the generating AI, and the operation procedure acquisition unit selects and provides any one of the plurality of language formats to the validity confirmation unit and the output unit, respectively.
8. The machining procedure verification device according to any one of claims 1 to 7, wherein the validity confirmation unit generates an image or a video simulating an operation screen of the machine tool as an execution result of the simulation, and the output unit further outputs the image or the video.
Citation Information
Patent Citations
Robot simulation device, robot simulation method, and robot simulation program
JP2015093345A
Machine tool for generating optimal acceleration and deceleration
JP2017068325A
Inverse and forward modeling machine learning-based generative design
US20220108185A1