Event analysis device, event analysis method, and analysis system

Abstract

Provided is an event analysis device capable of appropriately presenting information about various events. Therefore, a generation unit (2) in an event analysis device (60) generates an occurred event explanatory text (31) on the basis of: control structure information (11) indicating a control structure relating to a control method between a controlled object (72) and a control device (74) in a system (70) to be analyzed; event information (13) indicating an event occurred in the system (70) to be analyzed; and correspondence relationship information (12) including one or a plurality of pieces of correspondence information defining a correspondence relationship between a phrase included in the control structure information (11) and a phrase included in the event information (13).

Description

Event analysis device, event analysis method, and analysis system

[0001] The present invention relates to an event analysis device, an event analysis method, and an analysis system.

[0002] In the field of control systems, the scale and networking of controlled systems and the use of artificial intelligence (AI) technology are becoming increasingly common. However, when unexpected events such as malfunctions and breakdowns occur in large-scale, complex, and AI-enabled systems, the time and effort required for cause analysis and countermeasures increases, increasing the need for support using event resolution technology. For example, the abstract of Patent Document 1 states, "The system includes: a past fault information structuring unit that extracts multiple characteristic elements that represent a structure from a past fault information group and generates a hierarchically structured past fault information element group; a past fault information element organizing unit that compares the multiple elements with multiple keyword groups that are collections of keywords and generates multiple related element groups organized so as to assign the semantically closest keyword group; a past fault information database generation unit that adds reference information to each of the multiple related element groups to create multiple related reference groups; and a past fault information search unit that searches the multiple related reference groups for an input search target and outputs search results."

[0003] Japanese Patent Application Laid-Open No. 2023-116135

[0004] In Patent Document 1, the event analysis is limited to describing the fault information of software processing, and there are problems in applying it to control systems. It is difficult to perform analysis based on configuration information of controlled systems such as manufacturing line control systems and automated driving systems, control devices such as programmable logic controllers (PLCs) and electronic control units (ECUs), and information on their control operations. The present invention has been made in consideration of the above-mentioned circumstances, and aims to provide an event analysis device, an event analysis method, and an analysis system that can appropriately present information on various events.

[0005] In order to solve the above problems, the event analysis device of the present invention comprises: a generation unit that generates an occurring event description that explains an event that has occurred in an analyzed system that has a control object and a control device that controls the control object; and an improvement statement that proposes improvements to the analyzed system; and an output unit that outputs the occurring event description and the improvement statement generated by the generation unit, wherein the generation unit generates the occurring event description based on control structure information that indicates a control structure related to how the controlled object and the control device are controlled in the analyzed system, event information that represents an event that has occurred in the analyzed system, and correspondence information that includes one or more correspondence information that defines a correspondence between words included in the control structure information and words included in the event information.

[0006] According to the present invention, information about various events can be presented appropriately.

[0007] 1 is a block diagram showing an example of the configuration of an analysis system according to a first embodiment. FIG. 1 is a block diagram of a computer. FIG. 2 is a block diagram showing an example of a control structure. FIG. 3 is a diagram showing the contents of control structure information. FIG. 4 is a diagram showing the contents of event information. FIG. 5 is a diagram showing the contents of correspondence information. FIG. 6 is a diagram showing the contents of past event / explanation information. FIG. 7 is a diagram showing the contents of an occurred event explanation sentence. FIG. 8 is a diagram showing the contents of an improvement sentence. FIG. 9 is a diagram showing an example of an occurrence cause / control structure display screen displayed on the output unit. FIG. 10 is a diagram showing an example of an improvement proposal / control structure display screen displayed on the output unit. FIG. 11 is a diagram showing an example of a correspondence editing screen displayed on the output unit. FIG. 12 is a flowchart of a generated occurrence event explanation sentence. FIG. 13 is a flowchart of a generated cause / control structure data generation routine. FIG. 14 is a flowchart of an improvement sentence generation routine. FIG. 15 is a flowchart of an improvement proposal / control structure data generation routine.

[0008] 1 is a block diagram showing an example of the configuration of an analysis system 50 according to the first embodiment. The analysis system 50 includes an event analysis device 60 and an analysis target system 70. The event analysis device 60 (computer) is a device that explains events that occur in the analysis target system 70. Here, the analysis target system 70 includes a control target 72 and a control device 74.

[0009] 1 shows one control object 72 and one control device 74, but there may be multiple of these. The system under analysis 70 is, for example, an autonomous vehicle. The control object 72 is, for example, the steering wheel, brake pedal, accelerator pedal, etc. of this autonomous vehicle. The control device 74 is, for example, a vehicle controller, and controls the above-mentioned control object 72 to achieve autonomous driving.

[0010] The event analysis device 60 includes a memory unit 1, a generation unit 2 (generation process), an output unit 3 (output process), a large-scale language model unit 4, and a correspondence addition / change input unit 6. The memory unit 1 stores control structure information 11, correspondence information 12, event information 13, and past event / explanation information 14. The control structure information 11 is information describing the control structure (details of which will be described later) of the system 70 to be analyzed.

[0011] The event information 13 is a natural language description of an outline of an event that occurred in the system 70 under analysis. Here, "event" includes accidents, breakdowns, failures, malfunctions, etc. These may be called "accident events," "failure events," "fault events," or "malfunction events." The correspondence information 12 is information that defines the correspondence between the control structure information 11 and the event information 13. The past event / explanation information 14 is information such as the contents of past events, which are events that occurred in the past, explanations for the past events, and improvements to be made to the past events. The correspondence addition / change input unit 6 performs addition or change processing on the correspondence information 12 through processing described below. The output unit 3 displays various information to the user.

[0012] The generation unit 2 includes an occurrence event description generation unit 21 and an improvement statement generation unit 22. The occurrence event description generation unit 21 generates an occurrence event description 31 based on the information stored in the storage unit 1. Here, an "occurrence event" is, for example, a newly occurring event, and the occurrence event description 31 is a character string or the like that describes the cause of the occurrence event, etc.

[0013] The improvement statement generation unit 22 generates an improvement statement 32 based on the information stored in the storage unit 1 and the occurrence event explanation statement 31. The improvement statement 32 is a character string or the like that describes an improvement proposal for the control structure information 11. The generated occurrence event explanation statement 31 and improvement statement 32 are also written to the storage unit 1.

[0014] The large-scale language model unit 4 receives control structure information 11, correspondence information 12, event information 13, past event / explanation information 14, and instructions for generating an occurrence event description 31 or an improvement statement 32 from the generation unit 2, and executes the generation process for the occurrence event description 31 or the improvement statement 32 in accordance with the generation instructions.

[0015] Figure 2 is a block diagram of a computer 980. The event analysis device 60 shown in Figure 1 includes one or more computers 980 shown in Figure 2. In Figure 2, the computer 980 includes a CPU 981, a storage unit 982, a communication I / F (interface) 983, an input / output I / F 984, and a media I / F 985. Here, the storage unit 982 includes a RAM 982a, a ROM 982b, and an SSD (Solid State Drive) 982c.

[0016] The communication I / F 983 is connected to a communication circuit 986. The input / output I / F 984 is connected to an input / output device 987. The media I / F 985 reads and writes data from a recording medium 988. The ROM 982b stores an IPL (Initial Program Loader) and the like executed by the CPU. The SSD 982c stores application programs, various data, and the like. The CPU 981 realizes various functions by executing application programs and the like loaded from the SSD 982c into the RAM 982a.

[0017] The generation unit 2, large-scale language model unit 4, and correspondence addition / change input unit 6 shown in Fig. 1 are primarily blocks showing functions realized by application programs, etc. The storage unit 1 in Fig. 1 corresponds to the SSD 982c, and the output unit 3 corresponds to the input / output device 987.

[0018] <Data Structure of First Embodiment> Next, various data structures in the first embodiment will be described. FIG. 3 is a block diagram showing an example of a control structure 76. The control structure 76 has a graph structure, and the control content in the analysis target system 70 is represented by a plurality of nodes and one or more edges. In FIG. 3, the control structure 76 includes a plurality of nodes N1 to N7 (components) and edges E1 to E8 connecting these nodes. Node N1 corresponds to a human driver, node N2 corresponds to a steering wheel, node N3 corresponds to a brake pedal, node N4 corresponds to an AD controller (autonomous driving control device, AD controller), node N5 corresponds to a steering actuator, node N6 corresponds to a brake actuator, and node N7 corresponds to a vehicle.

[0019] Edge E1 starts from node N1 and ends at node N2. Hereinafter, the start and end nodes of each edge may be expressed in the format of "start node → end node". Edge E1 represents the driver's operation of the steering wheel (Handle steering wheel). Edge E2 from N1 to N3 represents the driver's depression of the brake pedal (Push brake pedal). Edge E3 from N2 to N4 represents the transmission of the steering wheel angle (Send steering wheel angle). Edge E4 from N4 to N5 represents the command to the steering actuator (Command steering actuator).

[0020] Edge E5 from N5 to N7 represents the actuation of the steering angle of the vehicle (Actuate Vehicle Steering Angle). Edge E6 from N4 to N6 represents a command to the brake actuator (Command brake actuator). Edge E7 from N6 to N7 represents the actuation of the brake system (Actuate brake system). Edge E8 from N7 to N4 represents sensor information (Sensor).

[0021] In this way, nodes N1 to N7 become "constituent parts" of the control structure 76. Furthermore, edges E1 to E8 and the operations within nodes N1 to N7 become "control operations" of these "constituent parts." Furthermore, in the example of FIG. 3, node N4 (AD controller) corresponds to the control device 74, but if multiple control devices 74 are provided, it is advisable to also include data transmitted and received between these control devices 74 in the edges.

[0022] 4 is a diagram showing the contents of the control structure information 111. The control structure information 111 is an example of the above-mentioned control structure information 11, and can be written in, for example, the PlantUML format, which is one type of structure information description. The control structure information 111 includes control structure node information 1111, control structure edge information 1112, and a control structure description section 1113. The control structure node information 1111 lists the node IDs and node names of nodes N1 to N7 in the above-mentioned control structure 76. The control structure edge information 1112 lists the edge IDs, edge names, start node IDs, and end node IDs of edges E1 to E8.

[0023] The control structure explanatory text section 1113 is text data that is an explanatory text of the control structure 76. Although part of the control structure explanatory text section 1113 is omitted in the figure, the full content is as follows: "The following is the draft model control loop structure of the target system. The Human Driver controls the steering wheel and brake to keep control vehicle dynamics. The steering wheel is operated by a human driver to control vehicle steering. The brake pedal is operated by a human driver to decrease vehicle speed. The AD controller is an Autonomous Driving controller that controls vehicle dynamics based on the sensor information of surrounding objects from the vehicle. The AD controller sends commands to the steering actuator and brake actuator in the Automatic Driving mode. In the manual driving mode, the AD controller sends commands to the steering wheel actuator based on the human-operated steering wheel inputs. The Steering Actuator controls vehicle steering based on the command from the AD controller. The Brake Actuator controls vehicle brake based on the controller command and human-operated pedal."The human-operated pedal can override the AD controller command. The vehicle drives automatically by the AD controller and observed by a human driver. The vehicle sends sensor information about surrounding objects. The following is an overview of the model control loop structure of the target system. A human driver operates the steering wheel and brake to control the vehicle dynamics. The steering wheel is operated by a human driver to control the steering of the vehicle. The brake pedal is operated by a human driver to reduce vehicle speed. The AD controller is an autonomous driving control device that controls the vehicle dynamics based on sensor information about objects around the vehicle. In autonomous driving mode, the AD controller sends commands to the steering actuator and brake actuator. In manual driving mode, it sends commands to the steering actuator based on human steering. The steering actuator controls the vehicle steering based on commands from the AD controller. The brake actuator controls the vehicle brakes based on commands from the controller and the human-operated pedal. The human-operated pedal can override the AD controller's commands. The vehicle is driven automatically by the AD controller and observed by a human driver. The vehicle sends sensor information about surrounding objects.

[0024] FIG. 5 is a diagram showing the contents of the event information 131. The event information 131 is an example of the event information 13 (see FIG. 1), and describes an event that occurred in the analysis target system 70 in normal text (character strings). The event information 131 describes the circumstances of a minor accident that occurred in the analysis target system 70, which is an autonomous vehicle, and its contents are as follows: "The autonomous vehicle lost control and came into contact with an obstacle (wall). The test driver attempted to steer to avoid the accident, but was unable to control the vehicle. The test driver tried to stop the vehicle by using the brake pedal, and the AD vehicle eventually came to a halt."

[0025] In this way, in this embodiment, the event information 13 can be written in natural language, which has the effect of making it possible to easily generate the event information 13 based on notes taken by field workers who have dealt with malfunctions, reports from users about breakdowns, and the like.

[0026] 6 is a diagram showing the contents of the correspondence information 121. The correspondence information 121 is an example of the correspondence information 12 (see FIG. 1), and its contents are as follows: "Please note that the output focuses on accident analysis using STAMP / CAST. What we want to know is the accident analysis results, not the system overview. Please note that in this analysis, the following nouns have the same meaning: {vehicle, autonomous vehicle} {human driver, test driver}"

[0027] The correspondence information 121 specifies that the term "vehicle" in the control structure information 111 and the term "autonomous vehicle (AD vehicle)" in the event information 131 have the same meaning. Similarly, the correspondence information 121 specifies that the term "human driver" in the control structure information 111 and the term "test driver" in the event information 131 have the same meaning. The correspondence information 121 also includes an instruction to use, if any, an analysis framework to be used in analyzing the event, for example, "accident analysis using STAMP / CAST."

[0028] As described above, according to this embodiment, the correspondence between the words in the control structure information 111 and the event information 131 can be defined by the correspondence information 121. The words in the control structure information 111 are set in advance at the design stage. According to this embodiment, by applying the correspondence information 121, it is not necessary to use the various words in the control structure information 111 exactly in the event information 131. This makes it possible to use reports from field workers and users as the event information 131, which has the effect of facilitating the input of the event information 131.

[0029] FIG. 7 is a diagram showing the contents of the past event / explanation information 141. The past event / explanation information 141 is an example of the past event / explanation information 14. The past event / explanation information 141 has multiple records (rows), and each record includes an event ID, past event data, past event explanation data, and improvement content data. The event ID is an identification number assigned to a past event. The past event data is a character string that explains the content of a past event, such as a malfunction or failure that occurred in an autonomous vehicle in the past. The past event explanation data is a character string that explains the cause of the past event. The improvement content data is a character string that explains the improvement content corresponding to the past event.

[0030] When outputting a generation instruction to the large-scale language model unit 4, the generation unit 2 also supplies past event / explanation information 14, such as past event / explanation information 141. This produces the effect of few-shot learning, making it possible to generate the occurred event explanation 31 and the improved sentence 32 while referring to past events, resulting in the effect of obtaining the occurred event explanation 31 and the improved sentence 32 with higher reliability.

[0031] FIG. 8 is a diagram showing the contents of the occurred event description 311. The occurred event description 311 is an example of the occurred event description 31 (see FIG. 1), and includes an event summary description section 3111, a non-safety control action description section 3112, and a cause candidate section 3113. The event summary description section 3111 describes an overview of the event. Although part of the event summary description section 3111 is omitted in the figure, the full content is as follows: "- Accident_description The autonomous vehicle lost control and moved towards an obstacle (wall). The human driver tried to control the steering wheel to avoid the accident but could not control the vehicle. The human driver then tried to stop the vehicle using the brake pedal, and the vehicle finally stopped."

[0032] The non-safety control action description section 3112 describes a non-safety control action that may be a direct cause of the occurrence of an event. Although part of the non-safety control action description section 3112 is omitted in the figure, the full content is as follows: - CAST_analysis - unsafe_control_actions - action AD controller failed to maintain proper vehicle control. - consequences Vehicle lost control and moved towards an obstacle. - action Human driver was unable to override the AD controller's steering control. - consequences Human driver could not avoid the accident by steering. - action Human driver had to use the brake pedal to stop the vehicle. - consequences The vehicle stopped, but not before getting dangerously close to the obstacle.

[0033] The cause candidate section 3113 describes cause candidates that may be factors that cause an unsafe control action to occur. In the illustrated example, the cause candidate section 3113 includes a cause candidate CN4 related to node N4 (AD controller), a cause candidate CN5 related to node N5 (steering actuator), and a cause candidate CN6 related to node N6 (brake actuator) in the control structure 76 (see FIG. 3).

[0034] The contents of the cause candidate CN4 are as follows: "- Cause: Malfunction of the AD controller or misinterpretation of sensor information. - Result: The AD controller was unable to maintain control of the vehicle. The human driver was unable to override the steering control of the AD controller."

[0035] The contents of the cause candidate CN5 are as follows: "-Cause: The steering actuator malfunctioned or did not respond to the human driver's input. -Result: The human driver was unable to override the AD controller's steering control."

[0036] The content of the cause candidate CN6 is as follows: "-Cause: The brake actuator reacted late to the human driver's input. -Result: The human driver had to use the brake pedal to stop the vehicle."

[0037] 9 is a diagram showing the contents of the improvement statement 321. The improvement statement 321 is an example of the improvement statement 32 (see FIG. 1), and describes a proposal to improve the cause candidate section 3113 (see FIG. 8) in the event description 31. In the illustrated example, the improvement statement 321 includes an improvement proposal PN4 related to the node N4 (AD controller) in the control structure 76 (see FIG. 3), an improvement proposal PN1 related to the node N1 (driver), particularly the override function, an improvement proposal PN6 related to the node N6 (brake actuator), and improvement proposals PA1 and PA2 related to the entire control structure 76.

[0038] The contents of improvement proposal PN4 are as follows: "1. Improving the AD controller's functionality: The failure of the AD controller to maintain proper vehicle control was a significant factor in the accident. It is recommended that the AD controller's functionality be enhanced, including its ability to correctly interpret sensor information. This could include improving the algorithms used by the AD controller and upgrading the sensors used in the vehicle."

[0039] The contents of improvement proposal PN1 are as follows: "2. Strengthening the override function: The accident could not be avoided because the human driver was unable to override the steering control of the AD controller. We recommend improving the system's override function so that the human driver can take control in an emergency. This could include introducing a more responsive override system or providing additional training to drivers on how to effectively use the override system."

[0040] The contents of improvement proposal PN6 are as follows: "3. Improving the responsiveness of the brake actuator: The delay in the reaction of the brake actuator to the input of the human driver was also a factor in the accident. It is recommended that the response time of the brake actuator be improved. This could be done by upgrading the brake actuator or by introducing a system that can apply the brakes immediately in an emergency."

[0041] The content of improvement proposal PA1 is as follows: "4. Regular inspection and maintenance of systems: Regular inspection and maintenance of vehicle control systems (including AD controllers, steering actuators, and brake actuators) will help identify and correct potential problems before they lead to an accident."

[0042] The contents of improvement proposal PA2 are as follows: "5. Strengthening safety procedures: Strengthen safety protocols and emergency procedures so that drivers and autonomous driving systems can respond effectively to unforeseen circumstances. This may include additional driver training, improving communication systems, and better integration of autonomous control systems and manual control systems."

[0043] <Display Screen> Next, various display screens in the first embodiment will be described. Fig. 10 is a diagram showing an example of a cause of occurrence / control structure display screen 80 displayed on the output unit 3. The cause of occurrence / control structure display screen 80 includes a control structure display field 82. The control structure display field 82 displays the contents of the control structure 76 (see Fig. 3). Therefore, the control structure display field 82 displays node images corresponding to the nodes N1 to N7 and edge images corresponding to the edges E1 to E8 shown in Fig. 3. The node images and edge images are assigned the same reference numerals as the nodes N1 to N7 and edges E1 to E8.

[0044] When the user performs a predetermined operation on any of the node images, the cause candidate corresponding to that node image is displayed on the occurrence cause / control structure display screen 80. In the example of Fig. 10, cause candidates CN4, CN5, and CN6 corresponding to node N4 (AD controller), node N5 (steering actuator), and node N6 (brake actuator), respectively, are displayed. The contents of these cause candidates CN4, CN5, and CN6 are as shown in the cause candidate section 3113 in Fig. 8.

[0045] In this way, on the cause / control structure display screen 80, the relationship between the nodes N1 to N7 and the edges E1 to E8 can be diagrammed as a directed graph in the control structure display field 82. Furthermore, the cause candidates described in the cause candidate section 3113 (see FIG. 8 ) can be displayed in correspondence with the nodes or edges described in the cause candidates. In this way, the control structure 76 is described using nodes and edges, and the control structure 76 can be diagrammed and superimposed with the cause candidates that are potential causes of the occurrence of an event. This allows the user to intuitively understand in which part of the control structure 76 a problem may be occurring, making analysis by the user easier.

[0046] 11 is a diagram showing an example of an improvement proposal / control structure display screen 90 displayed on the output unit 3. The improvement proposal / control structure display screen 90 includes a control structure display field 92. The contents of this control structure display field 92 are the same as those of the control structure display field 82 on the cause / control structure display screen 80 (see FIG. 10).

[0047] When the user performs a predetermined operation on any of the node images in the control structure display field 92, an improvement proposal corresponding to that node image is displayed on the improvement proposal / control structure display screen 90. In the example of Fig. 11, improvement proposals PN4, PN5, and PN6 corresponding to node N4 (AD controller), node N5 (steering actuator), and node N6 (brake actuator), respectively, are displayed. The contents of these improvement proposals PN4, PN5, and PN6 are as shown in Fig. 9.

[0048] In this way, on the improvement proposal / control structure display screen 90, similar to the cause / control structure display screen 80, the relationship between nodes N1 to N7 and edges E1 to E8 can be diagrammed as a directed graph in the control structure display field 92. Furthermore, the improvement proposals described in the improvement statements 321 (see FIG. 9 ) can be displayed in correspondence with the nodes or edges described in the improvement proposals. In this way, the control structure 76 is described using nodes and edges, and the control structure 76 can be diagrammed and the improvement proposals corresponding to the events can be superimposed and displayed. This allows the user to intuitively understand which parts of the control structure 76 need improvement or modification, thereby facilitating the user's repair and improvement of the control structure 76.

[0049] 12 is a diagram showing an example of a correspondence editing screen 150 displayed on the output unit 3. This correspondence editing screen 150 is displayed on the output unit 3 by the correspondence addition / change input unit 6. The correspondence editing screen 150 displays one or more records (rows), and each record includes a correspondence ID, a control structure phrase, and an event phrase. The correspondence ID is an identification number assigned to each record. The control structure phrase is a character string of any of the phrases included in the control structure explanation section 1113 (see FIG. 4). The event phrase is a phrase included in the event information 131 (see FIG. 5) that corresponds to (has the same meaning as) a control structure phrase in the same record.

[0050] That is, each record on the correspondence editing screen 150 becomes "correspondence information" that indicates the correspondence between a control structure phrase and an event phrase. The user can edit the correspondence between a control structure phrase and an event phrase by editing the control structure phrase or the event phrase of each record on the correspondence editing screen 150. The user can also define a new correspondence by adding a control structure phrase and an event phrase to a blank record with a correspondence ID of "3," for example.

[0051] In this embodiment, after the user checks the output occurrence event description 311 (see FIG. 8) and improvement statement 321 (see FIG. 9), the user can change or add word correspondences and regenerate the occurrence event description 311 and improvement statement 321. As a result, this embodiment has the effect of enabling the generation of an occurrence event description 311 and improvement statement 321 with even higher reliability.

[0052] <Operation of First Embodiment> Next, the operation of the first embodiment will be described. FIG. 13 is a flowchart of a routine for generating an occurred event explanation. This routine is executed by the occurred event explanation generation unit 21 in response to a predetermined instruction from the user. When the process proceeds to step S12 in FIG. 13, the occurred event explanation generation unit 21 reads the control structure information 11 from the storage unit 1. Next, when the process proceeds to step S14, the occurred event explanation generation unit 21 reads the event information 13 from the storage unit 1. Next, when the process proceeds to step S16, the occurred event explanation generation unit 21 reads the correspondence relationship information 12 from the storage unit 1. Next, when the process proceeds to step S18, the occurred event explanation generation unit 21 reads the past event / explanation information 14 from the storage unit 1.

[0053] Next, when the process proceeds to step S20, the occurred event explanation generation unit 21 instructs the large-scale language model unit 4 to generate the occurred event explanation 31. That is, the occurred event explanation generation unit 21 supplies the large-scale language model unit 4 with the information acquired in steps S12 to S18 and an instruction to generate the occurred event explanation 31. As a result, the large-scale language model unit 4 generates the occurred event explanation 31 as exemplified as the occurred event explanation 311 in FIG. 8. Then, the occurred event explanation generation unit 21 writes the generated occurred event explanation 31 to the storage unit 1. This completes the process of this routine.

[0054] As described above, according to this embodiment, the occurred event explanation generator 21 can generate the occurred event explanation 31 using the control structure information 11, the event information 13, and the correspondence relationship information 12. This makes it possible to generate the occurred event explanation 31 based on the control structure 76 (see FIG. 3 ) of the system 70 under analysis, which has the effect of improving the reliability of the occurred event explanation 31.

[0055] FIG. 14 is a flowchart of the routine for generating cause and control structure data. The "cause and control structure data" is data that specifies the contents of the cause and control structure display screen 80 (see FIG. 10). This routine is executed by the event description generator 21 in response to a predetermined instruction from the user after the event description 31 has been generated. When the process proceeds to step S32 in FIG. 14, the event description generator 21 reads the control structure information 11 from the storage unit 1. Next, when the process proceeds to step S34, the event description generator 21 reads the event description 31 from the storage unit 1.

[0056] Next, when the process proceeds to step S36, the event explanation generation unit 21 generates cause and control structure data by analyzing the correspondence between the explanation written in the event explanation 31 and the control structure information 11. The event explanation generation unit 21 then writes the generated cause and control structure data to the storage unit 1. This cause and control structure data can be used to display a cause and control structure display screen 80 such as the one shown in FIG. 10 on the output unit 3. This completes the process of this routine.

[0057] 15 is a flowchart of the improvement sentence generation routine. This routine is executed by the improvement sentence generation unit 22 in response to a predetermined instruction from the user after the occurred event explanation 31 has been generated. When the process proceeds to step S42 in FIG. 15, the improvement sentence generation unit 22 reads the control structure information 11 from the storage unit 1. Next, when the process proceeds to step S44, the improvement sentence generation unit 22 reads the occurred event explanation 31 from the storage unit 1. Next, when the process proceeds to step S46, the improvement sentence generation unit 22 reads the past event / explanation information 14 from the storage unit 1.

[0058] Next, when the process proceeds to step S48, the improved sentence generation unit 22 instructs the large-scale language model unit 4 to generate an improved sentence 32. That is, the improved sentence generation unit 22 supplies the large-scale language model unit 4 with the information acquired in steps S42 to S46 and an instruction to generate the improved sentence 32. As a result, the large-scale language model unit 4 generates an improved sentence 32 such as the improved sentence 321 exemplified in FIG. 9. Then, the improved sentence generation unit 22 writes the generated improved sentence 32 to the storage unit 1. This completes the process of this routine.

[0059] FIG. 16 is a flowchart of the improvement proposal / control structure data generation routine. The "improvement proposal / control structure data" is data that specifies the content of the improvement proposal / control structure display screen 90 (see FIG. 11). This routine is executed by the improvement statement generation unit 22 in response to a predetermined instruction from the user after the improvement statement 32 has been generated. When the process proceeds to step S52 in FIG. 16, the improvement statement generation unit 22 reads the control structure information 11 from the storage unit 1. Next, when the process proceeds to step S54, the improvement statement generation unit 22 reads the improvement statement 32 from the storage unit 1.

[0060] Next, when the process proceeds to step S56, the improvement statement generation unit 22 generates improvement proposal / control structure data by analyzing the correspondence between the improvement proposal described in the improvement statement 32 and the control structure information 11. Then, the improvement statement generation unit 22 writes the generated improvement proposal / control structure data into the storage unit 1. This improvement proposal / control structure data can be used to display an improvement proposal / control structure display screen 90 such as that shown in FIG. 11 on the output unit 3. This completes the process of this routine.

[0061] Effect of the embodiment As described above, according to the embodiment, the generation unit 2 generates the occurred event description 31 based on the control structure information 11 indicating the control structure 76 related to how the control target 72 and the control device 74 in the analysis target system 70 are controlled, the event information 13 indicating an event that has occurred in the analysis target system 70, and the correspondence relationship information 12 including one or more pieces of correspondence information that define the correspondence between the words and phrases included in the control structure information 11 and the words and phrases included in the event information 13. This allows the generation unit 2 to generate an appropriate occurred event description 31.

[0062] Furthermore, the control structure 76 specifies a plurality of components (N1 to N7) and the control actions in each of the components (N1 to N7), and the occurring event explanation 31 specifies the control action that caused the event to occur and the components (N1 to N7) that correspond to the control action, and it is more preferable that the generating unit 2 generates the occurring event explanation 31 based on the control structure information 11 and the event information 13. This allows the generating unit 2 to generate a more appropriate occurring event explanation 31 using the control structure information 11 and the event information 13.

[0063] It is more preferable that the generator 2 generates the improvement sentence 32 based on the occurred event description sentence 31. This allows appropriate improvement sentences 32 to be generated based on the occurred event description sentence 31.

[0064] Furthermore, it is more preferable that the events include failure events that are failures that occur in the system 70 under analysis, and that the correspondence information 12 is information describing the correspondence between words contained in the control structure information 11 and words having the same meaning contained in the event information 13. This makes it possible to use reports from field workers and users as the event information 13, making it easier to input the event information 13.

[0065] Furthermore, it is more preferable that the control structure information 11 has a graph structure including a plurality of nodes N1 to N7, each including the control target 72 and the control device 74, and a plurality of edges E1 to E8, each including a physical quantity measured by the control target 72 and supplied to the control device 74, a command supplied by the control device 74 to the control target 72, or data transmitted and received between the control devices 74. This makes it easier for the user to visually grasp the control structure information 11.

[0066] It is even more preferable to further provide a correspondence addition / change input unit 6 for adding new correspondence information to the correspondence information 12 or changing existing correspondence information. This allows appropriate correspondence information 12 to be constructed depending on the content of the event information 13.

[0067] Furthermore, it is more preferable that the generation unit 2 generates the improvement sentence 32 using the past event / explanation information 14 including past event data explaining the content of past events, which are events that occurred in the past, past event explanation data explaining the cause of the past event, and improvement content data explaining the content of improvements corresponding to the past event. This makes it possible to generate a more appropriate improvement sentence 32 based on the content of the past event / explanation information 14.

[0068] [Modifications] The present invention is not limited to the above-described embodiment, and various modifications are possible. The above-described embodiment is an example for explaining the present invention in an easy-to-understand manner, and is not necessarily limited to an embodiment having all of the described configurations. Furthermore, other configurations may be added to the configurations of the above-described embodiment, and some of the configurations may be replaced with other configurations. Furthermore, the control lines and information lines shown in the figures are those considered necessary for explanation, and do not necessarily represent all control lines and information lines necessary in the product. In reality, it can be assumed that almost all configurations are interconnected. Possible modifications of the above-described embodiment include, for example, the following.

[0069] (1) In the above embodiment, an example in which an autonomous vehicle is used as the system under analysis 70 has been described, but the system under analysis 70 is not limited to this. For example, the present invention can be applied to the field of maintenance of various types of equipment and can be applied to analyzing the causes of failures in various types of equipment. Furthermore, the present invention can be applied to the field of manufacturing various types of equipment and can be applied to analyzing the causes of defects in manufactured equipment.

[0070] (2) Since the hardware of the event analysis device 60 in the above embodiment can be realized by a general computer, the processes corresponding to the above-mentioned block diagrams and flowcharts, as well as programs for executing the various processes described above, may be stored on a storage medium (a computer-readable storage medium on which a program is recorded) or distributed via a transmission path.

[0071] (3) In the above embodiment, the processes corresponding to the block diagrams and flowcharts, as well as the various other processes described above, are described as software processes using a program. However, some or all of these processes may be replaced with hardware processes using an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array), etc.

[0072] (4) The various processes performed in the above embodiment may be executed by a server computer via a network not shown, and the various data stored in the memory unit 1 in the above embodiment may also be stored in the server computer.

[0073] 2 Generation unit (generation process) 3 Output unit (output process) 6 Correspondence addition / change input unit 11 Control structure information 12 Correspondence information 13 Event information 14 Past event / explanatory information 31 Occurring event explanation 32 Improvement statement 50 Analysis system 60 Event analysis device (computer) 70 System to be analyzed 72 Control object 74 Control device 76 Control structure E1 to E8 Edges N1 to N7 Nodes (component parts)

Claims

1. An event analysis device comprising: a generation unit that generates an occurrence event description that explains an event that has occurred in an analysis target system that has a control object and a control device that controls the control object, and an improvement statement that proposes improvements to the analysis target system; and an output unit that outputs the occurrence event description and the improvement statement generated by the generation unit, wherein the generation unit generates the occurrence event description based on control structure information that indicates a control structure related to how the control object and the control device are controlled in the analysis target system, event information that describes an event that has occurred in the analysis target system, and correspondence information that includes one or more pieces of correspondence information that define correspondence relationships between words included in the control structure information and words included in the event information.

2. An event analysis device according to claim 1, wherein the control structure identifies a plurality of components and a control operation in each of the components, and the description of the event that occurred identifies the control operation that caused the event to occur and the component that corresponds to the control operation.

3. An event analysis device according to claim 2, wherein the generation unit generates the improvement sentence based on the explanation of the occurred event.

4. An event analysis device according to claim 1, wherein the events include failure events that are failures that occur in the system under analysis, and the correspondence information is information that describes the correspondence between words contained in the control structure information and words contained in the event information that have the same meaning.

5. An event analysis device according to claim 1, characterized in that the control structure information has a graph structure comprising a plurality of nodes including the controlled object and the control device, and a plurality of edges including physical quantities measured by the controlled object and supplied to the control device, commands supplied by the control device to the controlled object, or data transmitted and received between the control devices.

6. An event analysis device according to claim 1, further comprising a correspondence addition / change input unit for adding new correspondence information to the correspondence information or changing existing correspondence information.

7. An event analysis device according to claim 1, wherein the generation unit generates the improvement sentence using past event / explanation information including past event data that explains the content of a past event, which is the event that occurred in the past, past event explanation data that explains the cause of the past event, and improvement content data that explains the content of an improvement corresponding to the past event.

8. An event analysis method that causes a computer to execute the following steps: a generation process that generates an occurrence event description that explains an event that has occurred in an analysis target system that has a control object and a control device that controls the control object, and an improvement statement that proposes improvements to the analysis target system; and an output process that outputs the occurrence event description and the improvement statement generated by the generation process, wherein the generation process generates the occurrence event description based on control structure information that indicates a control structure related to how the control object and the control device are controlled in the analysis target system, event information that describes an event that has occurred in the analysis target system, and correspondence information that includes one or more correspondence information that defines the correspondence between words included in the control structure information and words included in the event information.

9. An analysis system comprising an analysis target system and an event analysis device, wherein the analysis target system comprises a control target and a control device that controls the control target, and the event analysis device comprises: a generation unit that generates an occurred event description that explains an event that has occurred in the analysis target system and an improvement statement that proposes improvements to the analysis target system; and an output unit that outputs the occurred event description and the improvement statement generated by the generation unit, wherein the generation unit generates the occurred event description based on control structure information that indicates a control structure related to how the control target and the control device in the analysis target system are controlled, event information that indicates an event that has occurred in the analysis target system, and correspondence information that includes one or more correspondence information that defines a correspondence between words included in the control structure information and words included in the event information.

Images