Robot fault self-diagnosis and maintenance decision method and system based on large model

Abstract

This invention provides a robot fault self-diagnosis and maintenance decision-making method and system based on a large model, belonging to the field of robot fault diagnosis technology. The robot fault self-diagnosis and maintenance decision-making method includes: collecting and acquiring historical fault data of the robot; preprocessing the historical fault data and constructing a fault knowledge graph based on the preprocessed historical fault data; acquiring real-time early warning data of the robot; acquiring the real-time correlation values ​​of the robot's early warning nodes and multiple associated nodes based on the real-time early warning data and the fault knowledge graph; and using a method that determines the fault confidence of the early warning node based on multiple associated nodes, which can verify and corroborate whether the early warning node has failed, effectively improving the accuracy and effect of robot fault diagnosis, thereby avoiding misdiagnosis and reducing maintenance costs.

Description

Technical Field

[0001] This invention relates to the field of robot fault diagnosis technology, and more specifically to a robot fault self-diagnosis and maintenance decision-making method and system based on a large model. Background Technology

[0002] A robot is an intelligent machine capable of semi-autonomous or fully autonomous operation. Through programming and automatic control, robots can perform tasks such as manual labor or movement. With the continuous development of robotics technology, robots are increasingly being applied to various industries to assist in performing different tasks. For example, robots in the power grid sector can be used to de-ice cables, and robots in the medical field can assist in high-precision surgical procedures. Therefore, fault diagnosis and maintenance of robots are crucial for improving their operational safety and reliability.

[0003] In existing technologies, diagnosing robot malfunctions generally relies on real-time monitoring of the robot's own sensor parameters. These parameters, such as current, torque, temperature, and distance, are compared to corresponding threshold ranges. If a real-time sensor parameter exceeds the threshold range, the robot is considered faulty. However, the parameter values ​​of a single sensor at a given moment may exhibit normal fluctuations, potentially leading to misdiagnosis and resulting in low diagnostic accuracy and increased maintenance costs.

[0004] In the process of realizing this invention, the inventors of this application discovered that the above-mentioned solutions in the prior art have the defects of low accuracy in robot fault diagnosis and high maintenance costs. Summary of the Invention

[0005] The purpose of this invention is to provide a robot fault self-diagnosis and maintenance decision-making method and system based on a large model. This robot fault self-diagnosis and maintenance decision-making method and system based on a large model has the function of improving the accuracy of robot fault diagnosis and reducing maintenance costs.

[0006] To achieve the above objectives, embodiments of the present invention provide a robot fault self-diagnosis and maintenance decision-making method based on a large model, including:

[0007] Collect and acquire historical fault data of the robot;

[0008] The historical fault data is preprocessed, and a fault knowledge graph is constructed based on the preprocessed historical fault data.

[0009] Obtain real-time early warning data for the robot currently described;

[0010] Based on the real-time early warning data and the fault knowledge graph, the robot's early warning nodes and the real-time correlation values ​​of multiple associated nodes are obtained.

[0011] The fault confidence of the early warning node is determined based on multiple real-time correlation values;

[0012] The real-time fault status of the robot at the early warning node is determined based on the fault confidence level.

[0013] Based on the real-time fault situation, obtain the corresponding maintenance plan and execute the maintenance plan.

[0014] Optionally, preprocessing the historical fault data includes data cleaning.

[0015] Optionally, constructing a fault knowledge graph based on the preprocessed historical fault data includes:

[0016] Define entity types, relationship types, and entity attributes;

[0017] A pre-trained large language model is used to extract and align entities from the historical fault data.

[0018] A pre-trained large language model is used to extract and link relationships in the historical fault data.

[0019] Optionally, obtaining the real-time early warning data of the robot includes:

[0020] Obtain multiple time-series data of the robot within a preset time period;

[0021] According to formula (1), obtain multiple real-time parameters of the robot.

[0022] (1)

[0023] in, For the robot's first One real-time parameter, For the first preset quantity, For the first Peak and valley data for each real-time parameter within a corresponding preset time period. for Each real-time parameter precedes the peak and valley data for the corresponding preset time period. One data point, for Each real-time parameter is based on peak and valley data for a corresponding preset time period. One data point, , Numbered by integer;

[0024] Determine whether each of the real-time parameters is within the corresponding preset threshold range;

[0025] If the real-time parameters are not within the corresponding preset threshold range, the robot is suspected of malfunctioning, and an early warning is issued.

[0026] Obtain the real-time warning data corresponding to the real-time parameters.

[0027] Optionally, obtaining the robot's early warning nodes and the real-time association values ​​of multiple associated nodes based on the real-time early warning data and the fault knowledge graph includes:

[0028] Obtain the early warning nodes from the real-time early warning data;

[0029] The warning node is input into the fault knowledge graph for querying to obtain multiple associated nodes of the warning node;

[0030] Obtain the real-time association value for each of the associated nodes.

[0031] Optionally, obtaining the real-time association value of each of the associated nodes includes:

[0032] Obtain multiple time-series data of each associated node within a preset time period;

[0033] The real-time association parameters of each associated node are obtained according to formula (2).

[0034] (2)

[0035] in, For the first Real-time association parameters of the associated nodes, For the second preset quantity, For the first Peak and valley data of the associated nodes in the corresponding preset time period. For the first The associated nodes are in front of the peak and valley data for the corresponding preset time period. One data point, For the first After the peak and valley data of the associated nodes in the corresponding preset time period One data point, , Numbered by integer;

[0036] The real-time association value of each associated node is obtained according to formula (3).

[0037] (3)

[0038] in, For the first The real-time association value of each of the associated nodes. For the first The preset threshold range corresponding to each of the associated nodes.

[0039] Optionally, determining the fault confidence of the early warning node based on multiple real-time correlation values ​​includes:

[0040] Acquire the time-series data of the early warning node and multiple associated nodes when the early warning node fails;

[0041] The sensitivity coefficient between each associated node and the early warning node is obtained according to formula (4).

[0042] (4)

[0043] in, For the first The sensitivity coefficient between the associated nodes and the early warning nodes. When the warning node fails, the first The maximum parameter value of the associated nodes. When the warning node fails, the first The minimum parameter value of the associated nodes, This refers to the maximum parameter value of the early warning node when it fails. This refers to the minimum parameter value of the early warning node when a failure occurs.

[0044] The sensitivity coefficients of the multiple associated nodes and the early warning node are normalized;

[0045] The fault confidence level of the early warning node is obtained according to formula (5).

[0046] (5)

[0047] in, The fault confidence level of the early warning node. For the first The normalized value of the sensitivity coefficient between the associated node and the early warning node. The number of associated nodes.

[0048] Optionally, determining the current real-time fault status of the robot at the early warning node based on the fault confidence level includes:

[0049] Determine whether the fault confidence level is greater than or equal to the confidence threshold;

[0050] If the confidence level of the fault is greater than or equal to the confidence threshold, it is determined that the robot has a fault.

[0051] The fault status is obtained based on the real-time early warning data;

[0052] If the confidence level of the fault is less than the confidence threshold, the robot is determined to be fault-free.

[0053] On the other hand, the present invention also provides a robot fault self-diagnosis and maintenance decision-making system based on a large model, comprising:

[0054] The knowledge graph construction module is used to collect historical fault data and build a fault knowledge graph.

[0055] The controller, connected to the knowledge graph construction module, is used to execute any of the robot fault self-diagnosis and maintenance decision-making methods described above.

[0056] In another aspect, the present invention also provides a computer-readable storage medium storing instructions for being read by a machine to cause the machine to execute any of the robot fault self-diagnosis and maintenance decision-making methods described above.

[0057] Through the above technical solution, the robot fault self-diagnosis and maintenance decision-making method and system based on a large model provided by this invention collects and preprocesses historical fault data of the robot, enabling the construction of a corresponding fault knowledge graph based on a pre-trained large language model. Simultaneously, it collects real-time early warning data of the robot, locates early warning nodes and associated nodes in the fault knowledge graph based on the early warning data, and obtains the fault confidence of the early warning node based on the real-time association value of the associated nodes. Then, it determines the real-time fault status of the early warning node based on the fault confidence, generates a corresponding maintenance plan, and executes it. By using multiple associated nodes to determine the fault confidence of the early warning node, it can verify and corroborate whether the early warning node has failed, effectively improving the accuracy and effectiveness of robot fault diagnosis, thereby avoiding misdiagnosis and reducing maintenance costs.

[0058] Other features and advantages of the embodiments of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0059] The accompanying drawings are provided to further illustrate embodiments of the present invention and form part of the specification. They are used together with the following detailed description to explain the embodiments of the present invention, but do not constitute a limitation thereof. In the drawings:

[0060] Figure 1This is a flowchart of a robot fault self-diagnosis and maintenance decision-making method based on a large model according to an embodiment of the present invention;

[0061] Figure 2 This is a flowchart of constructing a fault knowledge graph in a robot fault self-diagnosis and maintenance decision-making method based on a large model according to an embodiment of the present invention;

[0062] Figure 3 This is a flowchart of obtaining early warning data in a robot fault self-diagnosis and maintenance decision-making method based on a large model according to an embodiment of the present invention;

[0063] Figure 4 This is a flowchart of obtaining associated nodes in a robot fault self-diagnosis and maintenance decision-making method based on a large model according to an embodiment of the present invention;

[0064] Figure 5 This is a flowchart of obtaining real-time correlation values ​​in a robot fault self-diagnosis and maintenance decision-making method based on a large model according to an embodiment of the present invention;

[0065] Figure 6 This is a flowchart of obtaining fault confidence in a robot fault self-diagnosis and maintenance decision-making method based on a large model according to an embodiment of the present invention;

[0066] Figure 7 This is a flowchart illustrating the process of obtaining fault information in a robot fault self-diagnosis and maintenance decision-making method based on a large model according to an embodiment of the present invention. Detailed Implementation

[0067] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the scope of the present invention.

[0068] It should be noted that the acquisition, transmission, storage, use, and processing of data in the technical solution of this application all comply with relevant laws and regulations. In the embodiments of this application, certain existing industry solutions such as software, components, and models may be mentioned. These should be considered exemplary, intended only to illustrate the feasibility of implementing the technical solution of this application, and do not imply that the applicant has already used or necessarily used such solutions.

[0069] Figure 1 This is a flowchart of a robot fault self-diagnosis and maintenance decision-making method based on a large model according to an embodiment of the present invention. Figure 1 The robot's fault self-diagnosis and maintenance decision-making method may include:

[0070] In step S1, historical fault data of the robot is collected and acquired. This historical fault data may include sources such as controller real-time logs, maintenance record texts, warranty records, and sensor historical databases.

[0071] In step S2, historical fault data is preprocessed, and a fault knowledge graph is constructed based on the preprocessed historical fault data. Preprocessing of historical fault data may include data cleaning and formatting, such as removing duplicate records and null values, and standardizing the time format. Specifically, the entities, relationships, and attributes of the knowledge graph are then defined, allowing the construction of the fault knowledge graph based on the historical fault data. Furthermore, a large language model pre-trained in the robotics domain can be used for entity extraction and relationship extraction from the historical fault data, such as using a BERT pre-trained model.

[0072] In step S3, real-time warning data of the current robot is acquired. This real-time warning data refers to the robot's fault warning triggered when a certain real-time parameter value exceeds a corresponding threshold range. Furthermore, the real-time warning data for this fault warning can be collected, including the warning component / location / node, warning parameters, etc.

[0073] In step S4, the robot's warning nodes and real-time association values ​​of multiple associated nodes are obtained based on real-time warning data and a fault knowledge graph. Specifically, the real-time warning data is input into the fault knowledge graph for querying / mapping, which yields the corresponding warning nodes and multiple associated / adjacent nodes. For example, for warning data indicating a joint x-tracking error of 0.5mm, the corresponding warning node is "Positioning Error Exceeds Limit," and its associated nodes may include reducer wear, encoder noise, etc. Furthermore, each associated node may include corresponding association parameters, such as housing temperature for reducer wear and encoder signal-to-noise ratio for encoder noise. Based on the association parameters of different associated nodes, the corresponding real-time association values ​​can be obtained.

[0074] In step S5, the fault confidence level of the early warning node is determined based on multiple real-time correlation values. Specifically, after obtaining the multiple real-time correlation values ​​for the early warning node, the fault confidence level of that early warning node can be calculated by comprehensively analyzing them.

[0075] In step S6, the real-time fault status of the robot at the warning node is determined based on the fault confidence level. A higher fault confidence level indicates a higher probability of a fault at the warning node, while a lower level indicates a lower probability. This means that the fault status of the warning node is verified by linking other nodes to accurately diagnose the robot's faults.

[0076] In step S7, the corresponding maintenance plan is obtained based on the real-time fault situation and then executed. Specifically, after identifying the fault, a large-scale model based on a fault knowledge graph can be used to analyze the fault to determine its root cause and formulate corresponding maintenance plans, such as replacing components or adjusting control parameters.

[0077] In steps S1 to S7, historical fault data of the robot is first acquired and preprocessed. Then, a large language model pre-trained in the robotics domain is used to extract entities and relationships from the historical fault data to construct a fault knowledge graph. Real-time early warning data of the robot is collected and input into the fault knowledge graph to obtain the corresponding early warning node and its associated / adjacent nodes. Simultaneously, the real-time parameter values ​​of the associated / adjacent nodes are acquired and converted into real-time association values. Based on multiple real-time association values, the fault confidence of the early warning node can be obtained. Finally, the fault status of the early warning node is verified and determined based on the fault confidence, and a corresponding maintenance plan is generated and executed based on the fault status.

[0078] Traditional robot fault diagnosis typically relies on real-time monitoring of the robot's sensor parameters. These parameters, such as current, torque, temperature, and distance, are compared to corresponding threshold ranges. If a parameter exceeds the threshold, the robot is considered faulty. However, individual sensor values ​​at any given moment may fluctuate, leading to misdiagnosis, low diagnostic accuracy, and increased maintenance costs. This embodiment of the invention uses a method that determines the fault confidence of a warning node based on multiple associated nodes. This verifies whether a warning node has malfunctioned, effectively improving the accuracy and effectiveness of robot fault diagnosis, thus avoiding misdiagnosis and reducing maintenance costs.

[0079] In this embodiment of the invention, after obtaining the preprocessed historical fault data, a fault knowledge graph can be constructed based on the historical fault data. Specifically, the construction steps can be as follows: Figure 2 As shown. Specifically, in Figure 2 Furthermore, the robot's fault self-diagnosis and maintenance decision-making method may also include:

[0080] In step S20, entity types, relation types, and entity attributes are defined. Entity types can include fault phenomena, fault modes, components, and parameter indicators. Fault phenomena can include abnormal joint vibration, excessive positioning error, etc. Fault modes can include reducer wear, encoder signal noise, etc. Components can include reducers, encoders, etc. Parameter indicators can include temperature, amplitude, etc. Relationship types can include "caused," "cause," etc. Entity attributes, such as fault phenomena, can include frequency of occurrence, duration, etc.

[0081] In step S21, a pre-trained large language model is used to extract and align entities from historical fault data. Specifically, a large language model pre-trained in the robotics field can be used to extract entities from structured or unstructured text and build a thesaurus for entity alignment and deduplication. For example, if wear of the reducer causes increased joint x-vibration, the fault mode is reducer wear, and the fault symptom is increased joint x-vibration.

[0082] In step S22, a pre-trained large language model is used to extract and link relationships from historical fault data. Specifically, a pre-trained large language model can be used to extract relationships from historical fault data and construct relationship chains between entities.

[0083] In this embodiment of the invention, the collection of robot warning data can originate from the robot's warning signals, and the warning signals can be judged. Specifically, the robot warning diagnosis steps can be as follows: Figure 3 As shown. Specifically, in Figure 3 Furthermore, the robot's fault self-diagnosis and maintenance decision-making method may also include:

[0084] In step S30, multiple time-series data points of the current robot within a preset time period are acquired. This preset time period can be determined based on actual needs, and multiple time-series data points within the preset time period can be collected at equal intervals to obtain the corresponding time values.

[0085] In step S31, multiple real-time parameters of the current robot are obtained according to formula (1).

[0086] (1)

[0087] in, For the robot's first Real-time parameters, such as displacement and temperature, For the first preset quantity, for Peak and valley data for each real-time parameter within a corresponding preset time period. for Each real-time parameter precedes the peak and valley data for the corresponding preset time period. One data point, for Each real-time parameter is based on peak and valley data for a corresponding preset time period. One data point, , The number is an integer. Specifically, it collects the values ​​before and after the peak or trough within the preset time period. These time values ​​effectively highlight the magnitude of change within the preset time period. Furthermore, It can be manually set, and must be less than or equal to 1 / 4 of the time series data within a preset time period. If the time series data before or after the peak / valley value is less than... Then, the time value with the furthest peak and valley value can be used as the subsequent time value for conversion.

[0088] In step S32, it is determined whether each real-time parameter is within the corresponding preset threshold range. The real-time parameters for the robot can include common fault monitoring parameters such as joint displacement and temperature. The preset threshold range can include, but is not limited to, numerical parameters, as well as text-based parameters such as fault codes. For example, real-time fault codes can be matched with fault codes in a library to determine if any suspected faults exist.

[0089] In step S33, if the real-time parameters are not within the corresponding preset threshold range, a suspected robot malfunction is determined, and an early warning is issued. Specifically, if the real-time parameters fluctuate abnormally or a fault code is matched, it can be determined that the robot may be malfunctioning, and an early warning can be issued.

[0090] In step S34, real-time warning data corresponding to the real-time parameters is obtained. For example, if the real-time parameter is a displacement distance error of 0.5mm, the displacement joint / component can be positioned, and the data can be summarized to obtain warning data, such as joint x tracking error of 0.5mm.

[0091] In steps S30 to S34, the early warning of potential robot malfunctions can be diagnosed based on traditional robot monitoring items, such as displacement and temperature monitoring. Simultaneously, multiple time-series data points for the corresponding monitoring items within a preset time period are used to comprehensively obtain the current real-time parameters, and a preset threshold is applied to each real-time parameter. If the comprehensive real-time parameters are outside the preset threshold range, it is determined that the robot may have a malfunction, and an early warning is issued simultaneously.

[0092] In this embodiment of the invention, after determining that the robot is suspected of having a malfunction, the real-time early warning data can be further verified, specifically as follows: Figure 4 As shown. Specifically, in Figure 4 Furthermore, the robot's fault self-diagnosis and maintenance decision-making method may also include:

[0093] In step S40, the warning nodes in the real-time warning data are obtained. For example, if the warning data is a joint x tracking error of 0.5mm, the corresponding warning node is a positioning error exceeding the limit.

[0094] In step S41, the warning node is input into the fault knowledge graph for querying to obtain multiple associated nodes of the warning node. Among them, associated nodes for positioning error exceeding the limit may include reducer wear, encoder noise, etc.

[0095] In step S42, the real-time association value of each associated node is obtained. The conversion of the real-time association value for each associated node can be as follows: Figure 5 As shown, specifically, in Figure 5 In this process, this step may include:

[0096] In step S420, multiple time-series data for each associated node within a preset time period are acquired. The preset time period for the associated node may have the same numerical range as the preset time period for the robot, but with a different meaning, as will be explained below. Furthermore, the time-series data may include time values ​​at equal intervals.

[0097] In step S421, the real-time association parameters of each associated node are obtained according to formula (2).

[0098] (2)

[0099] in, For the first Real-time association parameters of each associated node. For the second preset quantity, For the first Peak and valley data of each associated node in the corresponding preset time period For the first Each associated node is preceded by peak and valley data for the corresponding preset time period. One data point, For the first After the peak and valley data of each associated node in the corresponding preset time period One data point, , The number is an integer. Specifically, it uses the values ​​before and after the peak or trough within the preset time period. These time values ​​effectively highlight the magnitude of change within the preset time period. Furthermore, It can be manually set, and must be less than or equal to 1 / 4 of the time series data within a preset time period. If the time series data before or after the peak / valley value is less than... Then, the time value with the furthest peak and valley value can be used as the subsequent time value for conversion.

[0100] In step S422, the real-time association value of each associated node is obtained according to formula (3).

[0101] (3)

[0102] in, For the first Real-time association values ​​of each associated node. For the first Each associated node corresponds to a preset threshold range. Specifically, if the real-time association parameter of an associated node is not within the corresponding preset threshold range, the association value of that associated node can be set to 1, and vice versa. Furthermore, the preset threshold range of this associated node can be distinguished from the preset threshold range of the aforementioned real-time parameters, such as being designated as the first preset threshold range and the second preset threshold range, etc., which will not be elaborated further here.

[0103] In steps S40 to S42, the early warning nodes in the real-time early warning data are first input into the fault knowledge graph for querying, which yields multiple associated nodes of the early warning node and their corresponding real-time association parameters. Then, the corresponding real-time association value is obtained based on these parameters, facilitating subsequent calculation of the fault confidence level. A higher real-time association value indicates a more reliable fault.

[0104] In this embodiment of the invention, after obtaining multiple real-time correlation values ​​of the early warning node, the corresponding fault confidence level can be obtained based on these multiple real-time correlation values. The specific steps are as follows: Figure 6 As shown. Specifically, in Figure 6 In this process, this step may include:

[0105] In step S50, time-series data of the warning node and multiple associated nodes at the time of the warning node's failure are acquired. Specifically, based on the current real-time warning data / suspected fault data, parameter values ​​for the same fault can be obtained from historical fault data, including historical parameter values ​​of the warning node and multiple associated nodes, i.e., multiple time-series data / time values / parameter values ​​of the warning node and multiple associated nodes before and after the fault.

[0106] In step S51, the sensitivity coefficient between each associated node and the early warning node is obtained according to formula (4).

[0107] (4)

[0108] in, For the first The sensitivity coefficient between each associated node and the early warning node. To prevent the first [unclear] when a failure occurs at the early warning node The maximum parameter value of each associated node, which is also the maximum value of the historical parameters of the associated nodes before and after the failure. To prevent the first [unclear] when a failure occurs at the early warning node The minimum parameter value of each associated node, which is also the minimum historical parameter value of the associated node before and after the failure. This represents the maximum parameter value of the warning node when a failure occurs, which is also the maximum historical parameter value of the warning node before and after the failure. This refers to the minimum parameter value of the warning node when a failure occurs, which is also the minimum historical parameter value of the warning node before and after the failure. Specifically, the ratio of the change amplitude of the associated node and the warning node before and after the failure is used as the sensitivity coefficient, which can effectively reflect the correlation between the warning node and the associated node. The larger the change amplitude ratio, the stronger the correlation, and vice versa.

[0109] In step S52, the sensitivity coefficients of multiple associated nodes and early warning nodes are normalized.

[0110] In step S53, the fault confidence level of the early warning node is obtained according to formula (5).

[0111] (5)

[0112] in, The failure confidence level of the early warning node. For the first The normalized value of the sensitivity coefficient between each associated node and the early warning node. This represents the number of associated nodes. Specifically, this method limits the fault confidence of the warning node to between 0 and 1. The higher the fault confidence, the stronger the credibility that the warning node is faulty.

[0113] In steps S50 to S53, the relevant time-series data of the warning node's warning failure in the historical fault data is first obtained, that is, the historical parameter values ​​of the warning node and its associated nodes before and after the failure are obtained. Then, the correlation between the associated nodes and the warning node is determined based on the ratio of their fluctuation amplitudes before and after the failure, and the corresponding sensitivity coefficient is obtained. The failure confidence of the warning node is obtained by weighted summation based on the sensitivity coefficient of each associated node and the real-time correlation value.

[0114] In this embodiment of the invention, after calculating the fault confidence level of the early warning node, the final fault situation can be determined based on the fault confidence level. Specifically, it can be as follows: Figure 7 As shown. Specifically, in Figure 7 Furthermore, the robot's fault self-diagnosis and maintenance decision-making method may also include:

[0115] In step S60, it is determined whether the fault confidence level is greater than or equal to a confidence threshold. This confidence threshold may include 0.8-0.9, preferably 0.8.

[0116] In step S61, if the fault confidence level is greater than or equal to the confidence threshold, the robot is determined to be faulty. Specifically, if the fault confidence level is greater than or equal to the confidence threshold, it indicates that there are abnormal fluctuations in the synchronization of the associated nodes, and the robot's warning node can be directly determined to be faulty, and a corresponding maintenance / repair plan can be formulated and executed.

[0117] In step S62, the fault status is obtained based on real-time early warning data. This fault status may include the fault location / component, fault type, and fault time.

[0118] In step S63, if the fault confidence level is less than the confidence threshold, it is determined that the robot does not have a fault. Specifically, if the fault confidence level is less than the confidence threshold, it indicates that there are no abnormal fluctuations in the synchronization of the associated nodes. Therefore, it can be determined that the robot's warning node may be misjudged, and manual intervention is arranged to assist in the determination.

[0119] On the other hand, the present invention also provides a robot fault self-diagnosis and maintenance decision-making system based on a large model, which may include a knowledge graph construction module and a controller. Specifically, the knowledge graph construction module is used to collect historical fault data and construct a fault knowledge graph, and the controller is connected to the knowledge graph construction module to execute any of the robot fault self-diagnosis and maintenance decision-making methods described above.

[0120] In another aspect, the present invention also provides a computer-readable storage medium storing instructions for being read by a machine to cause the machine to execute any of the robot fault self-diagnosis and maintenance decision-making methods described above.

[0121] Through the above technical solution, the robot fault self-diagnosis and maintenance decision-making method and system based on a large model provided by this invention collects and preprocesses historical fault data of the robot, enabling the construction of a corresponding fault knowledge graph based on a pre-trained large language model. Simultaneously, it collects real-time early warning data of the robot, locates early warning nodes and associated nodes in the fault knowledge graph based on the early warning data, and obtains the fault confidence of the early warning node based on the real-time association value of the associated nodes. Then, it determines the real-time fault status of the early warning node based on the fault confidence, generates a corresponding maintenance plan, and executes it. By using multiple associated nodes to determine the fault confidence of the early warning node, it can verify and corroborate whether the early warning node has failed, effectively improving the accuracy and effectiveness of robot fault diagnosis, thereby avoiding misdiagnosis and reducing maintenance costs.

[0122] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0123] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0124] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0125] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0126] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0127] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, like read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0128] Computer-readable media include both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0129] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0130] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A robot fault self-diagnosis and maintenance decision-making method based on a large model, characterized in that, include: Collect and acquire historical fault data of the robot; The historical fault data is preprocessed, and a fault knowledge graph is constructed based on the preprocessed historical fault data. Obtain real-time early warning data for the robot currently described; Based on the real-time early warning data and the fault knowledge graph, the robot's early warning nodes and the real-time correlation values ​​of multiple associated nodes are obtained. The fault confidence of the early warning node is determined based on multiple real-time correlation values; The real-time fault status of the robot at the early warning node is determined based on the fault confidence level. Based on the real-time fault situation, obtain the corresponding maintenance plan and execute the maintenance plan.

2. The robot fault self-diagnosis and maintenance decision-making method according to claim 1, characterized in that, Preprocessing the historical fault data includes data cleaning.

3. The robot fault self-diagnosis and maintenance decision-making method according to claim 1, characterized in that, Constructing a fault knowledge graph based on the preprocessed historical fault data includes: Define entity types, relationship types, and entity attributes; A pre-trained large language model is used to extract and align entities from the historical fault data. A pre-trained large language model is used to extract and link relationships in the historical fault data.

4. The robot fault self-diagnosis and maintenance decision-making method according to claim 1, characterized in that, Obtaining the real-time early warning data of the robot includes: Obtain multiple time-series data of the robot within a preset time period; According to formula (1), obtain multiple real-time parameters of the robot. ，（1） in, For the robot's first One real-time parameter, For the first preset quantity, For the first Peak and valley data for each real-time parameter within a corresponding preset time period. for Each real-time parameter precedes the peak and valley data for the corresponding preset time period. One data point, for Each real-time parameter is based on peak and valley data for a corresponding preset time period. One data point, , Numbered by integer; Determine whether each of the real-time parameters is within the corresponding preset threshold range; If the real-time parameters are not within the corresponding preset threshold range, the robot is suspected of malfunctioning, and an early warning is issued. Obtain the real-time warning data corresponding to the real-time parameters.

5. The robot fault self-diagnosis and maintenance decision-making method according to claim 4, characterized in that, The real-time association values ​​of the robot's early warning nodes and multiple associated nodes are obtained based on the real-time early warning data and the fault knowledge graph, including: Obtain the early warning nodes from the real-time early warning data; The warning node is input into the fault knowledge graph for querying to obtain multiple associated nodes of the warning node; Obtain the real-time association value for each of the associated nodes.

6. The robot fault self-diagnosis and maintenance decision-making method according to claim 5, characterized in that, Obtaining the real-time association value for each of the associated nodes includes: Obtain multiple time-series data of each associated node within a preset time period; The real-time association parameters of each associated node are obtained according to formula (2). ，（2） in, For the first Real-time association parameters of the associated nodes, For the second preset quantity, For the first Peak and valley data of the associated nodes in the corresponding preset time period. For the first The associated nodes are in front of the peak and valley data for the corresponding preset time period. One data point, For the first After the peak and valley data of the associated nodes in the corresponding preset time period One data point, , Numbered by integer; The real-time association value of each associated node is obtained according to formula (3). ，（3） in, For the first The real-time association value of each of the associated nodes. For the first The preset threshold range corresponding to each of the associated nodes.

7. The robot fault self-diagnosis and maintenance decision-making method according to claim 6, characterized in that, Determining the fault confidence of the early warning node based on multiple real-time correlation values ​​includes: Acquire the time-series data of the early warning node and multiple associated nodes when the early warning node fails; The sensitivity coefficient between each associated node and the early warning node is obtained according to formula (4). ，（4） in, For the first The sensitivity coefficient between the associated nodes and the early warning nodes. When the warning node fails, the first The maximum parameter value of the associated nodes. When the warning node fails, the first The minimum parameter value of the associated nodes, This refers to the maximum parameter value of the early warning node when it fails. This refers to the minimum parameter value of the early warning node when a failure occurs. The sensitivity coefficients of the multiple associated nodes and the early warning node are normalized; The fault confidence level of the early warning node is obtained according to formula (5). ，（5） in, The fault confidence level of the early warning node. For the first The normalized value of the sensitivity coefficient between the associated node and the early warning node. The number of associated nodes.

8. The robot fault self-diagnosis and maintenance decision-making method according to claim 7, characterized in that, Determining the current real-time fault status of the robot at the early warning node based on the fault confidence level includes: Determine whether the fault confidence level is greater than or equal to the confidence threshold; If the confidence level of the fault is greater than or equal to the confidence threshold, it is determined that the robot has a fault. The fault status is obtained based on the real-time early warning data; If the confidence level of the fault is less than the confidence threshold, the robot is determined to be fault-free.

9. A robot fault self-diagnosis and maintenance decision-making system based on a large model, characterized in that, include: The knowledge graph construction module is used to collect historical fault data and build a fault knowledge graph. The controller, connected to the knowledge graph construction module, is used to execute the robot fault self-diagnosis and maintenance decision-making method as described in any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that are read by a machine to cause the machine to perform the robot fault self-diagnosis and maintenance decision-making method as described in any one of claims 1-8.

Images