Robot management method and device, terminal equipment and computer program product

Abstract

The invention is suitable for the technical field of robots, and provides a robot management method and device, terminal equipment and a computer program product, and the method comprises the steps: receiving multi-source data associated with a target robot; analyzing the multi-source data to obtain a health state analysis report of the target robot; performing fault prediction on the target robot based on the health state analysis report and the multi-source data to obtain a fault prediction result; generating a prediction operation and maintenance plan based on the fault prediction result; and executing a target operation and maintenance operation corresponding to the predicted operation and maintenance plan on the target robot. Compared with the prior art, the method has the advantages that the abnormity and health state of the robot can be analyzed in real time through multi-source data, so that fine-grained health changes of machine parts are sensed in real time, and a health state analysis report is generated. Meanwhile, potential faults can be predicted, and an operation and maintenance plan is automatically generated and scheduled for execution, so that shutdown loss and operation cost can be reduced.

Description

Technical Field

[0001] This application belongs to the field of robotics technology, and in particular relates to a robot management method, device, terminal equipment, and computer program product. Background Technology

[0002] With the widespread application of robotics technology in manufacturing, logistics, education services, healthcare, and public safety, the scale and complexity of robot swarms are constantly increasing. The operating status of each robot and the health status of its components have a significant impact on the overall stability of the system.

[0003] However, existing technologies typically rely on periodic inspections or simple threshold-based alarm mechanisms for robot maintenance. It is evident that existing technologies cannot perceive fine-grained health changes of machine components in real time, nor can they predict potential failures in advance. They often can only take action after a failure occurs, resulting in significant downtime losses and high operating costs. Summary of the Invention

[0004] This application provides a robot management method, device, terminal equipment, and computer program product to solve the problems of existing technologies that cannot perceive fine-grained health changes of machine parts in real time, nor can they predict potential faults in advance. Often, measures can only be taken after a fault occurs, resulting in large downtime losses and high operating costs.

[0005] In a first aspect, embodiments of this application provide a robot management method, including:

[0006] Receive multi-source data associated with the target robot; By analyzing multi-source data, a health status analysis report of the target robot is obtained; Based on health status analysis reports and multi-source data, fault prediction is performed on the target robot to obtain fault prediction results; Based on the fault prediction results, a predictive operation and maintenance plan is generated. Perform the target maintenance operations corresponding to the predicted maintenance plan on the target robot.

[0007] The beneficial effects of the embodiments in this application compared with the prior art are: This application provides a robot management method that involves receiving multi-source data associated with a target robot; analyzing the multi-source data to obtain a health status analysis report for the target robot; predicting faults in the target robot based on the health status analysis report and the multi-source data to obtain fault prediction results; generating a predictive maintenance plan based on the fault prediction results; and executing target maintenance operations corresponding to the predictive maintenance plan on the target robot. Compared with existing technologies, this application can analyze the robot's anomalies and health status in real time using multi-source data, thereby sensing fine-grained health changes of machine components in real time and generating a health status analysis report. Simultaneously, it can predict potential faults, automatically generate maintenance plans, and schedule their execution, thereby reducing downtime losses and operating costs.

[0008] In some possible embodiments, multi-source data is analyzed to obtain a health status analysis report of the target robot, including: Multi-source data undergoes preprocessing and feature extraction to obtain multi-dimensional indicators; The health status model of the target robot is obtained by modeling the state of the target robot based on multidimensional indicators. Anomaly detection and health assessment of the target robot are performed based on multidimensional indicators to obtain the anomaly risk level and current health index of the target robot; A health status analysis report is generated based on the health status model, abnormal risk level, and current health index.

[0009] In the above embodiments, data preprocessing filters out invalid and interfering data, and feature extraction extracts multidimensional indicators strongly correlated with robot health, thereby improving data quality. Then, a health status model is constructed based on these multidimensional indicators, transforming the robot's complex health status into quantifiable and comparable model parameters. This overcomes the limitations of binary judgments of either good or bad, achieving a refined and hierarchical description of the robot's health status. Subsequently, anomaly detection and health assessment are conducted simultaneously using multi-source data, ultimately outputting the anomaly risk level and current health index, clarifying both the severity of the risk and quantifying the current health level. Finally, the three core elements—health status model, anomaly risk level, and current health index—are integrated to generate an analysis report. This report includes both the overall model conclusions of robot health and specific risk classifications and quantitative indicators, enhancing the comprehensiveness, interpretability, and guidance of the health status analysis report.

[0010] In some possible embodiments, the state of the target robot is modeled based on multidimensional indicators to obtain a health state model of the target robot, including: Based on mechanical equations and multidimensional indices, a physical flow model is established between the degradation state variables and multidimensional indices corresponding to each component of the target robot; the physical flow model is a continuous-time state-space model. Based on multi-source data, the residual between the predicted value and the corresponding actual value is obtained; The data flow model is obtained by training the pre-constructed neural differential equations based on multidimensional indicators, residuals, and degenerate state variables. The physical flow model and the data flow model are fused together to obtain the health status model.

[0011] In the above embodiments, the physical flow model is constructed based on mechanical equations and multidimensional indicators, anchoring the physical laws between the degradation state quantities of robot components and monitoring indicators, thereby ensuring the reliability of the model's theoretical foundation. Meanwhile, the data flow model is trained based on neural differential equations, using residuals (the deviation between predicted and actual values) to capture actual working condition deviations not covered by the physical model, thus making up for the idealization shortcomings of the physical flow model. Finally, through model fusion, a dual guarantee of physical law constraints and actual data correction is achieved, enabling the final health state model to accurately represent the robot's true health state. It will not lead to misjudgment due to ignoring physical laws, nor will it be biased due to data noise, thereby providing a high-precision state benchmark for subsequent fault prediction.

[0012] In some possible embodiments, the anomaly risk level is obtained as follows: Anomaly detection is performed on multi-source data based on the target rule engine to obtain anomaly detection results; Multi-source data is input into the anomaly detection model for anomaly detection to obtain the target anomaly data; The target abnormal data is input into the abnormality type detection model for type detection, and the abnormality detection result is obtained; The anomaly judgment results, target anomaly data, and anomaly detection results are weighted and fused to obtain the anomaly risk level.

[0013] In the above embodiments, a basic anomaly determination is first performed using a target rule engine to quickly filter common anomalies that conform to known rules. Then, an anomaly identification model is used to mine latent anomaly features in multi-source data and output target anomaly data, thereby compensating for the rule engine's blind spot in recognizing unknown anomalies. Next, an anomaly type detection model is used to classify the target anomaly data, clarifying the specific type of anomaly. This ensures both rapid identification of common anomalies and accurate capture of rare and complex anomalies, while also minimizing false positive and false negative rates. Furthermore, the weighted fusion of the three core elements—rule engine determination results, target anomaly data, and anomaly type detection results—improves the scientific rigor and objectivity of risk level classification.

[0014] In some possible embodiments, the multi-source data includes historical operational data, and the health status analysis report includes a health status model; based on the health status analysis report and the multi-source data, fault prediction is performed on the target robot to obtain fault prediction results, including: The health status model and historical operating data are input into the remaining service life prediction model for processing to obtain degradation trend information of each component of the target robot. By analyzing the degradation trend information, fault prediction results are obtained.

[0015] In the above embodiments, the health status model and historical operating data are simultaneously input into the remaining service life prediction model. This not only anchors the current health status benchmark but also incorporates the patterns of historical degradation trends, enabling the accurate output of degradation trend information for each component, thereby improving the accuracy of subsequent fault prediction.

[0016] In some possible embodiments, the fault prediction results include multiple target fault types corresponding to each fault. Based on the fault prediction results, a predictive operation and maintenance plan is generated, including: The robot knowledge graph is used to filter out various target operation and maintenance strategies that are associated with the fault prediction results. The robot knowledge graph includes the relationships between the robot's various components, fault types, and operation and maintenance strategies. The priority of each target operation and maintenance strategy is obtained by prioritizing each target operation and maintenance strategy based on the genetic algorithm. Based on the operational strategies of each target and the corresponding task priorities, a predictive operational plan is generated.

[0017] In the above embodiments, the robot knowledge graph can be used to filter out target operation and maintenance strategies that match it and are highly targeted, without the need for manual retrieval. Then, the genetic algorithm can be used to achieve optimal scheduling of operation and maintenance resources when there are multiple target operation and maintenance strategies, minimizing the impact of failures on robot operation and controlling operation and maintenance costs. Finally, the predictive operation and maintenance plan generated based on each target operation and maintenance strategy and the corresponding task priority improves the executability and guidance of the operation and maintenance plan.

[0018] In some possible embodiments, the predictive maintenance plan includes multiple target maintenance strategies, and the target maintenance operations corresponding to the predictive maintenance plan are executed on the target robot, including: The operation and maintenance strategies for each target are analyzed to obtain the operation and maintenance execution entities corresponding to each target operation and maintenance strategy; Each target operation and maintenance strategy is sent to the corresponding operation and maintenance execution entity, so that each operation and maintenance execution entity can perform the corresponding target operation and maintenance operation on the target robot.

[0019] In the above embodiments, by analyzing the target operation and maintenance strategy, the corresponding operation and maintenance execution entity can be accurately matched, so that each operation and maintenance strategy is undertaken by an execution entity with corresponding capabilities, avoiding operational errors and task delays caused by mismatch, and improving the execution efficiency and accuracy of operation and maintenance operations.

[0020] In some possible embodiments, after performing the target maintenance operations corresponding to the predicted maintenance plan on the target robot, the method further includes: Receive work data from the target robot after it has completed the target maintenance operation; By analyzing the work data, the target operating status and target health index of the target robot can be obtained; The system visualizes the target's operational status and health index.

[0021] In the above embodiments, by receiving and analyzing the work data after maintenance, and outputting the target operating status and target health index, the actual effect of maintenance operations can be quantitatively verified. At the same time, the visual display of the target operating status and target health index also makes it convenient for relevant personnel to quickly and intuitively grasp the status of the robot after maintenance, without having to spend time analyzing complex data reports, thereby improving the response efficiency of maintenance management.

[0022] In some possible embodiments, after receiving the work data of the target robot after it has completed the target maintenance operation, the method further includes: Anomaly analysis is performed based on working data to obtain anomaly analysis results; When the anomaly analysis result indicates the presence of an anomaly, an alarm message is output; the alarm message is used to alert the target robot that an anomaly exists.

[0023] In the above embodiments, anomaly analysis can be performed on the work data after maintenance, thereby accurately identifying hidden problems that have not been resolved during maintenance operations and making up for the loophole of the process ending as soon as maintenance operations are completed; then, when an anomaly is detected, alarm information is immediately output to prompt relevant personnel to handle it, thereby improving the reliability of maintenance.

[0024] Secondly, embodiments of this application provide a robot management device, including: The receiving unit is used to receive multi-source data associated with the target robot; The first analysis unit is used to analyze the multi-source data to obtain a health status analysis report of the target robot. The fault prediction unit is used to predict the fault of the target robot based on the health status analysis report and the multi-source data, and obtain the fault prediction result. The first generation unit is used to generate a predictive operation and maintenance plan based on the fault prediction results; The first execution unit is used to perform target maintenance operations on the target robot that correspond to the predicted maintenance plan.

[0025] Thirdly, embodiments of this application provide a terminal device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the robot management method as described in any one of the first aspects above.

[0026] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the robot management method as described in any one of the first aspects above.

[0027] Fifthly, embodiments of this application provide a computer program product that, when run on a terminal device, enables the terminal device to execute the robot management method described in any one of the first aspects.

[0028] It should be noted that the specific beneficial effects corresponding to any of the second to fifth aspects mentioned above can be found in the beneficial effects described in the first aspect, and will not be repeated here. Attached Figure Description

[0029] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0030] Figure 1 This is a flowchart illustrating the implementation of a robot management method according to an embodiment of this application; Figure 2 This is a flowchart illustrating the specific implementation of step S102 in a robot management method provided in an embodiment of this application; Figure 3 This is a flowchart illustrating the specific implementation of step S202 in a robot management method provided in an embodiment of this application; Figure 4 This is a flowchart illustrating the specific implementation of step S103 in a robot management method provided in an embodiment of this application; Figure 5 This is a flowchart illustrating the specific implementation of step S104 in a robot management method provided in an embodiment of this application; Figure 6 This is a flowchart illustrating the implementation of a robot management method according to another embodiment of this application; Figure 7This is an overall flowchart of a robot management method provided in an embodiment of this application; Figure 8 This is a schematic diagram of the structure of a robot management device provided in an embodiment of this application; Figure 9 This is a schematic diagram of the structure of a terminal device provided in an embodiment of this application. Detailed Implementation

[0031] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0032] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0033] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0034] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0035] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0036] References to "one embodiment" or "some embodiments" in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0037] Please see Figure 1 , Figure 1 This is a flowchart illustrating the implementation of a robot management method according to an embodiment of this application. In this embodiment, the robot management method is executed by a terminal device. The terminal device includes, but is not limited to, devices such as laptops, desktop computers, and computers.

[0038] like Figure 1 As shown, a robot management method provided in one embodiment of this application may include steps S101 to S105, which are detailed below: In S101, multi-source data associated with the target robot is received.

[0039] In this embodiment of the application, the target robot is used to describe a robot that needs to be monitored for health at the current moment.

[0040] In practical applications, the aforementioned target robots include, but are not limited to, service robots and inspection robots. Service robots can be used in fields such as manufacturing, logistics, education services, healthcare, and public safety.

[0041] In this embodiment, the multi-source data includes, but is not limited to, various sensor data, robot operation logs, task execution data, and historical maintenance records.

[0042] It should be noted that the various sensor data include, but are not limited to: motor temperature, current, voltage, vibration parameters, and joint position.

[0043] Robot operation logs include, but are not limited to: the target robot's startup log, shutdown log, the target robot's hardware module (such as the central processing unit (CPU), memory, communication module, etc.) operation status log, software program execution, error and interface call logs, and abnormal alarm logs, etc.

[0044] Task execution data includes, but is not limited to: task instructions issued by the host computer, task execution progress and task completion, task execution parameters (such as work trajectory data, force / torque data, motion status, process parameters, energy consumption and heat data), task execution results (success / failure / interruption) and interruption reason records, etc.

[0045] Historical maintenance records include, but are not limited to, the target robot’s past fault records (including fault location, fault type, fault occurrence time and fault solution), operation and maintenance records (including operation and maintenance time, operation and maintenance content, replacement parts and operators, etc.), calibration records (such as calibration parameters and calibration time of sensors or motion modules), and downtime records (including downtime and reasons), etc.

[0046] In one implementation of this application, various sensors located at different parts of the target robot can collect various sensor data of the target robot in real time.

[0047] These sensors include, but are not limited to: temperature sensors, current sensors, voltage sensors, vibration sensors, and position sensors.

[0048] In practical applications, current sensors can be Hall current sensors, voltage sensors can be Hall voltage sensors, vibration sensors can be piezoelectric vibration sensors, and position sensors can include accelerometers and angular velocity sensors.

[0049] Therefore, in this embodiment, the terminal device can receive various sensor data of the target robot in real time through the various sensors that are wirelessly connected to it.

[0050] In another implementation of this application, the terminal device can obtain the robot's operation logs, task execution data, and historical maintenance records in real time through a server with which it is wirelessly connected. The server can be a robot management platform.

[0051] In S102, the multi-source data is analyzed to obtain a health status analysis report of the target robot.

[0052] In this embodiment of the application, after obtaining the above-mentioned multi-source data, the terminal device can perform data preprocessing on the multi-source data to obtain standard data, so as to eliminate data noise and fill in missing values, thereby purifying the data for subsequent analysis.

[0053] In some possible embodiments, data preprocessing includes, but is not limited to, data cleaning, data standardization, and data fusion.

[0054] Data cleaning involves removing abnormal fluctuations from various sensor data using filtering algorithms (such as Kalman filtering and median filtering); filling in missing data using time interpolation or the average of similar devices; and deleting duplicate and invalid operation logs and historical maintenance records.

[0055] Data standardization is the process of converting sensor data of different dimensions (such as temperature (°C), current (A), vibration (mm / s)) into the same dimension (such as the [0, 1] interval) to avoid the influence of dimensional differences on the analysis results.

[0056] Data fusion is based on device identification and timestamps, which integrates various sensor data, robot operation logs, task execution data and historical maintenance records of the target robot to form a robot full-dimensional dataset with time series as the core (such as the fused data of the robot's motor temperature, joint load, operation logs, task execution data and maintenance records at a certain moment).

[0057] In this embodiment, after obtaining standard data, the terminal device can extract features from the standard data to obtain multi-dimensional indicators. These multi-dimensional indicators include, but are not limited to, temperature rise rate, load fluctuation, and vibration amplitude.

[0058] Specifically, terminal devices can extract feature values ​​of each indicator from standard data through time-series feature analysis (such as calculating the mean, variance, trend slope, or abrupt change points of each indicator) and frequency domain analysis (such as performing Fourier transform on vibration data and extracting feature frequencies), thereby obtaining multidimensional indicators.

[0059] In this embodiment, after obtaining multi-dimensional indicators, the terminal device can compare each indicator with its corresponding health threshold to obtain the health monitoring result of each indicator. The health monitoring result includes, but is not limited to, whether the indicator is healthy, and the difference between the result and the health threshold.

[0060] Afterwards, the terminal device can score the health of each indicator based on the health monitoring results of the above-mentioned indicators.

[0061] Specifically, when a terminal device detects that the health monitoring result of a certain dimension indicator is healthy, it can determine that the health score of that dimension indicator is 100; when a terminal device detects that the health monitoring result of a certain dimension indicator is unhealthy, it can determine the health score of that dimension indicator based on the difference between that dimension indicator and its corresponding health threshold.

[0062] It should be noted that the difference between this dimension indicator and its corresponding health threshold is inversely proportional to the health score of this dimension indicator; that is, the larger the difference, the lower the health score, and the smaller the difference, the higher the health score.

[0063] In this embodiment of the application, after obtaining the health scores of each dimension, the terminal device can perform a weighted summation of the health scores of each dimension to obtain the overall health score of the target robot.

[0064] Afterwards, the terminal device can combine the multi-source data, multi-dimensional indicators, health scores of each dimension indicator, and overall health score of the target robot to generate a health status analysis report of the target robot.

[0065] In one embodiment of this application, the terminal device can specifically be configured as follows: Figure 2 Steps S201 to S204 shown implement step S102, as detailed below: In S201, data preprocessing and feature extraction are performed on multi-source data to obtain multi-dimensional indicators.

[0066] In this embodiment, the specific implementation process of step S201 can be referred to the specific implementation process of obtaining multi-dimensional indicators in step S102, and will not be repeated here.

[0067] In S202, the state of the target robot is modeled based on multidimensional indicators to obtain the health status model of the target robot.

[0068] In this embodiment, each dimension index may include sub-indicators at different times. The terminal device can construct multiple index curves based on the sub-indicators at different times in each dimension index. Then, the terminal device can combine the multiple index curves to obtain the health status model of the target robot.

[0069] In one embodiment of this application, the terminal device can specifically be configured as follows: Figure 3 Steps S301 to S304 shown implement step S202, as detailed below: In S301, a physical flow model is established between the degradation state variables and multidimensional indices corresponding to each component of the target robot, based on mechanical equations and multidimensional indices; the physical flow model is a continuous-time state-space model.

[0070] In this embodiment, the terminal device can identify the core components of the target robot that can affect its operational health based on the various components that have failed during the target robot's historical development. These core physical components include, but are not limited to, joint motors, drive wheels, power batteries, and built-in sensors.

[0071] Subsequently, the terminal device can determine the corresponding quantifiable degradation state quantity of each of the above-mentioned components based on their existing mechanical characteristics and operating rules.

[0072] Among them, degradation state quantities include, but are not limited to: motor efficiency attenuation coefficient, drive wheel friction coefficient, battery capacity attenuation rate, sensor response sensitivity coefficient, and sensor accuracy attenuation coefficient.

[0073] In this embodiment, for each component, the terminal device can establish a physical mapping relationship between multidimensional indices and the degradation state variables of the component based on existing mechanical equations, kinematic equations, and failure physics equations, thereby obtaining the physical equation corresponding to each degradation state variable. Then, the terminal device can fit the state function and observation function based on the physical equation, and combine them with the degradation state variables (i.e., state variables) and multidimensional indices (input variables) of the component to generate a continuous-time state-space sub-model.

[0074] Afterwards, the terminal device can integrate the continuous-time state-space sub-models of each component to obtain the final physical flow model.

[0075] In S302, the residual between the predicted value and the corresponding actual value is obtained based on multi-source data.

[0076] In this embodiment, the terminal device can input historical multi-source data corresponding to the multi-source data into the aforementioned physical flow model for calculation to obtain predicted values. These predicted values ​​may include predicted values ​​for degradation state quantities and predicted values ​​for multi-dimensional indicators.

[0077] The terminal device can then calculate the residual between the predicted value and the corresponding actual value in the multi-source data.

[0078] In S303, a data flow model is obtained by training a pre-constructed neural differential equation based on multidimensional indices, residuals, and degenerate state variables.

[0079] It should be noted that neural differential equations are a model that combines deep learning with differential equations. They can directly model continuous-time dynamic systems and can be adapted to the continuous-time state-space characteristics of the physical flow model mentioned above. There is no need to discretize the time series, and it can accurately capture the continuous change pattern of the target robot's state.

[0080] In this embodiment, after obtaining the above residuals, the terminal device can train the pre-constructed neural differential equations based on multidimensional indicators, residuals, and degenerate state variables to obtain a data flow model.

[0081] Specifically, the data flow model described above is as follows: ; in, This represents the degenerate state quantity output by the physical flow model. Indicates multidimensional indicators, Represents the residual. Represents the state vector of the data flow model. The first derivative of the state vector in the data flow model. This represents the residual fit value of the output of the data flow model. The state neural network is constructed using a multilayer perceptron and residual connections to fit the state change patterns over continuous time. The output neural network is constructed using a lightweight multilayer perceptron to fit the nonlinear behavior of the residuals.

[0082] In S304, the physical flow model and the data flow model are fused to obtain the health status model.

[0083] In this embodiment, the terminal device can fuse the physical flow model with physical interpretability and the data flow model with data fitting accuracy to obtain a target robot health state model that has both physical interpretability, data accuracy, and continuous time dynamic characteristics. This allows the fused health state model to retain the interpretability of the physical mechanism and have the accuracy of data-driven operation.

[0084] It should be noted that the fusion processing method can include weighted fusion.

[0085] In some possible embodiments, the terminal device may perform weighted fusion of variables of the same type in the physical flow model and the data flow model to obtain the final health status model.

[0086] Combining steps S301-S304, the physical flow model is constructed based on mechanical equations and multi-dimensional indicators, anchoring the physical laws between the degradation state quantities of robot components and monitoring indicators, thus ensuring the reliability of the model's theoretical foundation. Simultaneously, the data flow model is trained based on neural differential equations, utilizing residuals (the deviation between predicted and actual values) to capture actual working condition deviations not covered by the physical model, compensating for the idealization shortcomings of the physical flow model. Finally, through model fusion, a dual guarantee of physical law constraints and actual data correction is achieved, enabling the final health state model to accurately represent the robot's true health state. It avoids misjudgments due to ignoring physical laws and deviations due to data noise, thus providing a high-precision state benchmark for subsequent fault prediction.

[0087] In S203, anomaly detection and health assessment of the target robot are performed based on multi-dimensional indicators to obtain the anomaly risk level and current health index of the target robot.

[0088] In this embodiment, the terminal device can input multi-dimensional indicators into a trained anomaly detection model for processing to obtain the anomaly risk level of the target robot.

[0089] It should be noted that the anomaly detection model can be obtained by training a pre-built first deep learning model based on a preset sample set. Each sample in the preset sample set includes a multi-dimensional indicator and the corresponding anomaly risk level. When training the pre-built first deep learning model, the multi-dimensional indicator of each sample is used as the input, and the corresponding anomaly risk level is used as the output. Through training, the first deep learning model can learn the correspondence between all possible multi-dimensional indicators and anomaly risk levels, and the trained first deep learning model becomes the anomaly detection model.

[0090] The aforementioned first deep learning model may include an input layer, a feature layer, and an output layer. The input layer receives multi-dimensional indicators; the feature layer extracts features from the multi-dimensional indicators to obtain a feature vector; and the output layer performs detection on the feature vector to output the result.

[0091] In this embodiment, the abnormal risk levels include, but are not limited to, Level 1, Level 2, Level 3, and Level 4.

[0092] The first level (e.g., value 0) is used to represent no abnormality, the second level (e.g., value 1) is used to represent slight abnormality, the third level (e.g., value 2) is used to represent moderate abnormality, and the fourth level (e.g., value 3) is used to represent severe abnormality.

[0093] It should be noted that "no anomalies" is used to describe that all multidimensional indicators are normal.

[0094] Minor anomalies are used to describe anomalies that do not affect the normal operation of the target robot and allow it to continue running.

[0095] Moderate anomalies are used to describe anomalies that affect some operational performance of the target robot and require attention.

[0096] Severe anomalies are used to describe anomalies that affect the core operational capabilities of the target robot, pose potential faults, and require timely handling.

[0097] In one embodiment of this application, the terminal device can input multidimensional indicators into a trained health assessment model for processing to obtain the current health index of the target robot.

[0098] It should be noted that the health assessment model can be obtained by training a pre-built XGBoost model based on a preset sample set.

[0099] In another embodiment of this application, the terminal device can compare each dimension indicator with its corresponding health threshold to obtain the indicator health monitoring result for each dimension indicator. The indicator health monitoring result includes, but is not limited to, whether the indicator is healthy, and the difference between the result and the health threshold.

[0100] Afterwards, the terminal device can score the health of each indicator based on the health monitoring results of the above-mentioned indicators.

[0101] Specifically, when a terminal device detects that the health monitoring result of a certain dimension indicator is healthy, it can determine that the health score of that dimension indicator is 100; when a terminal device detects that the health monitoring result of a certain dimension indicator is unhealthy, it can determine the health score of that dimension indicator based on the difference between that dimension indicator and its corresponding health threshold.

[0102] It should be noted that the difference between this dimension indicator and its corresponding health threshold is inversely proportional to the health score of this dimension indicator; that is, the larger the difference, the lower the health score, and the smaller the difference, the higher the health score.

[0103] In this embodiment of the application, after obtaining the health scores of each dimension, the terminal device can perform a weighted summation of the health scores of each dimension to obtain the overall health score of the target robot, that is, the current health index of the target robot.

[0104] In another embodiment of this application, the terminal device may also obtain the abnormal risk level according to the following steps, detailed below: Anomaly detection is performed on multi-source data based on the target rule engine to obtain anomaly detection results; Multi-source data is input into the anomaly detection model for anomaly detection to obtain the target anomaly data; The target abnormal data is input into the abnormality type detection model for type detection, and the abnormality detection result is obtained; The anomaly judgment results, target anomaly data, and anomaly detection results are weighted and fused to obtain the anomaly risk level.

[0105] It should be noted that the target rule engine includes a data adaptation layer, a rule base, an inference engine, and a result output layer.

[0106] The data adaptation layer receives the aforementioned multi-source data and performs format conversion and dimension mapping to provide standardized data with a unified input format for the inference engine. The rule base stores all rules for robot anomaly detection, supporting rule creation, deletion, modification, retrieval, hierarchical configuration, enabling, and disabling. The inference engine, the core execution layer of the target rule engine, loads rules from the rule base and performs real-time matching and logical reasoning on the adapted multi-source data. The result output layer structures the inference engine's judgment results, outputting anomaly detection results with judgment criteria, rule identifiers, and anomaly locations.

[0107] The rule base includes, but is not limited to, sensor data rules, operation log rules, task execution data rules, and historical maintenance record association rules.

[0108] Among them, sensor data rules include, but are not limited to, single indicator threshold rules (such as motor temperature threshold to determine abnormal motor temperature, joint vibration threshold to determine abnormal joint, etc.), time-series trend rules (such as temperature rise rate threshold to determine abnormal motor temperature, load fluctuation threshold to determine abnormal load fluctuation, and vibration amplitude threshold to determine abnormal vibration, etc.), and multi-indicator combination rules (such as motor temperature threshold + motor load rate threshold + running time threshold to determine motor overload, etc.).

[0109] The operation log rules include, but are not limited to, error code rules (such as detecting error code E101 (overcurrent) for the motor drive module and error code C003 (timeout) for the lidar communication module, which are directly judged as abnormalities of the corresponding module), frequency rules (if the same module reports more than 5 errors within 10 minutes and the number of system startup failures is more than 2, it is judged as abnormal module stability), and status rules (such as the online rate of the core control module is less than 99% and the success rate of communication interface calls is less than 95%, which are judged as abnormal system operation).

[0110] The task execution data rules include, but are not limited to, task completion rules (such as judging whether task completion is abnormal by a single batch task completion rate threshold, judging the robot's operational capability as abnormal by N consecutive failures of core tasks), execution efficiency rules (such as judging task execution as abnormal by a time threshold that the task execution time exceeds the rated time), and interruption rules (such as judging hardware-related operation as abnormal by the number of task interruptions within a set time if the interruption reason is hardware failure).

[0111] Historical maintenance record association rules include, but are not limited to, maintenance cycle rules (such as judging anomalies when the running time of core components (reducers, bearings, etc.) exceeds the rated maintenance cycle threshold, and judging maintenance warnings when sensors exceed the calibration cycle by 30 days), and fault association rules (such as raising the anomaly judgment level if a component has experienced two similar faults in the past 3 months and real-time multi-source data triggers the relevant warning rules for that component).

[0112] In this embodiment, the target rule engine receives the aforementioned multi-source data through the data adaptation layer, the inference engine completes rule matching and logical reasoning, and finally outputs structured anomaly judgment results through the result output layer.

[0113] In this embodiment, the terminal device can also input multi-source data into the anomaly recognition model for anomaly recognition to obtain target anomaly data.

[0114] It should be noted that the anomaly detection model can be a temporal autoencoder (TAE) model that uses unsupervised learning and temporal feature extraction.

[0115] The aforementioned TAE includes an input layer, a temporal feature extraction layer, an encoding layer, a decoding layer, a reconstructed output layer, and an error calculation layer. Specifically, the input layer receives multi-source data; the temporal feature extraction layer uses a lightweight GRU layer (2 layers, with the number of hidden layer neurons being a preset multiple of the input dimension) to capture the temporal correlation features of the multi-source data and output a temporal fusion feature vector; the encoding layer uses a fully connected layer and ReLU activation to perform dimensionality reduction encoding on the above temporal fusion feature vector to obtain low-dimensional hidden layer features (the dimension is 1 / 4 to 1 / 2 of the input dimension) to mine the core features of the data; the decoding layer uses a combination of fully connected layers and GRU layers to perform dimensionality up-construction on the above low-dimensional hidden layer features, restoring them to a temporal feature vector with the same dimension as the input layer; the reconstruction output layer outputs the reconstructed multi-source data feature vector, which corresponds one-to-one with the multi-source data in the input layer; the error calculation layer calculates the dimensionality-wise reconstruction error (which can be MSE) between the multi-source data and its corresponding reconstructed feature vector, sets a reconstruction error threshold (based on the error distribution calibration of historical normal data), and determines an anomaly if the calculated reconstruction error exceeds the above reconstruction error threshold.

[0116] In this embodiment, after obtaining the target abnormal data, the terminal device can input the target abnormal data into the anomaly type detection model for type detection to obtain the anomaly detection result. The anomaly detection result includes, but is not limited to, the anomaly type. The anomaly type detection model is trained from a pre-built neural network model.

[0117] In this embodiment, after obtaining the anomaly determination result, the target anomaly data, and the anomaly detection result, the terminal device can determine a first risk level based on the anomaly determination result, a second risk level based on the target anomaly data, and a third risk level based on the anomaly detection result.

[0118] It should be noted that the first, second, and third risk levels mentioned above can all be represented by numerical values ​​(such as 0, 1, 2, and 3, etc.), and the larger the number, the higher the risk level.

[0119] Then, the terminal device can perform a weighted summation of the first risk level, the second risk level, and the third risk level to obtain the final abnormal risk level.

[0120] In the above embodiments, a basic anomaly determination is first performed using a target rule engine to quickly filter common anomalies that conform to known rules. Then, an anomaly identification model is used to mine latent anomaly features in multi-source data and output target anomaly data, thereby compensating for the rule engine's blind spot in recognizing unknown anomalies. Next, an anomaly type detection model is used to classify the target anomaly data, clarifying the specific type of anomaly. This ensures both rapid identification of common anomalies and accurate capture of rare and complex anomalies, while also minimizing false positive and false negative rates. Furthermore, the weighted fusion of the three core elements—rule engine determination results, target anomaly data, and anomaly type detection results—improves the scientific rigor and objectivity of risk level classification.

[0121] In S204, a health status analysis report is generated based on the health status model, abnormal risk level, and current health index.

[0122] In this embodiment, the terminal device can combine multi-source data, multi-dimensional indicators, health status models, abnormal risk levels, and current health indices to generate a health status analysis report.

[0123] Combining steps S201-S204, this embodiment filters out invalid interference data through data preprocessing, and extracts multidimensional indicators strongly correlated with robot health through feature extraction, thereby improving data quality. Then, a health status model is constructed based on these multidimensional indicators, transforming the robot's complex health status into quantifiable and comparable model parameters, overcoming the limitations of binary judgments of either good or bad, and achieving a refined and hierarchical description of the robot's health status. Next, anomaly detection and health assessment are conducted simultaneously using multi-source data, ultimately outputting the anomaly risk level and current health index, clarifying both the severity of the risk and quantifying the current health level. Finally, the three core elements—health status model, anomaly risk level, and current health index—are integrated to generate an analysis report, which includes both the overall model conclusions of robot health and specific risk classifications and quantitative indicators, enhancing the comprehensiveness, interpretability, and guidance of the health status analysis report.

[0124] In S103, fault prediction is performed on the target robot based on the health status analysis report and multi-source data to obtain the fault prediction results.

[0125] In this embodiment of the application, after the terminal device obtains the health status analysis report and multi-source data of the target robot, it can input the health status analysis report and multi-source data into the trained health status analysis model for processing to obtain the fault prediction result.

[0126] The aforementioned health status analysis model can be trained from a pre-built traditional machine learning model (such as logistic regression, random forest, etc.), a deep learning model (such as LSTM, GRU, etc.), or a lightweight model (such as decision tree, Bayesian classification, etc.).

[0127] It should be noted that the terminal device can use the target robot's historical multi-source data, health status data, and fault records as the training set, and label the fault type, fault location, and occurrence time. Then, the above model is trained through supervised learning, and the trained model is iteratively optimized using actual multi-source data to obtain the final health status analysis model.

[0128] In this embodiment, the fault prediction results include, but are not limited to, the location of the possible fault, the fault type, the time of occurrence of the fault, and the confidence level. The confidence level is used to characterize the probability that the aforementioned fault may occur.

[0129] It should be noted that the above-mentioned possible faults may be one or multiple.

[0130] In one embodiment of this application, when the multi-source data includes historical operational data and the health status analysis report includes a health status model, the terminal device can specifically use, as follows: Figure 4 Steps S401 to S402 shown implement step S103, as detailed below: In S401, the health status model and historical operating data are input into the remaining service life prediction model for processing to obtain degradation trend information of each component of the target robot.

[0131] In this embodiment, the terminal device can first extract the health status model from the health status analysis report as the model basis for component degradation analysis. This model, combined with historical operational data from multi-source data (covering the entire lifecycle degradation process of the component), is then input into the remaining service life prediction model. This remaining service life prediction model is used to perform time-series extrapolation of the fused degradation state quantities from the health status model, thereby obtaining degradation trend information for each component of the target robot. This degradation trend information includes, but is not limited to, the component's degradation rate, remaining service life, and key degradation nodes.

[0132] It should be noted that the aforementioned remaining useful life prediction model can utilize deep learning models (such as DCNN-LSTM fusion architecture, Transformer network) to capture degradation characteristics by analyzing health status models and multi-source data, thereby obtaining degradation trend information for each component.

[0133] In S402, the degradation trend information is analyzed to obtain fault prediction results.

[0134] In this embodiment, the terminal device can extract the degradation rate, remaining service life, and key degradation nodes of each component from the degradation trend information, and perform fault prediction on each component based on the degradation rate, remaining service life, and key degradation nodes to obtain fault prediction results.

[0135] Combining steps S401 to S402, this embodiment simultaneously inputs the health status model and historical operating data into the remaining service life prediction model. This not only anchors the current health status benchmark but also incorporates the patterns of historical degradation trends, enabling the accurate output of degradation trend information for each component, thereby improving the accuracy of subsequent fault prediction.

[0136] In S104, a predictive maintenance plan is generated based on the fault prediction results.

[0137] In this embodiment of the application, after obtaining the fault prediction result, the terminal device can determine the target maintenance requirement corresponding to the fault prediction result based on the fault prediction result and the pre-stored correspondence between different fault locations, different fault types and different maintenance requirements.

[0138] For example, assuming the fault prediction result is that there is an 85% probability that the No. 3 joint motor will experience an overload fault within the next 24 hours, the corresponding target maintenance requirements could be: detecting the cause of the motor load, reducing the load on the motor, and inspecting the joint transmission components of the target robot.

[0139] In some possible embodiments, the terminal device can input the aforementioned operation and maintenance requirements into a trained operation and maintenance plan generation model for processing to obtain a predicted operation and maintenance plan. The operation and maintenance plan generation model can be trained from a pre-built deep learning model.

[0140] In other possible embodiments, the terminal device can look up the predicted maintenance plan corresponding to the above maintenance requirements from a pre-built maintenance plan table.

[0141] It should be noted that the predicted maintenance plan includes, but is not limited to: the repair priority of the fault, the required parts, the respective maintenance actions, the maintenance execution entity, the execution time, and the responsible person.

[0142] In one embodiment of this application, when the fault prediction result includes multiple target fault types corresponding to each fault, the terminal device can specifically use methods such as... Figure 5 Steps S501 to S503 shown implement step S104, as detailed below: In S501, various target operation and maintenance strategies associated with fault prediction results are selected from the constructed robot knowledge graph; the robot knowledge graph includes the relationships between various robot components, fault types, and operation and maintenance strategies.

[0143] In S502, the priority of each target operation and maintenance strategy is sorted based on a genetic algorithm to obtain the task priority of each target operation and maintenance strategy.

[0144] In S503, predictive operation and maintenance plans are generated based on each target operation and maintenance strategy and the corresponding task priority.

[0145] It should be noted that the robot knowledge graph includes the relationships between the robot's various components, fault types, and operation and maintenance strategies. Specifically, the robot knowledge graph includes robot component entities, fault type entities, and operation and maintenance strategy entities, and these entities are semantically related through directed edges.

[0146] The robot component entity includes entity attributes (including component name, module, manufacturer and model, rated life, maintenance cycle, and associated components, such as [motor - joint module - model number - 10 years - 6 months - reducer and bearing]). The fault type entity includes entity attributes (including fault name, fault level, associated components, triggering cause, impact range, and fault probability), such as [abnormal motor temperature rise - moderate fault - motor - overload or poor heat dissipation - power system - 80%]). The maintenance strategy entity includes entity attributes (including strategy name, execution steps, applicable faults, compatible components, execution time, and required resources), such as [motor cooling module cleaning - disassembly → cleaning → reset - abnormal motor temperature rise - model number motor - 2 hours - wrench or high-pressure air gun]).

[0147] Semantic relationships include, but are not limited to, components and faults, faults and operation and maintenance policies, components and operation and maintenance policies, and operation and maintenance policies with each other.

[0148] The terms "Component and Fault" describe how a specific component causes a specific fault, such as "abnormal temperature rise in a motor." "Fault and Maintenance Strategy" describe how a specific fault requires a specific maintenance strategy, such as "abnormal temperature rise – heatsink module cleaning." "Component and Maintenance Strategy" describe how a specific component is adapted to a specific maintenance strategy, such as "specific motor model – heatsink module cleaning." "Maintenance Strategy and Maintenance Strategy" describe a specific strategy that is either pre-implemented or post-implemented, such as "heatsink module cleaning – motor temperature monitoring."

[0149] In this embodiment, the terminal device can filter out various target operation and maintenance strategies associated with various target fault types in the fault prediction results from the constructed robot knowledge graph.

[0150] It should be noted that the target operation and maintenance strategy includes, but is not limited to, basic strategy information, execution attribute information, and associated constraint information.

[0151] The basic strategy information includes, but is not limited to, strategy number, corresponding fault identifier, target fault type, associated component, strategy name, and core execution steps. Execution attribute information includes, but is not limited to, execution duration, maintenance costs (including labor and material costs), required resources (including tools, parts, and the execution entity), and historical execution success rate. Related constraint information includes, but is not limited to, pre-implementation maintenance strategies, post-implementation maintenance strategies, execution time windows (e.g., completion X days before fault prediction), and job interruption requirements (e.g., whether system shutdown or job suspension is required).

[0152] In this embodiment, after obtaining each target operation and maintenance strategy, in order to determine the order of execution of each target operation and maintenance strategy, the terminal device can prioritize each target operation and maintenance strategy based on a genetic algorithm to obtain the task priority of each target operation and maintenance strategy.

[0153] Specifically, the terminal device can construct a multi-objective optimization genetic algorithm model, and combine it with the various objective operation and maintenance strategies to execute an iterative process of initializing the population → fitness calculation → selection operation → crossover operation → mutation operation → constraint check → optimal retention operation → iterative convergence to obtain the task priority of each objective operation and maintenance strategy.

[0154] In this embodiment, the terminal device can generate a final predictive maintenance plan based on each target maintenance strategy and the corresponding task priority, as well as each target maintenance strategy.

[0155] Combining steps S501 to S503, this embodiment can filter out target operation and maintenance strategies that match and are highly targeted through the robot knowledge graph, without the need for manual retrieval. Then, through a genetic algorithm, optimal scheduling of operation and maintenance resources can be achieved when multiple target operation and maintenance strategies exist, minimizing the impact of faults on robot operation and controlling operation and maintenance costs. Finally, the predictive operation and maintenance plan generated based on each target operation and maintenance strategy and the corresponding task priority improves the executability and guidance of the operation and maintenance plan.

[0156] In S105, the target maintenance operation corresponding to the predicted maintenance plan is executed on the target robot.

[0157] In this embodiment of the application, after obtaining the predictive maintenance plan, the terminal device can perform the target maintenance operation corresponding to the predictive maintenance plan on the target robot, so as to realize predictive maintenance of the target robot.

[0158] Specifically, the terminal device can send the aforementioned predictive maintenance plan to the maintenance execution entity, so that the maintenance execution entity can perform the corresponding target maintenance operation on the target robot based on the predictive maintenance plan.

[0159] In one embodiment of this application, when the predicted operation and maintenance plan includes multiple target operation and maintenance strategies, the terminal device can specifically implement step S105 according to the following steps, as detailed below: The operation and maintenance strategies for each target are analyzed to obtain the operation and maintenance execution entities corresponding to each target operation and maintenance strategy; Each target operation and maintenance strategy is sent to the corresponding operation and maintenance execution entity, so that each operation and maintenance execution entity can perform the corresponding target operation and maintenance operation on the target robot.

[0160] In this embodiment, after obtaining the predicted operation and maintenance plan, the terminal device can extract multiple target operation and maintenance strategies from the plan and perform execution subject analysis on each strategy to obtain the corresponding operation and maintenance execution subject. Then, the terminal device can send each target operation and maintenance strategy to the corresponding execution subject, so that each execution subject can perform the corresponding target operation and maintenance operation on the target robot.

[0161] In the above embodiments, by analyzing the target operation and maintenance strategy, the corresponding operation and maintenance execution entity can be accurately matched, so that each operation and maintenance strategy is undertaken by an execution entity with corresponding capabilities, avoiding operational errors and task delays caused by mismatch, and improving the execution efficiency and accuracy of operation and maintenance operations.

[0162] As can be seen from the above, the robot management method provided in this application involves receiving multi-source data associated with a target robot; analyzing the multi-source data to obtain a health status analysis report for the target robot; predicting faults in the target robot based on the health status analysis report and the multi-source data to obtain fault prediction results; generating a predictive maintenance plan based on the fault prediction results; and executing target maintenance operations corresponding to the predictive maintenance plan on the target robot. Compared with the prior art, this application can analyze the robot's anomalies and health status in real time through multi-source data, thereby perceiving fine-grained health changes of machine components in real time and generating a health status analysis report. Simultaneously, it can predict potential faults, automatically generate maintenance plans, and schedule their execution, thereby reducing downtime losses and operating costs.

[0163] Please see Figure 6 , Figure 6This is a flowchart illustrating the implementation of a robot management method according to another embodiment of this application. Compared to... Figure 1 In a corresponding embodiment, this embodiment may further include steps S601 to S603 after S105, as detailed below: In S601, the work data of the target robot after completing the target operation is received.

[0164] In S602, the working data is analyzed to obtain the target operating status and target health index of the target robot.

[0165] In S603, the target's operating status and target health index are visualized.

[0166] It should be noted that the working data includes, but is not limited to, various sensor data, robot operation logs, task execution data, and historical maintenance records during the operation phase of the target robot after it has completed the target maintenance operation.

[0167] In this embodiment, after obtaining the aforementioned working data, the terminal device can analyze the working data, that is, execute step S102 based on the working data to obtain a current health status analysis report. Then, based on the multi-dimensional indicators and health scores of each indicator in the current health status analysis report, the terminal device can determine the target operating state of the target robot; simultaneously, the terminal device can determine the target health index of the target robot based on the overall health score in the current health status analysis report.

[0168] The target operating state is used to describe the degree of recovery of the target robot.

[0169] In this embodiment, the terminal device can visualize the target's operating status and target health index through a health status dashboard.

[0170] As can be seen from the above, the robot management method provided in this embodiment can quantitatively verify the actual effect of operation and maintenance by receiving and analyzing the work data after operation and maintenance, and outputting the target operating status and target health index. At the same time, the visual display of the target operating status and target health index also makes it convenient for relevant personnel to quickly and intuitively grasp the status of the robot after operation and maintenance, without having to spend time analyzing complex data reports, thereby improving the response efficiency of operation and maintenance management.

[0171] In one embodiment of this application, after step S601, the terminal device may further perform the following steps, detailed below: Anomaly analysis is performed based on working data to obtain anomaly analysis results; When the anomaly analysis result indicates the presence of an anomaly, an alarm message is output; the alarm message is used to alert the target robot that an anomaly exists.

[0172] In this embodiment, after obtaining the work data, the terminal device can perform anomaly analysis on the work data, that is, execute step S102 based on the work data to obtain a health status analysis report at this time.

[0173] Subsequently, the terminal device can obtain anomaly analysis results based on the health scores of various indicators in the current health status analysis report. These anomaly analysis results include, but are not limited to, the presence of anomalies and anomalous data.

[0174] It should be noted that when the health score of a certain dimension is not 100, it can be determined that there is an anomaly, and that dimension is identified as abnormal data; when the health scores of all dimensions are 100, it can be determined that there is no anomaly, that is, there is no abnormal data.

[0175] In this embodiment, the terminal device can output alarm information in response to an anomaly analysis result indicating the presence of an anomaly. The alarm information serves to alert the target robot to the presence of an anomaly.

[0176] Afterwards, the terminal device can send the above alarm information to the user terminal of the operation and maintenance personnel via message push, so that the operation and maintenance personnel can perform corresponding operation and maintenance on the target robot.

[0177] The message push methods include, but are not limited to: SMS, email, and applications associated with the target robot.

[0178] In the above embodiments, anomaly analysis can be performed on the work data after maintenance, thereby accurately identifying hidden problems that have not been resolved during maintenance operations and making up for the loophole of the process ending as soon as maintenance operations are completed; then, when an anomaly is detected, alarm information is immediately output to prompt relevant personnel to handle it, thereby improving the reliability of maintenance.

[0179] In another embodiment of this application, after step S503, during the execution of each target operation and maintenance strategy, the terminal device can also virtually map the operating status of the target robot based on digital twin technology to achieve remote visual diagnosis and simulation verification, and optimize each target operation and maintenance strategy according to the results of the visual diagnosis and simulation verification, thereby optimizing the operation and maintenance strategy without downtime and improving operation and maintenance efficiency.

[0180] Please see Figure 7 , Figure 7 This is an overall flowchart of a robot management method provided in one embodiment of this application. Figure 7As shown, the terminal device can receive multi-source data such as sensor data, operation logs, and stored historical maintenance records of the target robot (as in step S101). Then, the terminal device can perform status analysis on the aforementioned multi-source data (as in step S102), predict faults using a predictive model (as in step S103), and generate a plan (as in step S104). Afterward, the terminal device can achieve predictive maintenance of the target robot through scheduling execution (as in step S105).

[0181] After the above scheduling is executed, the terminal device can visualize the target robot's target operating status and target health index through a visual interface (as in steps S601~S603), and issue anomaly alarms based on the anomaly analysis results.

[0182] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0183] Corresponding to the robot management method described in the above embodiments, Figure 8 A schematic diagram of a robot management device according to an embodiment of this application is shown. For ease of explanation, only the parts relevant to the embodiment of this application are shown. (Refer to...) Figure 8 The robot management device 700 includes: a receiving unit 71, a first analysis unit 72, a fault prediction unit 73, a first generation unit 74, and a first execution unit 75. Wherein: The receiving unit 71 is used to receive multi-source data associated with the target robot.

[0184] The first analysis unit 72 is used to analyze the multi-source data to obtain a health status analysis report of the target robot.

[0185] The fault prediction unit 73 is used to predict the faults of the target robot based on the health status analysis report and the multi-source data, and obtain the fault prediction result.

[0186] The first generation unit 74 is used to generate a predictive operation and maintenance plan based on the fault prediction results.

[0187] The first execution unit 75 is used to perform target maintenance operations on the target robot that correspond to the predicted maintenance plan.

[0188] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, and they will not be repeated here.

[0189] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0190] Figure 9 This is a schematic diagram of the structure of a terminal device provided in an embodiment of this application. Figure 9 As shown, the terminal device 8 of this embodiment includes: at least one processor 80 ( Figure 9 (Only one is shown in the diagram), memory 81, and computer program 82 stored in said memory 81 and executable on said at least one processor 80, wherein said processor 80 executes said computer program 82 to implement the steps in any of the above embodiments of robot management methods.

[0191] The terminal device 8 may include, but is not limited to, a processor 80 and a memory 81. Those skilled in the art will understand that... Figure 9 This is merely an example of terminal device 8 and does not constitute a limitation on terminal device 8. It may include more or fewer components than shown in the figure, or combine certain components, or different components, such as input / output devices, network access devices, etc.

[0192] The processor 80 may be a Central Processing Unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0193] In some embodiments, the memory 81 may be an internal storage unit of the terminal device 8, such as the RAM of the terminal device 8. In other embodiments, the memory 81 may be an external storage device of the terminal device 8, such as a plug-in hard drive, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the terminal device 8. Furthermore, the memory 81 may include both internal and external storage units of the terminal device 8. The memory 81 is used to store the operating system, applications, bootloader, data, and other programs, such as the program code of the computer program. The memory 81 can also be used to temporarily store data that has been output or will be output.

[0194] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps described in the various method embodiments above.

[0195] This application provides a computer program product that, when run on a terminal device, enables the terminal device to implement the steps described in the above-described method embodiments.

[0196] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or device capable of carrying computer program code to a terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks.

[0197] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0198] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A robot management method, characterized in that, include: Receive multi-source data associated with the target robot; The health status analysis report of the target robot is obtained by analyzing the multi-source data. Based on the health status analysis report and the multi-source data, fault prediction is performed on the target robot to obtain fault prediction results; Based on the fault prediction results, a predictive operation and maintenance plan is generated; Perform the target maintenance operation corresponding to the predicted maintenance plan on the target robot.

2. The robot management method as described in claim 1, characterized in that, The analysis of the multi-source data to obtain the health status analysis report of the target robot includes: Data preprocessing and feature extraction are performed on the multi-source data to obtain multi-dimensional indicators; The state of the target robot is modeled based on the multidimensional indicators to obtain the health status model of the target robot; Based on the aforementioned multidimensional indicators, anomaly detection and health assessment are performed on the target robot to obtain the anomaly risk level and current health index of the target robot. Based on the health status model, the abnormal risk level, and the current health index, the health status analysis report is generated.

3. The robot management method as described in claim 2, characterized in that, The process of modeling the state of the target robot based on the multidimensional indicators to obtain a health state model of the target robot includes: Based on the mechanical equations and the multidimensional indices, a physical flow model is established between the degradation state variables corresponding to each component of the target robot and the multidimensional indices; the physical flow model is a continuous-time state-space model. Based on the multi-source data, the residual between the predicted value and the corresponding actual value is obtained; The pre-constructed neural differential equation is trained based on the multidimensional index, the residual, and the degenerate state quantity to obtain a data flow model; The physical flow model and the data flow model are fused together to obtain the health status model.

4. The robot management method as described in claim 2, characterized in that, The abnormal risk level is obtained according to the following method: Anomaly detection is performed on the multi-source data based on the target rule engine to obtain anomaly detection results; The multi-source data is input into the anomaly identification model for anomaly identification to obtain the target anomaly data; The target abnormal data is input into the abnormality type detection model for type detection to obtain the abnormality detection result; The anomaly determination result, the target anomaly data, and the anomaly detection result are weighted and fused to obtain the anomaly risk level.

5. The robot management method according to any one of claims 1-4, characterized in that, The multi-source data includes historical operational data, and the health status analysis report includes a health status model; the step of performing fault prediction on the target robot based on the health status analysis report and the multi-source data to obtain fault prediction results includes: The health status model and the historical operating data are input into the remaining service life prediction model for processing to obtain the degradation trend information of each component of the target robot. The degradation trend information is analyzed to obtain the fault prediction result.

6. The robot management method according to any one of claims 1-4, characterized in that, The fault prediction results include multiple target fault types corresponding to each fault. The step of generating a predictive maintenance plan based on the fault prediction results includes: Each target operation and maintenance strategy associated with the fault prediction result is selected from the constructed robot knowledge graph; the robot knowledge graph includes the relationships between the robot's various components, fault types, and operation and maintenance strategies. The priority of each target operation and maintenance strategy is obtained by prioritizing each target operation and maintenance strategy based on a genetic algorithm. Based on each of the target operation and maintenance strategies and the corresponding task priorities, the predictive operation and maintenance plan is generated.

7. The robot management method according to any one of claims 1-4, characterized in that, The predictive maintenance plan includes multiple target maintenance strategies. The step of executing the target maintenance operation corresponding to the predictive maintenance plan on the target robot includes: Analyze each of the target operation and maintenance strategies to obtain the operation and maintenance execution entity corresponding to each target operation and maintenance strategy; Each of the target operation and maintenance strategies is sent to the corresponding operation and maintenance execution entity, so that each operation and maintenance execution entity performs the corresponding target operation and maintenance operation on the target robot.

8. The robot management method according to any one of claims 1-4, characterized in that, After performing the target maintenance operation corresponding to the predicted maintenance plan on the target robot, the method further includes: Receive the work data of the target robot after it has completed the target maintenance operation; The target operating status and target health index of the target robot are obtained by analyzing the working data. The target's operating status and health index are displayed visually.

9. The robot management method as described in claim 8, characterized in that, After receiving the work data of the target robot after completing the target maintenance operation, the method further includes: Anomaly analysis is performed based on the aforementioned working data to obtain anomaly analysis results; When the anomaly analysis result indicates the presence of an anomaly, an alarm message is output; the alarm message is used to indicate that the target robot has an anomaly.

10. A robot management device, characterized in that, include: The receiving unit is used to receive multi-source data associated with the target robot; The first analysis unit is used to analyze the multi-source data to obtain a health status analysis report of the target robot. The fault prediction unit is used to predict the fault of the target robot based on the health status analysis report and the multi-source data, and obtain the fault prediction result. The first generation unit is used to generate a predictive operation and maintenance plan based on the fault prediction results; The first execution unit is used to perform target maintenance operations on the target robot that correspond to the predicted maintenance plan.

11. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the robot management method as described in any one of claims 1 to 9.

12. A computer program product, characterized in that, It includes a computer program that, when run, implements the robot management method as described in any one of claims 1 to 9.

Images