A data acquisition system and method

Abstract

The present application relates to the field of industrial robot monitoring and fault diagnosis, and particularly relates to a data acquisition system and method, the present application obtains the running data of the corresponding controller of the industrial robot under the conventional monitoring acquisition frequency, filters the suspicious running time domain segment, extracts the corresponding execution job of the suspicious running time domain segment, evaluates the running abnormality representation value of the industrial robot based on the job response characteristics of the execution job, and divides the corresponding abnormality level, and adaptively adjusts the monitoring acquisition frequency of the corresponding controller of the industrial robot and the subsequent depth analysis based on the abnormality level. The present application can reduce the conventional monitoring overhead while improving the efficiency and accuracy of the abnormality detection and depth diagnosis of the industrial robot based on the adaptive adjustment of the monitoring acquisition frequency, the job execution chain analysis and the log sequence matching.

Description

Technical Field

[0001] This invention relates to the field of industrial robot monitoring and fault diagnosis, and in particular to a data acquisition system and method. Background Technology

[0002] Industrial robots are core execution units in fields such as intelligent manufacturing, automated assembly, and precision machining. To ensure their long-term, stable, and high-precision operation, continuous monitoring of their operating status is usually required, and data acquisition is a fundamental component supporting robot monitoring and fault diagnosis systems.

[0003] During the operation of industrial robots, the monitoring system relies on the controller to periodically acquire operational data according to a set acquisition frequency to reflect the robot's dynamic behavior and health status, thereby providing a basis for identifying anomalies or faults. As the complexity and real-time requirements of robot operations continue to increase, monitoring and data acquisition schemes need to strike a balance between information integrity and system overhead. On the one hand, too low an acquisition frequency may miss transient anomalies or early warning signs of faults, weakening the timeliness and effectiveness of monitoring; on the other hand, too high an acquisition frequency will generate massive amounts of data, placing a heavy burden on storage, communication, and subsequent analysis and processing, and even affecting the real-time performance of the monitoring system itself. Therefore, how to optimize the acquisition strategy while ensuring monitoring effectiveness has become a common key consideration in the design of industrial robot monitoring and data acquisition methods.

[0004] Furthermore, different work nodes often have different requirements for the granularity of monitoring data collection, and abnormal behaviors are usually closely related to specific work nodes or work stages. Therefore, incorporating an understanding of the work context into monitoring and data collection methods helps to improve the targeting of data collection and the efficiency of data utilization, thereby enhancing the overall intelligence level of the monitoring system.

[0005] For example, Chinese Patent Publication No. CN121300331A discloses a robot remote adaptive data acquisition control system and method based on dynamic value assessment. This system calculates the dynamic information value factor of data points in real time through a processing unit located locally on the robot. This factor integrates data change rate, uncertainty, task context, and anomaly probability. Based on this factor, the system dynamically adjusts the sampling frequency, compression strategy, and transmission priority. A remote center is responsible for data reconstruction and fusion, and uses machine learning closed-loop optimization of acquisition parameters. This invention solves the problems of resource waste and missing key information caused by fixed-frequency sampling, achieving an optimal balance between data acquisition efficiency and quality under limited bandwidth.

[0006] However, the following problems still exist in the existing technology. The robot lacks adaptive learning capabilities and cannot dynamically adjust the monitoring and diagnostic strategies based on its operating status. Summary of the Invention

[0007] To address this, the present invention provides a data acquisition system and method to overcome the problems of low utilization of monitoring resources and poor accuracy of fault diagnosis in the prior art.

[0008] To achieve the above objectives, the present invention provides a data acquisition method, comprising: Acquire the operating data of the industrial robot's corresponding controller at the regular monitoring and acquisition frequency, and filter out suspicious runtime domain segments; Extract the execution jobs corresponding to the suspicious runtime domain segments, and evaluate the operation anomaly characterization values ​​of the industrial robot based on the job response characteristics of the execution jobs, so as to classify the anomaly level of the industrial robot; Based on the aforementioned anomaly level, adjust the monitoring and data acquisition frequency for the corresponding controller of the industrial robot, including: If the industrial robot is classified as having a high anomaly level, the monitoring and data collection frequency for the corresponding controller of the industrial robot will be increased, and in-depth diagnostic analysis of the industrial robot will be performed, including: Extract the job execution chain of the industrial robot within the suspected runtime domain segment and assess whether it has job continuity; Extract the data fluctuation characteristics and time interval uniformity corresponding to the abnormal operation nodes, calculate the operation anomaly characterization value, and narrow down the suspicious runtime domain segment; Based on the log sequence within the deviation time domain, the similarity of the log sequence is matched with the fault mode library, and the fault type corresponding to the industrial robot is output. If the industrial robot is at a low anomaly level, the monitoring and data collection frequency for the corresponding controller of the industrial robot will be maintained.

[0009] Furthermore, the process of filtering suspicious runtime domain segments includes: Construct the time-domain variation curve of the joint torque of the industrial robot; Extract the starting point of the sharp increase in torque and the adjacent starting point of the decrease to lock the time domain segment of abnormal fluctuation; The torque fluctuation is determined based on the change in torque increase in each sub-time domain segment within the fluctuating anomaly time domain segment. If the torque fluctuation is greater than or equal to the torque fluctuation threshold, then the abnormal fluctuation time domain segment is selected as the suspicious runtime domain segment. The torque fluctuation is determined based on the statistical variance of the change in torque increase.

[0010] Furthermore, the process of evaluating the abnormal operation characteristics of the industrial robot includes: The ratio of the deviation value of the trajectory tracking error to the error deviation threshold is used as the first abnormal operation feature value; The ratio of the number of instruction response timeouts to the instruction response timeout threshold is used as the second operational anomaly characteristic value; The weighted sum of the first operational anomaly feature value and the second operational anomaly feature value is determined as the operational anomaly characterization value. The operation response characteristics include the deviation value of the trajectory tracking error and the number of command response timeouts.

[0011] Furthermore, the abnormality levels of the industrial robot are classified, including: If the abnormal operation characteristic value of an industrial robot is greater than or equal to the abnormal operation characteristic threshold, the industrial robot is classified as a high abnormality level. If the abnormal operation characteristic value of the industrial robot is less than the abnormal operation characteristic threshold, the industrial robot will be classified as a low abnormality level.

[0012] Furthermore, the process of assessing whether there is operational continuity includes: If an industrial robot meets the conditions for sequential operation, then it is determined that it has operational continuity. The timing coherence condition includes that the number of consecutive errors of the executed job node is less than the consecutive error number threshold, and the baseline overlap rate of the executed job node is greater than the baseline overlap rate threshold.

[0013] Furthermore, the process of identifying abnormal operation nodes includes: Compare the job execution chain with the standard execution chain and calculate the longest matching step sequence; Based on the longest matching step sequence, the execution job nodes in the job execution chain are time-aligned with the standard job nodes in the standard execution chain to establish matching node pairs; If an abnormal situation exists, the corresponding execution job node will be identified as the abnormal job node; The abnormal situations include missing abnormalities, redundancy abnormalities, sequence abnormalities, duplication abnormalities, and duration abnormalities.

[0014] Furthermore, the process of calculating the anomaly characterization value includes: The ratio of the time interval uniformity of abnormal operation nodes to the time interval uniformity threshold is used as the first operation abnormality feature value. The sum of the ratio of the deviation in execution time to the deviation threshold and the ratio of the corresponding torque oscillation amplitude to the torque oscillation amplitude threshold is used as the second abnormal operation characteristic value. The first job anomaly feature value and the second job anomaly feature value are weighted and summed to determine the job anomaly characterization value; The data fluctuation characteristics include the deviation of execution time and the torque oscillation amplitude.

[0015] Furthermore, the suspected runtime domain segment is narrowed down, including: If the job anomaly characterization value is greater than or equal to the job anomaly characterization threshold, then the suspected runtime domain segment is restricted. The time domain segment defined by the start timestamp of the first abnormal operation node as the starting point and the end timestamp of the last abnormal operation node as the ending point is determined as the deviation time domain segment.

[0016] Furthermore, the process of matching and outputting the fault type corresponding to the industrial robot includes: Logs within the deviation time domain are sorted chronologically and matched with samples in the fault mode library; The similarity is determined by coupling the occurrence order matching degree of adjacent log pairs with the causal semantic connectivity degree of adjacent logs. If the similarity is greater than or equal to the similarity threshold, the corresponding fault type in the fault mode library is matched and output as the fault type of the industrial robot.

[0017] Furthermore, systems for applying data acquisition methods are also provided, including, The time domain extraction module is used to obtain the operating data of the corresponding controller of the industrial robot under the normal monitoring and acquisition frequency, and to filter out suspicious runtime domain segments. An anomaly assessment module is used to extract the execution jobs corresponding to the suspicious runtime domain segments, and to assess the anomaly characterization value of the industrial robot based on the job response characteristics of the execution jobs, so as to classify the anomaly level of the industrial robot. The adjustment analysis module is used to adjust the monitoring and data acquisition frequency for the corresponding controller of the industrial robot based on the anomaly level, including: If the industrial robot is classified as having a high anomaly level, the monitoring and data collection frequency for the corresponding controller of the industrial robot will be increased, and in-depth diagnostic analysis of the industrial robot will be performed, including: Extract the job execution chain of the industrial robot within the suspected runtime domain segment and assess whether it has job continuity; Extract the data fluctuation characteristics and time interval uniformity corresponding to the abnormal operation nodes, calculate the operation anomaly characterization value, and narrow down the suspicious runtime domain segment; Based on the log sequence within the deviation time domain, the similarity of the log sequence is matched with the fault mode library, and the fault type corresponding to the industrial robot is output. If the industrial robot is at a low anomaly level, the monitoring and data collection frequency for the corresponding controller of the industrial robot will be maintained.

[0018] Compared with existing technologies, this invention sets up a time-domain extraction module, an anomaly assessment module, and an adjustment analysis module. The time-domain extraction module acquires the operational data of the industrial robot's corresponding controller at the normal monitoring and acquisition frequency, filtering out suspicious runtime domain segments. The anomaly assessment module extracts the execution tasks corresponding to the suspicious runtime domain segments, and based on the task response characteristics, evaluates the anomaly characterization values ​​of the industrial robot's operation to classify the corresponding anomaly levels. The adjustment analysis module adaptively adjusts the monitoring and acquisition frequency and subsequent in-depth analysis for the industrial robot's corresponding controller based on the anomaly level. This invention, based on adaptive adjustment of the monitoring and acquisition frequency, combined with task execution chain analysis and log sequence matching, can improve the efficiency and accuracy of industrial robot anomaly detection and in-depth diagnosis while reducing the overhead of conventional monitoring.

[0019] In particular, this invention filters time-domain intervals that may contain abnormal operating behaviors, captures transient conflict anomalies, and effectively identifies short-term, high-amplitude impact anomalies by analyzing the sharp rise and adjacent fall points of joint torque. Furthermore, it calculates torque fluctuations based on the statistical variance of the torque increase change to reflect the severity and dispersion of the fluctuations, distinguishing between normal load fluctuations and genuine abnormal oscillations. This provides data support for subsequent targeted anomaly assessment and in-depth diagnosis, thereby improving the efficiency of overall monitoring and fault diagnosis.

[0020] In particular, this invention further extracts the corresponding execution information of the industrial robot for the suspicious runtime domain segments identified in the previous screening, and evaluates the anomaly characterization value based on key response features to more comprehensively and accurately reflect the real-time health status of the industrial robot. In terms of execution accuracy, the deviation value of the trajectory tracking error reflects the degree of deviation between the actual pose and the reference pose of the industrial robot when executing the corresponding trajectory. The larger the value of this indicator, the worse the motion control accuracy of the industrial robot, which may correspond to abnormal states such as mechanical wear, overload, or joint transmission failure. This characterizes the accuracy of the industrial robot's motion control and its mechanical dynamic response capability. In terms of real-time response, the number of times the time from command issuance to industrial robot response execution exceeds the reference time is quantified by the number of command response timeouts. The higher the value of this indicator, the more frequent the delay or loss of control commands during the industrial robot's reception and processing, and the worse the real-time performance and reliability of the communication link between the controller and the execution system or the task scheduling system, which may indicate anomalies such as bus conflicts, controller overload, communication interference, or drive response blocking. Furthermore, based on the aforementioned two dimensions of characteristics, the anomaly assessment values ​​comprehensively characterize the overall operational health status and severity of anomalies of industrial robots during operation, providing data support for subsequent classification of anomaly levels. Moreover, it provides a clear grading basis for subsequent adaptive monitoring and in-depth diagnostics, thereby enabling on-demand allocation of monitoring resources, avoiding resource waste caused by indiscriminate high-frequency data collection, and improving the targeting and efficiency of fault diagnosis.

[0021] In particular, this invention supplements the anomaly detection at the node sequence level of the execution task. By analyzing the correctness of the execution task node sequence and the overall matching degree, it identifies task-level anomalies such as missing, redundant, disordered sequence, and repeated execution. In terms of the logical correctness of the task sequence, the number of consecutive errors in the execution task node sequence quantifies the number of times the order of two adjacent actually executed task nodes in the task execution chain is inconsistent with the order of corresponding nodes in the standard execution chain. When a reversed order, unexpected jump, or repetition occurs, it is counted as one consecutive error. The more consecutive errors, the more severe the program logic anomaly, state transition disorder, or instruction sequence misalignment. In terms of the overall sequence matching integrity, the baseline overlap rate of the executed task nodes reflects the overall matching degree between the actual executed node sequence and the standard node sequence, i.e., the completion coverage rate of correctly executed nodes in the actual task, used to determine whether there are missing or redundant steps. The lower the baseline overlap rate, the more anomalies such as missing nodes, redundant insertions, or serious misalignments exist. This quantifies the severity of problems such as incomplete task execution, omission of key steps, or chaotic task logic. This avoids unnecessary in-depth diagnostics triggered by transient interference during normal operations, thus improving the system's anti-interference capability. Furthermore, the evaluation results provide crucial screening criteria for subsequent abnormal operation node localization and time-domain segment limitation, ensuring that in-depth diagnostics only target intervals where operational logic problems truly exist, thereby improving the overall efficiency and accuracy of fault diagnosis.

[0022] In particular, this invention precisely narrows down suspicious runtime domain segments. In terms of time consumption, the deviation of the execution time of abnormal work nodes reflects whether the time consumed from start to finish deviates from the normal baseline, reflecting the timeliness and smoothness of work execution. A larger value indicates possible slow execution due to jamming, increased resistance, or instruction blocking; a smaller value indicates possible abnormalities such as nodes being skipped or prematurely terminated. In terms of torque fluctuation, the amplitude of torque oscillation reflects the severity of joint torque fluctuations during work node execution, reflecting the stability of force control and the smoothness of mechanical transmission. A larger amplitude indicates more severe torque fluctuations, potentially indicating vibration, impact, frictional abrupt changes, or gap collisions. In terms of event occurrence rhythm, the uniformity of time intervals reflects whether the distribution of multiple abnormal work nodes on the time axis is uniform, reflecting the regularity and periodicity of abnormal events. Furthermore, this invention quantifies the severity and distribution patterns of abnormal operation nodes using the aforementioned three features, thereby narrowing down the suspicious runtime segments and eliminating redundant intervals within these segments that lack abnormal behavior. This makes the data window for subsequent fault mode matching more focused, improving the signal-to-noise ratio. In addition, the introduction of time interval uniformity enhances the ability to distinguish between periodic weak anomalies and sudden faults, further improving the sensitivity and accuracy of fault diagnosis.

[0023] In particular, this invention improves the accuracy and robustness of fault type identification by comprehensively measuring matching similarity by evaluating the sequential consistency and causal logic rationality of the corresponding sequences of logs generated by industrial robots performing operations. From a sequential logic perspective, the occurrence order matching degree of adjacent log pairs reflects whether the order of occurrence of two adjacent logs in the actual log sequence is consistent with the order of adjacent logs of the corresponding fault samples in the fault pattern library, i.e., whether the temporal arrangement of the logs is correct. A larger value indicates a closer match between the event occurrence order of the actual logs and the order of the sample fault patterns, used to eliminate erroneous matches caused by log order reversal. From a causal logic perspective, the causal semantic coherence degree of adjacent logs reflects whether there is a logical causal relationship between two adjacent logs, i.e., whether the occurrence of the previous log has a high probability of historically leading to the occurrence of the subsequent log. A higher value of this feature indicates a stronger causal dependency between the two logs in the fault propagation chain, such as motor overheating-protection tripping, which can avoid misjudging accidental co-occurrence as a fault. Furthermore, the similarity between the actual log sequence and the samples in the fault pattern library is quantified by coupling the above two features. Within the time domain of precise positioning deviation, log sequences are used to automatically identify the most likely specific fault type of the industrial robot, thereby improving the automation level and accuracy of fault diagnosis and shortening downtime for troubleshooting. Attached Figure Description

[0024] Figure 1 This is a schematic diagram illustrating the steps of a data acquisition method according to an embodiment of the invention; Figure 2 A logic block diagram for classifying the anomaly levels of an industrial robot according to an embodiment of the invention; Figure 3 A logic block diagram for evaluating whether there is operational coherence in an embodiment of the invention; Figure 4 This is a logic block diagram illustrating the restriction of suspicious runtime domain segments according to an embodiment of the invention. Detailed Implementation

[0025] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0026] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0027] Please see Figure 1 The diagram illustrates the steps of a data acquisition method according to an embodiment of the invention. The data acquisition method of this embodiment includes: Step S1: Obtain the operating data of the industrial robot's corresponding controller at the normal monitoring and acquisition frequency, and filter out suspicious runtime domain segments; Step S2: Extract the execution job corresponding to the suspicious runtime domain segment, and evaluate the operation anomaly characterization value of the industrial robot based on the job response characteristics of the execution job, so as to classify the anomaly level of the industrial robot. Step S3: Based on the anomaly level, adjust the monitoring and data acquisition frequency for the corresponding controller of the industrial robot, including: If the industrial robot is classified as having a high anomaly level, the monitoring and data collection frequency for the corresponding controller of the industrial robot will be increased, and in-depth diagnostic analysis of the industrial robot will be performed, including: Extract the job execution chain of the industrial robot within the suspected runtime domain segment and assess whether it has job continuity; Extract the data fluctuation characteristics and time interval uniformity corresponding to the abnormal operation nodes, calculate the operation anomaly characterization value, and narrow down the suspicious runtime domain segment; Based on the log sequence within the deviation time domain, the similarity of the log sequence is matched with the fault mode library, and the fault type corresponding to the industrial robot is output. If the industrial robot is at a low anomaly level, the monitoring and data collection frequency for the corresponding controller of the industrial robot will be maintained.

[0028] Specifically, the job execution chain refers to a sequence of job nodes executed sequentially in chronological order during the execution of a complete job task by an industrial robot. Furthermore, the job execution chain records the complete process of a job task from start to finish, serving as the data foundation for subsequent time-series alignment with standard execution chains, calculation of the longest matching step sequence, and identification of anomalies such as missing, redundant, sequential, repetitive, and duration-related errors.

[0029] Specifically, the operational data includes torque increase, operational response characteristics, number of consecutive errors, benchmark overlap rate, data fluctuation characteristics, time interval uniformity, occurrence order matching degree, and causal semantic coherence degree, etc.

[0030] Specifically, the controller can collect and record several relevant data, such as joint torque values, trajectory tracking errors, relevant timestamps, and system log records, to provide basic support for determining the relevant characteristics corresponding to the operating data.

[0031] In this embodiment, the monitoring and acquisition frequency of the controller corresponding to the industrial robot is increased, and the increase is dynamically determined based on the magnitude of the abnormal operation characteristics. Multiple abnormal operation characteristic value ranges are pre-defined, each range corresponding to a acquisition frequency increase coefficient, which is a multiple of the conventional monitoring and acquisition frequency. The higher the abnormal operation characteristic value, the larger the corresponding increase coefficient, and the greater the increase in acquisition frequency. For example, when the abnormal operation characteristic value is in (S0, 1.5S0], the increase coefficient is set to 2 times; when it is in (1.5S0, 2S0], the increase coefficient is set to 5 times; and when it exceeds 2S0, the acquisition frequency is increased to the highest sampling frequency supported by the controller. Furthermore, the specific values ​​of the range division and the increase coefficient can be dynamically adjusted according to the robot model, operating conditions, and historical fault data.

[0032] Specifically, the process of filtering suspicious runtime domain segments includes: Construct the time-domain variation curve of the joint torque of the industrial robot; Extract the starting point of the sharp increase in torque and the adjacent starting point of the decrease to lock the time domain segment of abnormal fluctuation; The torque fluctuation is determined based on the change in torque increase in each sub-time domain segment within the fluctuating anomaly time domain segment. If the torque fluctuation is greater than or equal to the torque fluctuation threshold, then the abnormal fluctuation time domain segment is selected as the suspicious runtime domain segment. The torque fluctuation is determined based on the statistical variance of the change in torque increase.

[0033] In this embodiment, the purpose of setting a torque fluctuation threshold is to identify suspicious runtime segments in which the joint torque of the industrial robot exhibits abnormal fluctuations and requires further screening. Considering that during normal operation, joint torque will experience random fluctuations due to load changes and acceleration / deceleration, setting the threshold too low would lead to a large number of normal operating periods being misjudged as suspicious, increasing the burden of unnecessary subsequent analysis. Therefore, the threshold is set higher than the average torque fluctuation during historical normal operation to reduce sensitivity, triggering a suspicious determination only when the torque fluctuation significantly exceeds the normal level.

[0034] Specifically, by acquiring torque fluctuation data of the industrial robot over several historical normal operating cycles, the average torque fluctuation is calculated as a benchmark value. Based on the purpose of setting the torque fluctuation threshold, the torque fluctuation threshold is determined to be the product of the average torque fluctuation and the torque deviation coefficient.

[0035] The torque deviation coefficient is selected within the range [1.1, 1.3]. During normal operation, the torque fluctuation is small and stable. A coefficient slightly higher than the average value can effectively eliminate normal load fluctuations while retaining the ability to identify truly abnormal fluctuations, such as impacts and jamming. In practice, a value of 1.2 is preferred.

[0036] The selection of the above-mentioned preferred values ​​balances the timeliness and reliability of identification, achieving a good balance between filtering normal fluctuations and accurately identifying suspicious periods. Furthermore, those skilled in the art can readjust the value range or adjust the torque deviation coefficient within a given range according to the operating load characteristics of the industrial robot and the required sensitivity for anomaly detection.

[0037] Specifically, the process of constructing the time-domain variation curve of the joint torque of an industrial robot includes: Construct a rectangular coordinate system with time as the horizontal axis and joint torque as the vertical axis; The joint torque coordinate points at each moment are marked in the rectangular coordinate system; Connect the coordinate points with a smooth curve to obtain the time-domain variation curve of the joint torque.

[0038] Specifically, there are no restrictions on the method for constructing the time-domain variation curve of joint torque. For example, the time-domain curve can be fitted using Matlab, Python, or other software with curve fitting capabilities, which will not be elaborated further.

[0039] Specifically, the moment when the upward slope of torque over time first exceeds the upward slope threshold is defined as the starting point of a sharp increase in torque; the moment when the torque immediately following its peak begins to decrease, and the downward slope first exceeds the downward slope threshold, is defined as the starting point of an adjacent decrease. The downward slope is taken as its absolute value.

[0040] The determination of the rising slope threshold and the falling slope threshold can be achieved by collecting joint torque data from multiple runs of the industrial robot under normal operating conditions, calculating the maximum slope of the torque rising segment during each normal operation, and taking the sum of the mean of the maximum slopes and 2 to 3 times the standard deviation as the rising slope threshold. Similarly, the sum of the mean of the minimum slopes of the torque falling segment and 2 to 3 times the standard deviation is calculated as the falling slope threshold. Furthermore, those skilled in the art can adaptively adjust the values ​​of the rising slope threshold and the falling slope threshold according to the robot's operational stability, historical data volume, and real-time requirements.

[0041] Specifically, the process of evaluating the abnormal operation characteristics of the industrial robot includes: The ratio of the deviation value of the trajectory tracking error to the error deviation threshold is used as the first abnormal operation feature value; The ratio of the number of instruction response timeouts to the instruction response timeout threshold is used as the second operational anomaly characteristic value; The weighted sum of the first operational anomaly feature value and the second operational anomaly feature value is determined as the operational anomaly characterization value. The operation response characteristics include the deviation value of the trajectory tracking error and the number of command response timeouts.

[0042] Specifically, the deviation value of trajectory tracking error directly reflects the motion control accuracy and operation quality of industrial robots, and is a core indicator for measuring execution performance. It encompasses various fault types such as mechanical wear, servo mismatch, and overload, exhibiting high fault sensitivity and continuous monitoring value. In contrast, the number of command response timeouts primarily reflects communication or controller load issues, with a relatively limited impact. Therefore, in the weighted summation, a higher weight coefficient is assigned to the first operational anomaly characteristic value to capture key anomalies affecting operation quality earlier and more accurately, while also considering system real-time risks. This coefficient is selected within the range of 0.6 to 0.8, and is set to 0.7 in this embodiment. Correspondingly, the weight coefficient for the second operational anomaly characteristic value is selected within the range of 0.2 to 0.4, and is set to 0.3 in this embodiment. Furthermore, the specific values ​​can be adjusted based on the specific robot model, operating conditions, and historical fault data statistics, or adaptively optimized based on real-time feedback during system operation.

[0043] In this embodiment, the purpose of setting the error deviation threshold and the command response timeout threshold is to characterize situations where the overall operational health of the industrial robot is poor and the severity of anomalies is high during the execution of tasks. Considering that during normal operation, the deviation value of trajectory tracking error may fluctuate slightly, and the number of command response timeouts may also occur sporadically due to communication interference, setting the thresholds too low would lead to a large number of normal robots being misjudged as having high anomaly levels, triggering unnecessary high-frequency monitoring and in-depth diagnostics, and wasting system resources. Therefore, each threshold is set to an appropriate multiple of the historical average of the corresponding feature quantity to reduce sensitivity, triggering a high anomaly level judgment only when the anomaly level significantly exceeds the normal level.

[0044] Specifically, by acquiring deviation data of trajectory tracking errors and command response timeout counts from several historical normal operation cycles, the mean error deviation and the mean command response timeout counts are calculated as benchmark values. Based on the purpose of setting the above two thresholds, the error deviation threshold is determined as the product of the mean error deviation and the error deviation coefficient, and the command response timeout count threshold is determined as the product of the mean command response timeout counts and the count deviation coefficient.

[0045] The error deviation coefficient is selected within the interval [1.1, 1.3]. Under normal operating conditions, the deviation value of the trajectory tracking error fluctuates relatively little, and a coefficient slightly higher than the mean can effectively filter out random errors. In practice, a coefficient of 1.2 is preferred.

[0046] The deviation coefficient for the number of times is selected within the range [1.2, 1.4]. The number of command response timeouts fluctuates greatly due to interference from the communication environment, requiring a larger amplification factor to distinguish between occasional timeouts and systematic response problems. In practice, a value of 1.3 is preferred.

[0047] The selection of the above-mentioned preferred values ​​balances the timeliness and reliability of identification, achieving a good balance between effectively filtering normal fluctuations and accurately identifying high anomalies. Furthermore, the value range can be readjusted or the coefficients adjusted within a given range based on the operational accuracy requirements of the industrial robot and the stability of the communication environment.

[0048] Specifically, please refer to Figure 2 As shown, it is a logic block diagram for classifying the anomaly levels of an industrial robot according to an embodiment of the present invention. Classifying the anomaly levels of the industrial robot includes: If the abnormal operation characteristic value of an industrial robot is greater than or equal to the abnormal operation characteristic threshold, the industrial robot is classified as a high abnormality level. If the abnormal operation characteristic value of the industrial robot is less than the abnormal operation characteristic threshold, the industrial robot will be classified as a low abnormality level.

[0049] To determine the anomaly characterization threshold S0, the anomaly characterization values ​​for each cycle are extracted from relevant data of the same industrial robot over several normal operating cycles. The mean and standard deviation of the sequence are calculated, and the sum of the product of the standard deviation and the sensitivity coefficient k and the mean is used as the anomaly characterization threshold. Furthermore, the anomaly characterization threshold should be dynamically updated periodically with the industrial robot's operating time to adapt to mechanical wear and controller characteristic drift. Simultaneously, the anomaly characterization threshold should be coordinated with the error deviation threshold and the command response timeout threshold, and can be iteratively optimized by verifying the misjudgment ratio of normal samples.

[0050] The sensitivity coefficient k can be selected within the range [1.5, 3.0] based on the tolerance for missed detections of high anomalies; for example, k = 2.0. If sufficient historical data is available, the 95th percentile can also be used as the threshold.

[0051] Specifically, please refer to Figure 3 The diagram shown is a logical block diagram for evaluating whether there is operational continuity according to an embodiment of the present invention. The process for evaluating whether there is operational continuity includes: If an industrial robot meets the conditions for sequential operation, then it is determined that it has operational continuity. The timing coherence condition includes that the number of consecutive errors of the executed job node is less than the consecutive error number threshold, and the baseline overlap rate of the executed job node is greater than the baseline overlap rate threshold.

[0052] In this embodiment, the purpose of setting the threshold for the number of continuity errors and the baseline overlap rate threshold is to characterize situations where the sequential correctness and overall matching degree of the executed job nodes are low. To avoid judging occasional minor errors during normal job execution as inconsistencies, the thresholds are shifted in a direction that makes it difficult to meet the consistency conditions, reducing sensitivity, and inconsistency judgment is only triggered when there is a significant abnormality in the job logic.

[0053] Specifically, by acquiring data on the number of continuity errors and the baseline overlap rate of the industrial robot over several historical normal operating cycles, the average number of continuity errors and the average baseline overlap rate are calculated as baseline values. Based on the purpose of setting the above two thresholds, the threshold for the number of continuity errors is determined as the product of the average number of continuity errors and the error deviation coefficient, and the threshold for the baseline overlap rate is determined as the product of the average baseline overlap rate and the overlap deviation coefficient.

[0054] The error deviation coefficient is selected within the interval [1.1, 1.3]. During normal operation, the number of consecutive errors is small and stable, and a coefficient slightly higher than the average can eliminate occasional sequence disorder. In practice, a coefficient of 1.2 is preferred.

[0055] The overlap deviation coefficient is selected within the range [0.85, 0.95], and a threshold slightly lower than the mean value allows for minor deviations during normal operation. In practice, a value of 0.9 is preferred.

[0056] The selection of the above-mentioned optimal values ​​balances the timeliness and reliability of identification, achieving a good balance between effectively filtering normal fluctuations and accurately identifying true operational continuity issues. Furthermore, those skilled in the art can readjust the value range or adjust the coefficients within a given range based on the operational complexity of the industrial robot and the sensitivity requirements for determining operational continuity.

[0057] Specifically, the process of identifying abnormal job nodes includes: Compare the job execution chain with the standard execution chain and calculate the longest matching step sequence; Based on the longest matching step sequence, the execution job nodes in the job execution chain are time-aligned with the standard job nodes in the standard execution chain to establish matching node pairs; If an abnormal situation exists, the corresponding execution job node will be identified as the abnormal job node; The abnormal situations include missing abnormalities, redundancy abnormalities, sequence abnormalities, duplication abnormalities, and duration abnormalities.

[0058] Specifically, the longest matching step sequence is determined by the longest common subsequence algorithm, which is existing technology and will not be elaborated further.

[0059] Specifically, the missing anomaly refers to a situation where a job node that should exist in the standard execution chain is not executed in the actual execution chain; The redundancy anomaly refers to the situation where a job node that does not exist in the standard execution chain is executed in the actual execution chain; The so-called sequence anomaly refers to a situation where the actual sequence of adjacent work nodes is inconsistent with the standard sequence. The duplication anomaly refers to the situation where the same job node is executed multiple times in the actual execution chain, while it only needs to be executed once in the standard execution chain; The term "duration anomaly" refers to a situation where the actual execution time of a job node deviates from the standard or historical average duration by more than a preset threshold.

[0060] For example, the preset threshold can be determined by collecting the duration data of multiple executions of the same task node by the industrial robot under normal operating conditions, and calculating the average and standard deviation of this set of duration data. The preset threshold is set as the average value plus or minus a preset multiple of the standard deviation, where the preset multiple is determined empirically, and a value of 2 or 3 times is recommended. When the actual execution time exceeds this threshold range, i.e., below the average value minus the preset multiple of the standard deviation, or above the average value plus the preset multiple of the standard deviation, it is determined that the task node has an abnormal duration. In addition, those skilled in the art can adjust it according to the robot's operational stability, historical data volume, and real-time requirements.

[0061] Specifically, the process of calculating the anomaly representation value of a job includes: The ratio of the time interval uniformity of abnormal operation nodes to the time interval uniformity threshold is used as the first operation abnormality feature value. The sum of the ratio of the deviation in execution time to the deviation threshold and the ratio of the corresponding torque oscillation amplitude to the torque oscillation amplitude threshold is used as the second abnormal operation characteristic value. The first job anomaly feature value and the second job anomaly feature value are weighted and summed to determine the job anomaly characterization value; The data fluctuation characteristics include the deviation of execution time and the torque oscillation amplitude.

[0062] Specifically, by calculating the coefficient of variation of the start timestamps corresponding to adjacent abnormal job nodes, the difference between 1 and the coefficient of variation is used as the time interval uniformity. The higher this value, the more regular the occurrence time of abnormal job nodes.

[0063] The absolute value of the difference between the actual execution time and the baseline time is used as the deviation of the execution time to reflect the degree to which the execution time of the abnormal work node deviates from the normal baseline. The baseline time can be obtained by collecting multiple operation data of the industrial robot performing the same work node under normal operating conditions, recording the operation time of each operation, and calculating the arithmetic mean, which serves as the corresponding baseline time.

[0064] The difference between the maximum and minimum torque values ​​is used as the torque oscillation amplitude, reflecting the severity of torque fluctuations.

[0065] Specifically, the second operational anomaly characteristic value integrates the deviation in execution time and the torque oscillation amplitude, quantifying the execution deviation and force fluctuation severity of the abnormal operational node itself. It is a core indicator for assessing the severity of the node's anomaly. Meanwhile, the time interval uniformity reflects the temporal distribution pattern among multiple abnormal nodes, primarily used to assist in determining the periodicity and suddenness of the fault. Therefore, assigning a higher weighting coefficient to the second operational anomaly characteristic value can more accurately reflect the actual severity of the abnormal node, while also considering the ability to identify periodic faults. The coefficient is selected within the range of 0.6 to 0.8, and is set to 0.7 in this embodiment. Correspondingly, the weighting coefficient of the first operational anomaly characteristic value is selected within the range of 0.2 to 0.4, and is set to 0.3 in this embodiment. Furthermore, the specific values ​​can be adjusted based on the robot model, operating conditions, and historical fault data statistics.

[0066] In this embodiment, the purpose of setting the time interval uniformity threshold, deviation threshold, and torque oscillation amplitude threshold is to quantify the severity of abnormal operation nodes and to determine whether to restrict suspicious runtime domain segments. To avoid misjudging minor fluctuations within the normal range as serious anomalies, the relevant thresholds are adjusted as follows: Specifically, by acquiring time interval uniformity data, execution duration deviation data, and torque oscillation amplitude data of the industrial robot over several historical normal operation cycles, the mean of time interval uniformity, the mean of deviation, and the mean of torque oscillation amplitude are calculated as benchmark values. Based on the purpose of setting the above three thresholds, the time interval uniformity threshold is determined as the product of the mean time interval uniformity and the uniformity deviation coefficient; the deviation threshold is determined as the product of the mean deviation and the operation deviation coefficient; and the torque oscillation amplitude threshold is determined as the product of the mean torque oscillation amplitude and the oscillation deviation coefficient.

[0067] The uniformity deviation coefficient is selected within the range [0.85, 0.95]. The uniformity of the time interval is usually high during normal operation, and a threshold slightly lower than the mean can effectively filter out slight non-uniformities. In practice, a value of 0.9 is preferred.

[0068] The operation deviation coefficient is selected within the range [1.1, 1.3]. The deviation in execution time is less affected by the work content, and a coefficient slightly higher than the mean can eliminate random deviations. In practice, a coefficient of 1.2 is preferred.

[0069] The oscillation deviation coefficient is selected within the range [1.2, 1.4]. The torque oscillation amplitude fluctuates greatly due to the influence of mechanical characteristics, requiring a large amplification factor to distinguish between normal vibration and abnormal impact. In practice, a value of 1.3 is preferred.

[0070] The selection of the above-mentioned optimal values ​​balances the timeliness and reliability of identification, achieving a good balance between effectively filtering normal fluctuations and accurately identifying severely abnormal operation nodes. Furthermore, those skilled in the art can readjust the value range or adjust the coefficients within a given range according to the operational characteristics of the industrial robot and the requirements for limiting sensitivity.

[0071] Specifically, please refer to Figure 4 As shown, this is a logic block diagram for narrowing suspicious runtime domain segments according to an embodiment of the present invention. Narrowing the suspicious runtime domain segments includes: If the job anomaly characterization value is greater than or equal to the job anomaly characterization threshold, then the suspected runtime domain segment is restricted. The time domain segment defined by the start timestamp of the first abnormal operation node as the starting point and the end timestamp of the last abnormal operation node as the ending point is determined as the deviation time domain segment.

[0072] To determine the threshold for anomaly characterization, the anomaly characterization values ​​for each suspicious runtime segment are extracted from relevant data of the same industrial robot across several normal operating cycles. The mean and standard deviation of the sequence are calculated, and the sum of the product of the standard deviation and the sensitivity coefficient e and the mean is used as the anomaly characterization threshold. Furthermore, the anomaly characterization threshold should be dynamically updated periodically with the robot's operating cycle to adapt to changes in job content and equipment state drift. Simultaneously, the anomaly characterization threshold should be coordinated with the time interval uniformity threshold, deviation threshold, and torque oscillation amplitude threshold, and can be iteratively optimized by verifying the misjudgment ratio of normal samples.

[0073] The sensitivity coefficient e can be selected within the range [1.5, 3.0] based on the tolerance for missed detection of abnormal operations; for example, e = 2.0. If sufficient historical data is available, the 95th percentile can also be used as the threshold.

[0074] Specifically, the process of matching and outputting the fault type corresponding to the industrial robot includes: Logs within the deviation time domain are sorted chronologically and matched with samples in the fault mode library; The similarity is determined by coupling the occurrence order matching degree of adjacent log pairs with the causal semantic connectivity degree of adjacent logs. If the similarity is greater than or equal to the similarity threshold, the corresponding fault type in the fault mode library is matched and output as the fault type of the industrial robot.

[0075] In this implementation, the order matching degree of adjacent log pairs and the causal semantic connectivity degree of adjacent logs are coupled in the following ways: The similarity is obtained by assigning corresponding weight coefficients to the order matching degree of adjacent log pairs and the causal semantic connection degree of adjacent logs, and then multiplying the two by the corresponding weight coefficients and summing them.

[0076] The sum of the two weighting coefficients is 1. For example, the weighting coefficient for sequential matching can be set to 0.6, and the weighting coefficient for causal semantic coherence can be set to 0.4, reflecting that sequential consistency is slightly more important than causal logical coherence. Furthermore, these weighting coefficients can be adjusted based on the confidence level of the fault mode library or historical matching accuracy.

[0077] In this embodiment, the purpose of setting a similarity threshold is to determine the degree of matching between log sequences within the deviation time domain and samples in the fault mode library, thereby outputting the corresponding fault type. Considering that in actual operation, log sequences may exhibit accidental local similarities due to noise and communication interference, setting the threshold too low would lead to a large number of false matches and the output of incorrect fault types. Therefore, the threshold is set higher than the average similarity value during historical normal matches to reduce sensitivity. Only when sequences are highly similar is a high degree of matching considered, thus improving the accuracy of fault diagnosis.

[0078] Specifically, by acquiring similarity data from several historical cases of successful matching of known faults involving industrial robots, the mean similarity is calculated as a benchmark value. Based on the purpose of setting the similarity threshold, the similarity threshold is determined to be the product of the mean similarity and the similarity deviation coefficient.

[0079] The similarity deviation coefficient is selected within the range [1.05, 1.15]. During normal matching, the similarity value is already at a high level, and a threshold slightly higher than the mean can effectively exclude accidental approximate matches, while avoiding overly conservative approaches that may miss real faults. A value of 1.1 is preferred in practice.

[0080] The selection of the above-mentioned optimal values ​​balances the reliability and timeliness of identification, achieving a good balance between false alarm rate and false negative rate. Furthermore, those skilled in the art can readjust the value range or adjust the similarity deviation coefficient within a given range based on the sample quality of the fault mode library and the system's tolerance to false alarms.

[0081] Specifically, the occurrence order matching degree of adjacent log pairs measures the degree of consistency between the actual order in which adjacent log pairs appear and the order corresponding to the fault samples in the fault mode library. The occurrence order matching degree is calculated by determining the total number of pairs with correct order and pairs with incorrect order, and then using the ratio of the number of pairs with correct order to the total number of pairs.

[0082] The correct order refers to the order of the actual adjacent log pairs being consistent with the order in the fault sample. The incorrect sequence refers to the actual sequence being the opposite of or not corresponding to the sequence in the fault sample.

[0083] Causal semantic coherence measures the strength of the logical causal relationship between two adjacent log entries in an actual log sequence. It is determined based on historical co-occurrence frequency, that is, the probability of a subsequent log entry appearing adjacent to a preceding log entry.

[0084] Specifically, systems that provide application data acquisition methods also include, The time domain extraction module is used to obtain the operating data of the corresponding controller of the industrial robot under the normal monitoring and acquisition frequency, and to filter out suspicious runtime domain segments. An anomaly assessment module, which is connected to the time domain extraction module, is used to extract the execution job corresponding to the suspicious runtime domain segment, and evaluate the operation anomaly characterization value of the industrial robot based on the job response characteristics of the execution job, so as to classify the anomaly level of the industrial robot. An adjustment analysis module, connected to the anomaly assessment module, is used to adjust the monitoring and data acquisition frequency for the corresponding controller of the industrial robot based on the anomaly level, including: If the industrial robot is classified as having a high anomaly level, the monitoring and data collection frequency for the corresponding controller of the industrial robot will be increased, and in-depth diagnostic analysis of the industrial robot will be performed, including: Extract the job execution chain of the industrial robot within the suspected runtime domain segment and assess whether it has job continuity; Extract the data fluctuation characteristics and time interval uniformity corresponding to the abnormal operation nodes, calculate the operation anomaly characterization value, and narrow down the suspicious runtime domain segment; Based on the log sequence within the deviation time domain, the similarity of the log sequence is matched with the fault mode library, and the fault type corresponding to the industrial robot is output. If the industrial robot is at a low anomaly level, the monitoring and data collection frequency for the corresponding controller of the industrial robot will be maintained.

[0085] It should be noted that the multiple functional modules involved in this application are only a logical division based on the functions implemented according to the present invention, and are not a strict limitation on the physical structure; in practical applications, the above functional modules can be implemented by one or more integrated circuits, a processor executing program code in memory, or a combination of the above devices.

[0086] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A data acquisition method, characterized in that, include: Acquire the operating data of the industrial robot's corresponding controller at the regular monitoring and acquisition frequency, and filter out suspicious runtime domain segments; Extract the execution jobs corresponding to the suspicious runtime domain segments, and evaluate the operation anomaly characterization values ​​of the industrial robot based on the job response characteristics of the execution jobs, so as to classify the anomaly level of the industrial robot; Based on the aforementioned anomaly level, adjust the monitoring and data collection frequency for the corresponding controller of the industrial robot. include, If the industrial robot is classified as having a high anomaly level, the monitoring and data collection frequency for the corresponding controller of the industrial robot will be increased, and in-depth diagnostic analysis of the industrial robot will be performed, including: Extract the job execution chain of the industrial robot within the suspected runtime domain segment and assess whether it has job continuity; Extract the data fluctuation characteristics and time interval uniformity corresponding to the abnormal operation nodes, calculate the operation anomaly characterization value, and narrow down the suspicious runtime domain segment; Based on the log sequence within the deviation time domain, the similarity of the log sequence is matched with the fault mode library, and the fault type corresponding to the industrial robot is output. If the industrial robot is at a low anomaly level, the monitoring and data collection frequency for the corresponding controller of the industrial robot will be maintained.

2. The data acquisition method according to claim 1, characterized in that, The process of filtering suspicious runtime domain segments includes: Construct the time-domain variation curve of the joint torque of the industrial robot; Extract the starting point of the sharp increase in torque and the adjacent starting point of the decrease to lock the time domain segment of abnormal fluctuation; The torque fluctuation is determined based on the change in torque increase in each sub-time domain segment within the fluctuating anomaly time domain segment. If the torque fluctuation is greater than or equal to the torque fluctuation threshold, then the abnormal fluctuation time domain segment is selected as the suspicious runtime domain segment. The torque fluctuation is determined based on the statistical variance of the change in torque increase.

3. The data acquisition method according to claim 1, characterized in that, The process of evaluating the malfunction characteristics of the industrial robot includes: The ratio of the deviation value of the trajectory tracking error to the error deviation threshold is used as the first abnormal operation feature value; The ratio of the number of instruction response timeouts to the instruction response timeout threshold is used as the second operational anomaly characteristic value; The weighted sum of the first operational anomaly feature value and the second operational anomaly feature value is determined as the operational anomaly characterization value. The operation response characteristics include the deviation value of the trajectory tracking error and the number of command response timeouts.

4. The data acquisition method according to claim 3, characterized in that, The abnormality levels of the industrial robot are classified as follows: If the abnormal operation characteristic value of an industrial robot is greater than or equal to the abnormal operation characteristic threshold, the industrial robot is classified as a high abnormality level. If the abnormal operation characteristic value of the industrial robot is less than the abnormal operation characteristic threshold, the industrial robot will be classified as a low abnormality level.

5. The data acquisition method according to claim 1, characterized in that, The process of assessing whether there is operational continuity includes: If an industrial robot meets the conditions for sequential operation, then it is determined that it has operational continuity. The timing coherence condition includes that the number of consecutive errors of the executed job node is less than the consecutive error number threshold, and the baseline overlap rate of the executed job node is greater than the baseline overlap rate threshold.

6. The data acquisition method according to claim 1, characterized in that, The process of identifying abnormal job nodes includes: Compare the job execution chain with the standard execution chain and calculate the longest matching step sequence; Based on the longest matching step sequence, the execution job nodes in the job execution chain are time-aligned with the standard job nodes in the standard execution chain to establish matching node pairs; If an abnormal situation exists, the corresponding execution job node will be identified as the abnormal job node; The abnormal situations include missing abnormalities, redundancy abnormalities, sequence abnormalities, duplication abnormalities, and duration abnormalities.

7. The data acquisition method according to claim 1, characterized in that, The process of calculating job anomaly representation values ​​includes: The ratio of the time interval uniformity of abnormal operation nodes to the time interval uniformity threshold is used as the first operation abnormality feature value. The sum of the ratio of the deviation in execution time to the deviation threshold and the ratio of the corresponding torque oscillation amplitude to the torque oscillation amplitude threshold is used as the second abnormal operation characteristic value. The first job anomaly feature value and the second job anomaly feature value are weighted and summed to determine the job anomaly characterization value; The data fluctuation characteristics include the deviation of execution time and the torque oscillation amplitude.

8. The data acquisition method according to claim 7, characterized in that, The narrowing of the suspected runtime domain segment includes: If the job anomaly characterization value is greater than or equal to the job anomaly characterization threshold, then the suspected runtime domain segment is restricted. The time domain segment defined by the start timestamp of the first abnormal operation node as the starting point and the end timestamp of the last abnormal operation node as the ending point is determined as the deviation time domain segment.

9. The data acquisition method according to claim 1, characterized in that, The process of matching and outputting the fault type corresponding to the industrial robot includes: Logs within the deviation time domain are sorted chronologically and matched with samples in the fault mode library; The similarity is determined by coupling the occurrence order matching degree of adjacent log pairs with the causal semantic connectivity degree of adjacent logs. If the similarity is greater than or equal to the similarity threshold, the corresponding fault type in the fault mode library is matched and output as the fault type of the industrial robot.

10. A system applying the data acquisition method according to any one of claims 1-9, characterized in that, include, The time domain extraction module is used to obtain the operating data of the corresponding controller of the industrial robot under the normal monitoring and acquisition frequency, and to filter out suspicious runtime domain segments. An anomaly assessment module is used to extract the execution jobs corresponding to the suspicious runtime domain segments, and to assess the anomaly characterization value of the industrial robot based on the job response characteristics of the execution jobs, so as to classify the anomaly level of the industrial robot. The adjustment analysis module is used to adjust the monitoring and data acquisition frequency for the corresponding controller of the industrial robot based on the anomaly level, including: If the industrial robot is classified as having a high anomaly level, the monitoring and data collection frequency for the corresponding controller of the industrial robot will be increased, and in-depth diagnostic analysis of the industrial robot will be performed, including: Extract the job execution chain of the industrial robot within the suspected runtime domain segment and assess whether it has job continuity; Extract the data fluctuation characteristics and time interval uniformity corresponding to the abnormal operation nodes, calculate the operation anomaly characterization value, and narrow down the suspicious runtime domain segment; Based on the log sequence within the deviation time domain, the similarity of the log sequence is matched with the fault mode library, and the fault type corresponding to the industrial robot is output. If the industrial robot is at a low anomaly level, the monitoring and data collection frequency for the corresponding controller of the industrial robot will be maintained.

Images