Anomaly data management method and system for industrial robot

Abstract

The application relates to the technical field of intelligent management of industrial equipment, and particularly discloses an abnormal data management method and system for an industrial robot, which comprises the following steps: receiving operation data fed back by a monitoring sensor built in the industrial robot, positioning an abnormal moment according to the operation data; positioning a camera according to the installation position of the industrial robot, establishing a connection channel with the storage space of the camera, and reading a motion video with the abnormal moment as the center; identifying the motion video to obtain an abnormal trajectory; calculating the matching degree of the operation data and the abnormal trajectory, determining an extension interval according to the matching degree, and obtaining and storing the operation data, the motion video and an operation process based on the extension interval. On the basis of original identification, the visual identification process based on an existing monitoring system is introduced, the precision requirement of the two identification modes is reduced through the introduction of the matching degree calculation process, and the identification cost is reduced.

Description

Technical Field

[0001] This invention relates to the field of intelligent management technology for industrial equipment, specifically a method and system for managing abnormal data of industrial robots. Background Technology

[0002] Industrial robots are multi-jointed manipulators or multi-degree-of-freedom robotic devices widely used in industrial fields. They possess a certain degree of automation and can perform various industrial processing and manufacturing functions using their own power and control capabilities. Industrial robots are widely used in various industrial sectors such as electronics, logistics, and chemicals.

[0003] The operation of industrial robots requires real-time monitoring, but most existing monitoring methods are independent monitoring processes, such as monitoring by intelligent modules built into the industrial robot. This method places high demands on the performance of the intelligent modules and related data processing algorithms, which makes the cost very high. How to reduce monitoring costs while ensuring the accuracy of data monitoring is the technical problem that the present invention aims to solve. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for managing abnormal data of industrial robots, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A method for managing abnormal data of an industrial robot, the method comprising:

[0007] It receives operational data from monitoring sensors built into industrial robots and locates abnormal moments based on the operational data;

[0008] The camera is located based on the installation position of the industrial robot, and a connection channel is established with the camera's storage space to read motion video at abnormal moments.

[0009] The motion video is identified to obtain abnormal trajectories;

[0010] Calculate the matching degree between the running data and the abnormal trajectory, determine the extension interval based on the matching degree, and acquire and store the running data, motion video and calculation process based on the extension interval;

[0011] The computational process includes an abnormal moment location process based on runtime data and a motion video recognition process.

[0012] As a further aspect of the present invention: the step of receiving operational data fed back from monitoring sensors built into the industrial robot and locating abnormal moments based on the operational data includes:

[0013] It receives operational data from monitoring sensors built into the industrial robot and maps the operational data to the same interval based on the same mapping function;

[0014] Based on the same time axis, all running data are statistically analyzed to obtain motion curves;

[0015] The derivative and mean of the derivative of the motion curve are calculated progressively. When the mean of the derivative reaches a preset mean threshold, the calculation process is stopped and the order of the derivative of the curve is recorded.

[0016] The outliers at each time point are calculated based on the derivatives of each order, and the time when the outlier reaches the preset outlier threshold is selected as the outlier time.

[0017] As a further aspect of the present invention: the mapping function is:

[0018] In the formula, Y i Let X be the value after the i-th mapping. i The value before the i-th mapping; D is the interval span of the same interval, which is a preset value;

[0019] The calculation process for the outlier is as follows: In the formula, A t Let α be the outlier at time t. j The weighting coefficient of the j-th derivative, f j (t) is the j-th derivative at time t, and M is the total order of differentiation.

[0020] As a further aspect of the present invention: the step of identifying the abnormal trajectory in the motion video includes:

[0021] The distance between the camera and a preset calibration point is obtained in real time based on a rangefinder installed in the camera; the calibration point is set on the industrial robot.

[0022] The spatial dividing plane is determined based on the distance, and the spatial unit is determined based on the spatial dividing plane;

[0023] The proportion of robots in each spatial unit is obtained based on the camera;

[0024] The proportion of robots in each spatial unit is input into a preset analysis model to determine the pose of the industrial robot.

[0025] By analyzing the poses of industrial robots in chronological order, abnormal trajectories can be determined.

[0026] As a further aspect of the present invention: the step of determining abnormal trajectories by statistically analyzing the poses of the industrial robot in chronological order includes:

[0027] The pose of the industrial robot at each moment is read, and the motion characteristics of the industrial robot are calculated based on the pose at each moment; the motion characteristics include the moving parts, their motion direction and motion amplitude; the moving parts are one part of the industrial robot.

[0028] By connecting the motion characteristics at each moment in chronological order, the abnormal trajectory is obtained.

[0029] As a further aspect of the present invention: the steps of calculating the matching degree between the running data and the abnormal trajectory, determining the extension interval based on the matching degree, and acquiring and storing the running data, motion video, and calculation process based on the extension interval include:

[0030] Focusing on the abnormal moment, read the running data with the same duration as the motion video;

[0031] According to a preset selection frequency, select data to be tested from the running data, input the data to be tested into the trained neural network model, and obtain the allowed pose set;

[0032] Iterate through the allowed poses in the allowed pose set and determine whether an allowed pose is contained in an abnormal trajectory. If an allowed pose is contained in an abnormal trajectory, exit the traversal process and set the matching result to one. If none of the allowed poses are contained in an abnormal trajectory, set the matching result to zero.

[0033] Calculate the judgment results corresponding to all the data to be inspected to obtain a judgment array;

[0034] The extension radius is determined based on the judgment array, and the extension interval is determined with the abnormal moment as the center.

[0035] Read and store the running data, motion video and calculation process within the extended range.

[0036] The present invention also provides an abnormal data management system for industrial robots, the system comprising:

[0037] The abnormal moment location module is used to receive operating data fed back from the monitoring sensors built into the industrial robot and locate the abnormal moment based on the operating data.

[0038] The video reading module is used to locate the camera based on the installation position of the industrial robot and establish a connection channel with the camera's storage space to read motion video at abnormal moments.

[0039] A video recognition module is used to identify the motion video and obtain abnormal trajectories;

[0040] The data reading module is used to calculate the matching degree between the running data and the abnormal trajectory, determine the extension interval based on the matching degree, and acquire and store the running data, motion video and calculation process based on the extension interval;

[0041] The computational process includes an abnormal moment location process based on runtime data and a motion video recognition process.

[0042] As a further aspect of the present invention: the abnormal moment location module includes:

[0043] The data mapping unit is used to receive the operating data fed back by the monitoring sensors built into the industrial robot, and to map the operating data to the same interval based on the same mapping function;

[0044] The curve generation unit is used to statistically analyze all running data based on the same time axis to obtain motion curves;

[0045] The derivative calculation unit is used to progressively calculate the derivative and mean of the motion curve. When the mean of the derivative reaches a preset mean threshold, the calculation process stops and the order of the derivative of the curve is recorded.

[0046] The time selection unit is used to calculate the outlier value at each time based on the derivatives of each order, and select the time when the outlier value reaches the preset outlier threshold as the outlier time.

[0047] As a further aspect of the present invention: the video recognition module includes:

[0048] A distance measurement unit is used to acquire the distance between the camera and a preset calibration point in real time based on a rangefinder installed in the camera; the calibration point is set on the industrial robot.

[0049] A spatial segmentation unit is used to determine a spatial segmentation surface based on the distance, and to determine a spatial unit based on the spatial segmentation surface.

[0050] The percentage acquisition unit is used to acquire the percentage of robots in each spatial unit based on the camera.

[0051] The pose determination unit is used to input the proportion of the robot in each spatial unit into a preset analysis model to determine the pose of the industrial robot.

[0052] The pose statistics unit is used to count the pose of the industrial robot in chronological order and determine abnormal trajectories.

[0053] As a further aspect of the present invention: the pose statistics unit includes:

[0054] The feature acquisition subunit is used to read the pose of the industrial robot at each moment and calculate the motion features of the industrial robot based on the pose at each moment; the motion features include the moving parts, their motion direction and motion amplitude; the moving parts are one part of the industrial robot.

[0055] The feature connection subunit is used to connect the motion features at each moment in chronological order to obtain the abnormal trajectory.

[0056] Compared with the prior art, the beneficial effects of the present invention are: the present invention introduces a visual recognition process based on the existing monitoring system on the basis of the original recognition, and by introducing a matching degree calculation process, the accuracy requirements of the two recognition methods are reduced, thereby indirectly reducing the recognition cost. Attached Figure Description

[0057] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention.

[0058] Figure 1 A flowchart illustrating the abnormal data management method for industrial robots.

[0059] Figure 2 This is the first sub-flowchart of the abnormal data management method for industrial robots.

[0060] Figure 3 This is the second sub-flowchart of the abnormal data management method for industrial robots.

[0061] Figure 4 This is the third sub-flowchart of the abnormal data management method for industrial robots.

[0062] Figure 5 This is a block diagram showing the structural composition of an abnormal data management system for industrial robots. Detailed Implementation

[0063] To make the technical problems to be solved, the technical solutions, and the beneficial effects of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present invention and are not intended to limit the present invention.

[0064] Figure 1 This is a flowchart of an abnormal data management method for industrial robots. In this embodiment of the invention, an abnormal data management method for industrial robots includes:

[0065] Step S100: Receive operating data from the monitoring sensors built into the industrial robot, and locate the abnormal moment based on the operating data;

[0066] Industrial robots are intelligent working components installed on intelligent production lines. Most existing industrial robots have multiple built-in monitoring sensors to collect operational data. There are many types of monitoring sensors, which can acquire different types of data. The simplest are physical parameters such as current and voltage. If cost permits, microprocessors can be used to replace monitoring sensors. In this case, in addition to current and voltage, different control signals can also be captured. By analyzing the operational data, abnormal data can be identified, and the time stamps of these data can be read as the abnormal moments.

[0067] Step S200: Locate the camera according to the installation position of the industrial robot and establish a connection channel with the camera's storage space to read motion video centered on abnormal moments;

[0068] In a smart factory, multiple cameras are installed to capture images of the production process. The location process involves finding the camera that can capture images of the industrial robot based on its installation location. Once the camera is located, a connection is established with its storage space to read the motion video.

[0069] It should be noted that the video reading process is essentially a multi-moment image reading process. Video analysis always involves analyzing video over a period of time. Therefore, when reading motion video, it is necessary to focus on the abnormal moment and read the motion video over a period of time as the subsequent identification object. The time span of the motion video is a preset value input by the industrial robot.

[0070] Step S300: Identify the motion video to obtain abnormal trajectories;

[0071] By identifying motion videos, the motion trajectory of industrial robots in the videos can be obtained. Since the motion videos are based on videos read at abnormal moments, the obtained motion trajectories are called abnormal trajectories.

[0072] Step S400: Calculate the matching degree between the running data and the abnormal trajectory, determine the extension interval based on the matching degree, and acquire and store the running data, motion video and calculation process based on the extension interval;

[0073] Operational data is the internal data of an industrial robot, while abnormal trajectories are external data. There is a certain correspondence between the two, meaning that when operational data is abnormal, abnormal trajectories can be predicted. Based on the matching of operational data and abnormal trajectories, it can be determined whether the current anomaly identification process is accurate. If only operational data is abnormal or only abnormal trajectories are abnormal, then it can be considered that there is a problem with the data acquisition process, and the acquired data will be identified as invalid data. The advantage of this architecture is that the identification process of operational data and the identification process of motion video do not need to be very complicated, and even with low cost, relatively accurate identification results can be obtained.

[0074] The computational process includes an abnormal moment location process based on runtime data and a motion video recognition process.

[0075] In one example of the technical solution of this invention, the final result is running data, motion video and calculation process, which are collectively referred to as abnormal data. Taking the abnormal moment as the center, a more suitable time span (extended interval) is determined, and the data within the time span is read and stored.

[0076] Figure 2 The first sub-flowchart of the abnormal data management method for industrial robots includes steps S101 to S104, which involve receiving operational data from monitoring sensors built into the industrial robot and locating the moment of abnormality based on the operational data.

[0077] Step S101: Receive the operation data fed back by the monitoring sensors built into the industrial robot, and map the operation data to the same interval based on the same mapping function;

[0078] Step S102: Statistically analyze all running data based on the same time axis to obtain the motion curve;

[0079] Step S103: Progressively calculate the derivative and mean of the derivative of the motion curve. When the mean of the derivative reaches the preset mean threshold, stop the calculation process and record the order of the derivative of the curve.

[0080] Step S104: Calculate the outlier values ​​at each time point based on the derivatives of each order, and select the time when the outlier values ​​reach the preset outlier threshold as the outlier time.

[0081] The above content describes in detail the localization process at abnormal moments. First, the operation data fed back by the monitoring sensors is received. The data format and value range of the operation data are not fixed. Therefore, it is necessary to convert it into values ​​within the same range.

[0082] After each type of operational data is transformed, a corresponding value can be obtained. Statistical analysis of the obtained values ​​based on the same time axis yields multiple motion curves. For each motion curve, progressive differentiation is performed, that is, the first derivative is calculated first, then the second derivative, and so on, until the average value of the final derivative is sufficiently small. At this point, based on the values ​​of each derivative at each time point, the data fluctuation at each time point can be calculated. The more drastic the data fluctuation, the higher the probability of an anomaly and the higher the outlier. The outlier is compared with a preset anomaly threshold. When the outlier reaches the preset anomaly threshold, the corresponding time point is marked as an anomaly time.

[0083] Specifically, In the formula, Y i Let X be the value after the i-th mapping. i The value before the i-th mapping; D is the interval span of the same interval, which is a preset value;

[0084] The calculation process for the outlier is as follows: In the formula, A t Let α be the outlier at time t. j The weighting coefficient of the j-th derivative, f j (t) is the j-th derivative at time t, and M is the total order of differentiation.

[0085] Figure 3 The second sub-flowchart of the abnormal data management method for industrial robots includes the step of identifying the abnormal trajectory from the motion video:

[0086] Step S301: The distance between the camera and a preset calibration point is obtained in real time based on the rangefinder installed in the camera; the calibration point is set on the industrial robot.

[0087] Step S302: Determine the spatial dividing plane based on the distance, and determine the spatial unit based on the spatial dividing plane;

[0088] Step S303: Obtain the proportion of robots in each spatial unit based on the camera;

[0089] Step S304: Input the proportion of the robot in each spatial unit into the preset analysis model to determine the pose of the industrial robot;

[0090] Step S305: Analyze the pose of the industrial robot according to the time sequence to determine abnormal trajectories.

[0091] In one example of the technical solution of this invention, the recognition process of motion video is specifically defined. The principle is that the distance between the camera and the industrial robot (calibration point) is obtained by the rangefinder. The distance determines some mutually perpendicular planes, and the space is distinguished by the planes. Generally, the space is divided into nine spatial units of three by three. The current pose of the industrial robot can be determined according to the filling situation of the working robot in each spatial unit. The more spatial units there are, the higher the accuracy of the determined pose. By statistically analyzing the pose of the industrial robot at each moment, the behavioral information of the industrial robot can be obtained.

[0092] Specifically, the step of determining abnormal trajectories by statistically analyzing the poses of the industrial robot in chronological order includes:

[0093] The pose of the industrial robot at each moment is read, and the motion characteristics of the industrial robot are calculated based on the pose at each moment; the motion characteristics include the moving parts, their motion direction and motion amplitude; the moving parts are one part of the industrial robot.

[0094] By connecting the motion characteristics at each moment in chronological order, the abnormal trajectory is obtained.

[0095] The above content provides a specific scheme for determining abnormal trajectories. In layman's terms, abnormal trajectories refer to the direction and distance of translation and rotation of each robotic arm of an industrial robot. The generation logic is to statistically analyze the position of each component at each moment according to the time sequence and connect them to obtain the motion trajectory of multiple components, which is called abnormal trajectory.

[0096] Figure 4 The third sub-flow diagram of the abnormal data management method for industrial robots includes the following steps: calculating the matching degree between the running data and the abnormal trajectory, determining the extension interval based on the matching degree, and acquiring and storing the running data, motion video, and calculation process based on the extension interval.

[0097] Step S401: Centered on the abnormal moment, read the running data with the same duration as the motion video;

[0098] When comparing runtime data and motion videos, data of the same duration should be used as the two sides for comparison. The motion video is obtained by focusing on the abnormal moment. Based on this, the same runtime data as the motion video can be read based on the same time span.

[0099] Step S402: Select data to be inspected from the running data according to the preset selection frequency, and input the data to be inspected into the trained neural network model to obtain the allowed pose set;

[0100] The amount of running data is large. During the analysis process, it is not necessary to analyze all the running data. Only a portion of the data needs to be selected for analysis. The selection process is adjusted by the parameter of selection frequency. The selection frequency refers to the selection of running data at several moments within a time period. The selection method can be uniform selection or random selection. The selected running data is called the data to be tested. The data to be tested is input into the trained neural network model to obtain the set of allowed poses.

[0101] The neural network model is a mapping relationship between running data and allowed poses, which is obtained by staff through statistics. The statistical method is generally a reverse statistical method, that is, different poses are counted, then the running data corresponding to different poses are queried, and the running data is compared and merged to establish the relationship between running data and allowed poses.

[0102] Step S403: Traverse the allowed poses in the allowed pose set and determine whether the allowed poses are contained in the abnormal trajectory; when there is an allowed pose contained in the abnormal trajectory, exit the traversal process and determine the matching result as one; when all allowed poses are not contained in the abnormal trajectory, determine the matching result as zero.

[0103] In practical applications, the allowed poses corresponding to the running data at a certain moment are read, and it is determined whether they are contained in the abnormal trajectory. As long as there is an allowed pose contained in the abnormal trajectory, the running data and the motion video can be determined to match. When a match is determined, it is represented by one; when a mismatch is determined, it is represented by zero.

[0104] Step S404: Calculate the judgment results corresponding to all the data to be inspected to obtain the judgment array;

[0105] By reading the judgment results corresponding to all the data to be inspected, we can obtain an array consisting of one and zero, which is called the judgment array.

[0106] Step S405: Determine the extension radius based on the judgment array, and determine the extension interval with the abnormal moment as the center;

[0107] The judgment array is analyzed, for example, by summing the results. The extension radius is determined based on the summation result. The determination process can be carried out using a linear formula. The extension radius is proportional to the summation result. Taking the abnormal moment as the center, the extension interval can be obtained by combining the extension radius.

[0108] Step S406: Read and store the running data, motion video and calculation process within the extended interval;

[0109] Finally, the running data, motion video, and calculation process are read and stored.

[0110] Figure 5This is a block diagram illustrating the structural composition of an abnormal data management system for industrial robots. In this embodiment of the invention, an abnormal data management system for industrial robots, system 10, includes:

[0111] The abnormal moment location module 11 is used to receive the operating data fed back by the monitoring sensors built into the industrial robot and locate the abnormal moment based on the operating data.

[0112] The video reading module 12 is used to locate the camera according to the installation position of the industrial robot and establish a connection channel with the camera's storage space to read motion video centered on abnormal moments.

[0113] Video recognition module 13 is used to identify the motion video and obtain abnormal trajectories;

[0114] The data reading module 14 is used to calculate the matching degree between the running data and the abnormal trajectory, determine the extension interval based on the matching degree, and acquire and store the running data, motion video and calculation process based on the extension interval;

[0115] The computational process includes an abnormal moment location process based on runtime data and a motion video recognition process.

[0116] Furthermore, the abnormal time location module 11 includes:

[0117] The data mapping unit is used to receive the operating data fed back by the monitoring sensors built into the industrial robot, and to map the operating data to the same interval based on the same mapping function;

[0118] The curve generation unit is used to statistically analyze all running data based on the same time axis to obtain motion curves;

[0119] The derivative calculation unit is used to progressively calculate the derivative and mean of the motion curve. When the mean of the derivative reaches a preset mean threshold, the calculation process stops and the order of the derivative of the curve is recorded.

[0120] The time selection unit is used to calculate the outlier value at each time based on the derivatives of each order, and select the time when the outlier value reaches the preset outlier threshold as the outlier time.

[0121] Furthermore, the video recognition module 13 includes:

[0122] A distance measurement unit is used to acquire the distance between the camera and a preset calibration point in real time based on a rangefinder installed in the camera; the calibration point is set on the industrial robot.

[0123] A spatial segmentation unit is used to determine a spatial segmentation surface based on the distance, and to determine a spatial unit based on the spatial segmentation surface.

[0124] The percentage acquisition unit is used to acquire the percentage of robots in each spatial unit based on the camera.

[0125] The pose determination unit is used to input the proportion of the robot in each spatial unit into a preset analysis model to determine the pose of the industrial robot.

[0126] The pose statistics unit is used to count the pose of the industrial robot in chronological order and determine abnormal trajectories.

[0127] Specifically, the pose statistics unit includes:

[0128] The feature acquisition subunit is used to read the pose of the industrial robot at each moment and calculate the motion features of the industrial robot based on the pose at each moment; the motion features include the moving parts, their motion direction and motion amplitude; the moving parts are one part of the industrial robot.

[0129] The feature connection subunit is used to connect the motion features at each moment in chronological order to obtain the abnormal trajectory.

[0130] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for managing abnormal data of an industrial robot, characterized in that, The method includes: It receives operational data from monitoring sensors built into industrial robots and locates abnormal moments based on the operational data; The camera is located based on the installation position of the industrial robot, and a connection channel is established with the camera's storage space to read motion video at abnormal moments. The motion video is identified to obtain abnormal trajectories; Calculate the matching degree between the running data and the abnormal trajectory, determine the extension interval based on the matching degree, and acquire and store the running data, motion video and calculation process based on the extension interval; The computational process includes an abnormal moment location process based on runtime data and a motion video recognition process. The steps of calculating the matching degree between the running data and the abnormal trajectory, determining the extension interval based on the matching degree, and acquiring and storing the running data, motion video, and calculation process based on the extension interval include: Focusing on the abnormal moment, read the running data with the same duration as the motion video; According to a preset selection frequency, select data to be tested from the running data, input the data to be tested into the trained neural network model, and obtain the allowed pose set; Iterate through the allowed poses in the allowed pose set and determine whether an allowed pose is contained in an abnormal trajectory. If an allowed pose is contained in an abnormal trajectory, exit the traversal process and set the matching result to one. If none of the allowed poses are contained in an abnormal trajectory, set the matching result to zero. Calculate the judgment results corresponding to all the data to be inspected to obtain a judgment array; The extension radius is determined based on the judgment array, and the extension interval is determined with the abnormal moment as the center. Read and store the running data, motion video and calculation process within the extended range.

2. The abnormal data management method for industrial robots according to claim 1, characterized in that, The step of receiving operational data from monitoring sensors built into the industrial robot and locating abnormal moments based on the operational data includes: It receives operational data from monitoring sensors built into the industrial robot and maps the operational data to the same interval based on the same mapping function; Based on the same time axis, all running data are statistically analyzed to obtain motion curves; The derivative and mean of the derivative of the motion curve are calculated progressively. When the mean of the derivative reaches a preset mean threshold, the calculation process is stopped and the order of the derivative of the curve is recorded. The outliers at each time point are calculated based on the derivatives of each order, and the time when the outlier reaches the preset outlier threshold is selected as the outlier time.

3. The abnormal data management method for industrial robots according to claim 2, characterized in that, The mapping function is: In the formula, Let i be the value after the mapping. The value before the i-th mapping; D is the interval span of the same interval, which is a preset value; The calculation process for the outlier is as follows: In the formula, The outlier at time t The weighting coefficient of the j-th derivative, ; Let be the j-th derivative at time t, and M be the total order of differentiation.

4. The abnormal data management method for industrial robots according to claim 1, characterized in that, The step of identifying abnormal trajectories from the motion video includes: The distance between the camera and a preset calibration point is obtained in real time based on a rangefinder installed in the camera; the calibration point is set on the industrial robot. The spatial dividing plane is determined based on the distance, and the spatial unit is determined based on the spatial dividing plane; The proportion of robots in each spatial unit is obtained based on the camera; The proportion of robots in each spatial unit is input into a preset analysis model to determine the pose of the industrial robot. By analyzing the poses of industrial robots in chronological order, abnormal trajectories can be determined.

5. The abnormal data management method for industrial robots according to claim 4, characterized in that, The step of determining abnormal trajectories by statistically analyzing the pose of the industrial robot in chronological order includes: The pose of the industrial robot at each moment is read, and the motion characteristics of the industrial robot are calculated based on the pose at each moment; the motion characteristics include the moving parts, their motion direction and motion amplitude; the moving parts are one part of the industrial robot. By connecting the motion characteristics at each moment in chronological order, the abnormal trajectory is obtained.

6. An abnormal data management system for an industrial robot, characterized in that, The system includes: The abnormal moment location module is used to receive operating data fed back from the monitoring sensors built into the industrial robot and locate the abnormal moment based on the operating data. The video reading module is used to locate the camera based on the installation position of the industrial robot and establish a connection channel with the camera's storage space to read motion video at abnormal moments. A video recognition module is used to identify the motion video and obtain abnormal trajectories; The data reading module is used to calculate the matching degree between the running data and the abnormal trajectory, determine the extension interval based on the matching degree, and acquire and store the running data, motion video and calculation process based on the extension interval; The computational process includes an abnormal moment location process based on runtime data and a motion video recognition process. The calculation of the matching degree between the running data and the abnormal trajectory, the determination of the extension interval based on the matching degree, and the acquisition and storage of running data, motion video, and calculation process based on the extension interval include: Focusing on the abnormal moment, read the running data with the same duration as the motion video; According to a preset selection frequency, select data to be tested from the running data, input the data to be tested into the trained neural network model, and obtain the allowed pose set; Iterate through the allowed poses in the allowed pose set and determine whether an allowed pose is contained in an abnormal trajectory. If an allowed pose is contained in an abnormal trajectory, exit the traversal process and set the matching result to one. If none of the allowed poses are contained in an abnormal trajectory, set the matching result to zero. Calculate the judgment results corresponding to all the data to be inspected to obtain a judgment array; The extension radius is determined based on the judgment array, and the extension interval is determined with the abnormal moment as the center. Read and store the running data, motion video and calculation process within the extended range.

7. The abnormal data management system for industrial robots according to claim 6, characterized in that, The abnormal moment location module includes: The data mapping unit is used to receive the operating data fed back by the monitoring sensors built into the industrial robot, and to map the operating data to the same interval based on the same mapping function; The curve generation unit is used to statistically analyze all running data based on the same time axis to obtain motion curves; The derivative calculation unit is used to progressively calculate the derivative and mean of the motion curve. When the mean of the derivative reaches a preset mean threshold, the calculation process stops and the order of the derivative of the curve is recorded. The time selection unit is used to calculate the outlier value at each time based on the derivatives of each order, and select the time when the outlier value reaches the preset outlier threshold as the outlier time.

8. The abnormal data management system for industrial robots according to claim 6, characterized in that, The video recognition module includes: A distance measurement unit is used to acquire the distance between the camera and a preset calibration point in real time based on a rangefinder installed in the camera; the calibration point is set on the industrial robot. A spatial segmentation unit is used to determine a spatial segmentation surface based on the distance, and to determine a spatial unit based on the spatial segmentation surface. The percentage acquisition unit is used to acquire the percentage of robots in each spatial unit based on the camera. The pose determination unit is used to input the proportion of the robot in each spatial unit into a preset analysis model to determine the pose of the industrial robot. The pose statistics unit is used to count the pose of the industrial robot in chronological order and determine abnormal trajectories.

9. The abnormal data management system for industrial robots according to claim 8, characterized in that, The pose statistics unit includes: The feature acquisition subunit is used to read the pose of the industrial robot at each moment and calculate the motion features of the industrial robot based on the pose at each moment; the motion features include the moving parts, their motion direction and motion amplitude; the moving parts are one part of the industrial robot. The feature connection subunit is used to connect the motion features at each moment in chronological order to obtain the abnormal trajectory.

Images