Industrial robot predictive monitoring and maintenance system based on digital twinning

Abstract

The invention relates to the technical field of industrial robot data processing, in particular to an industrial robot predictive monitoring and maintenance system based on digital twinning, and the system comprises a data obtaining module which is used for obtaining multiple pieces of core data; the data processing module is used for determining the initial reference degree of each piece of real data and correcting the initial reference degree of each piece of real data to obtain an effective reference degree; the data calculation module is used for determining the abnormal delay degree of each item of twin data and calculating the effective reference degree of each item of twin data; the machine learning module is used for extracting the data feature of each item of core data, setting the weight value of the data feature, and inputting the data feature and the weight value of each item of core data into a machine learning model to obtain a state monitoring result; and the monitoring and maintenance module is used for carrying out predictive monitoring and maintenance on key parts of the industrial robot according to the state monitoring result. According to the invention, accurate and reliable predictive monitoring and maintenance of the industrial robot can be realized.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial robot data processing, and particularly relates to a predictive monitoring and maintenance system for industrial robots based on digital twin. Background Technique

[0002] Digital twin is a technology that creates a virtual model of a physical entity through digital means and uses real-time data for simulation, monitoring, and optimization. In the context of digital twin, the condition monitoring and maintenance of key parts of industrial robots are crucial.

[0003] In the related art, during the process of monitoring and maintaining the key parts of industrial robots, it is usually realized only based on the real data collected by data acquisition devices such as sensors. However, a single data source is difficult to comprehensively reflect the health status of the key parts of industrial robots, and the real data collected by data acquisition devices such as sensors is easily interfered by various factors, resulting in insufficient accuracy and reliability in the process of monitoring and maintaining the key parts of industrial robots. Summary of the Invention

[0004] In order to solve the technical problems that the traditional monitoring and maintenance scheme has a single data source and the real data is easily interfered by various factors, resulting in insufficient accuracy and reliability in the process of monitoring and maintaining the key parts of industrial robots, the purpose of the present invention is to provide a predictive monitoring and maintenance system for industrial robots based on digital twin, and the specific technical solutions adopted are as follows:

[0005] A predictive monitoring and maintenance system for industrial robots based on digital twin, the system includes:

[0006] A data acquisition module, configured to acquire multiple core data of key parts of an industrial robot; wherein, each core data includes real data and twin data;

[0007] A data processing module, configured to respectively determine the initial reference degree and the missing core degree of each real data, and based on the missing core degree, correct the initial reference degree of each real data to obtain the effective reference degree of each real data;

[0008] A data calculation module, configured to determine the abnormal delay degree of each twin data, and calculate the effective reference degree of each twin data based on the abnormal delay degree of each twin data and the effective reference degree of the corresponding real data;

[0009] A machine learning module, configured to extract data features of each core data item, set weight values for the corresponding data features according to the effective reference degrees of the real data and the twin data in each core data item, input the data features and weight values of each core data item into a pre-constructed machine learning model, and obtain a status monitoring result output by the machine learning model;

[0010] A monitoring and maintenance module, configured to perform predictive monitoring and maintenance on key parts of an industrial robot according to the status monitoring result.

[0011] For a predictive monitoring and maintenance system of an industrial robot based on digital twin provided by the present invention, the data processing module determines the initial reference degree of each real data item, including:

[0012] Determine the theoretical quantity, actual quantity, time series sequence of the real data in each core data item, and the time series sequence of the twin data;

[0013] Divide the actual quantity by the theoretical quantity to obtain a quantity ratio;

[0014] Calculate the dynamic time warping distance between the time series sequence of the real data and the time series sequence of the twin data in each core data item;

[0015] Add 1 to the dynamic time warping distance and take the reciprocal to obtain a distance correlation value;

[0016] Multiply the quantity ratio by the distance correlation value to calculate the initial reference degree of each real data item.

[0017] For a predictive monitoring and maintenance system of an industrial robot based on digital twin provided by the present invention, the data processing module determines the missing core degree of each real data item, including:

[0018] Determine the fluctuation instability of each element data in the time series sequence of each real data item;

[0019] Based on the fluctuation instability, perform clustering processing on all data elements in the time series sequence of each real data item to generate multiple clusters;

[0020] Determine the fluctuation instability and data missing amount corresponding to each cluster;

[0021] Based on the fluctuation instability and data missing amount of each cluster corresponding to each real data item, calculate the missing core degree of each real data item.

[0022] For a predictive monitoring and maintenance system of an industrial robot based on digital twin provided by the present invention, to determine the fluctuation instability of each element data in the time series sequence of each real data item, including:

[0023] Determine the data value, slope value of each element data in the time series of each piece of real data, and the data value and slope value of each surrounding data corresponding to each element data;

[0024] Calculate the absolute value of the difference between the slope value of each element data and the slope value of each surrounding data to obtain the slope difference value of each surrounding data under each element data;

[0025] Calculate the absolute value of the difference between the data value of each element data and the data value of each surrounding data to obtain the numerical difference value of each surrounding data under each element data;

[0026] Multiply the slope difference value corresponding to each surrounding data by the numerical difference value to obtain the difference product value corresponding to each surrounding data;

[0027] Take the average of the difference product values corresponding to all surrounding data under each element data to obtain the fluctuation instability degree of each element data in the time series of each piece of real data.

[0028] According to a predictive monitoring and maintenance system for industrial robots based on digital twin provided by the present invention, based on the fluctuation instability degree and data missing amount of each cluster corresponding to each piece of real data, calculate the missing core degree of each piece of real data, including:

[0029] Sum up the fluctuation instability degrees of all clusters corresponding to each piece of real data to obtain the cluster fluctuation sum value;

[0030] Divide the fluctuation instability degree of each cluster by the cluster fluctuation sum value to obtain the missing value influence degree;

[0031] Multiply the data missing amount of each cluster by the missing value influence degree to obtain the cluster missing importance degree of each cluster;

[0032] Take the average of the cluster missing importance degrees of all clusters corresponding to each piece of real data and perform normalization processing to calculate the missing core degree of each piece of real data.

[0033] According to a predictive monitoring and maintenance system for industrial robots based on digital twin provided by the present invention, the data processing module corrects the initial reference degree of each piece of real data based on the missing core degree to obtain the effective reference degree of each piece of real data, including:

[0034] Multiply the missing core degree of each piece of real data by the initial reference value to obtain the effective reference degree of each piece of real data.

[0035] A predictive monitoring and maintenance system for industrial robots based on digital twins provided by the present invention, wherein the data calculation module determines the abnormal delay degree of each piece of twin data, including:

[0036] Obtain the acquisition time of the last element data in the time series of each piece of real data, the generation time of the first element data in the time series of each piece of twin data, the theoretical data volume of each piece of real data, and the data generation volume of each piece of twin data;

[0037] Calculate the absolute value of the difference between the acquisition time of the last element data in the time series of each piece of real data and the generation time of the first element data in the time series of the corresponding piece of twin data to obtain the time difference value of each piece of twin data;

[0038] Calculate the absolute value of the difference between the theoretical data volume of each piece of real data and the data generation volume of the corresponding piece of twin data to obtain the data volume difference value of each piece of twin data;

[0039] Based on the time difference value and the data volume difference value of each piece of twin data, calculate the abnormal delay degree of each piece of twin data.

[0040] A predictive monitoring and maintenance system for industrial robots based on digital twins provided by the present invention, based on the time difference value and the data volume difference value of each piece of twin data, calculate the abnormal delay degree of each piece of twin data, including:

[0041] Divide the time difference value of each piece of twin data by a preset time difference threshold to obtain a time difference ratio;

[0042] Divide the data volume difference value of each piece of twin data by a preset data volume difference threshold to obtain a data volume difference ratio;

[0043] Multiply the time difference ratio by the data volume difference ratio to calculate the abnormal delay degree of each piece of twin data.

[0044] A predictive monitoring and maintenance system for industrial robots based on digital twins provided by the present invention, wherein the data calculation module calculates the effective reference degree of each piece of twin data based on the abnormal delay degree of each piece of twin data and the effective reference degree of the corresponding piece of real data, including:

[0045] Based on the abnormal delay degree of each piece of twin data, determine the first intermediate parameter value of each piece of twin data;

[0046] Multiply the first intermediate parameter value by a preset adjustment coefficient to obtain the second intermediate parameter value of each piece of twin data;

[0047] The effective reference degree of each piece of real data is subtracted from the second intermediate parameter value of the corresponding piece of twin data, and the effective reference degree of each piece of twin data is calculated.

[0048] According to an industrial robot predictive monitoring and maintenance system based on digital twin provided by the present invention, the monitoring and maintenance module includes:

[0049] A data comparison unit, configured to compare each state prediction value in the state monitoring result with its corresponding state reference value to obtain a comparison result;

[0050] A monitoring and maintenance unit, configured to generate a monitoring and maintenance plan for predictive monitoring and maintenance of key parts of the industrial robot when at least one state prediction value in the comparison result deviates from the corresponding state reference value.

[0051] The present invention has the following beneficial effects:

[0052] Through the data acquisition module, a plurality of core data including real data and twin data of the key parts of the industrial robot are acquired. Through the data processing module, the initial reference degree and the missing core degree of each piece of real data are respectively determined. Based on the missing core degree, the initial reference degree of each piece of real data is corrected to obtain the effective reference degree of each piece of real data. Furthermore, through the data calculation module, the abnormal delay degree of each piece of twin data is determined. Based on the abnormal delay degree of each piece of twin data and the effective reference degree of the corresponding piece of real data, the effective reference degree of each piece of twin data is calculated. Through the machine learning module, the data features of each piece of core data are extracted, and the weight value of the corresponding data feature is set according to the effective reference degree of the real data and the effective reference degree of the twin data in each piece of core data. The data features and weight values of each piece of core data are input into a pre-constructed machine learning model to obtain the state monitoring result output by the machine learning model. Finally, through the monitoring and maintenance module, predictive monitoring and maintenance of the key parts of the industrial robot are carried out according to the state monitoring result. Since the core data on which the monitoring and maintenance link of the key parts of the industrial robot is based not only includes real data but also involves twin data generated through digital twin technology, the data source is more diverse. Subsequently, through the calculation of the effective reference degrees of real data and twin data, more accurate data support can be provided for the generation of the state monitoring result. Combined with the machine learning model, the state monitoring result of the key parts of the industrial robot can be obtained efficiently and accurately, and thus more precise and reliable predictive monitoring and maintenance of the key parts of the industrial robot can be carried out. Description of the Drawings

[0053] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0054] Figure 1 The system structure diagram of a predictive monitoring and maintenance system for industrial robots based on digital twin provided by an embodiment of the present invention;

[0055] Figure 2 It is a distribution schematic diagram of all element data on the time series of the m-th piece of real data;

[0056] Figure 3 For Figure 2 The schematic diagram of the clustering result of all element data on the shown time series. Detailed implementation manners

[0057] In order to further elaborate on the technical means and effects adopted by the present invention to achieve the intended invention purpose, the following will, in conjunction with the drawings and preferred embodiments, detail the specific implementation manners, structures, features, and effects of a predictive monitoring and maintenance system for industrial robots based on digital twin proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0058] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs.

[0059] The following will Figures 1 to 3 Specifically illustrate the specific solution of a predictive monitoring and maintenance system for industrial robots based on digital twin provided by the present invention.

[0060] Please refer to Figure 1 , which shows the system structure diagram of a predictive monitoring and maintenance system for industrial robots based on digital twin provided by an embodiment of the present invention. As Figure 1 shown, the predictive monitoring and maintenance system for industrial robots based on digital twin provided by the embodiment of the present invention specifically includes:

[0061] A data acquisition module 110, configured to acquire multiple core data of key parts of the industrial robot; among them, each core data includes real data and twin data.

[0062] In this embodiment, the key part can be the joint part of an industrial robot. The joint part usually adopts a complex mechatronic design and is mainly composed of a servo motor and a reducer.

[0063] In practical applications, multiple core data can be data that can characterize the health status of the key parts of an industrial robot. Among them, the real data is the data collected by data acquisition devices such as sensors, and the twin data is the data generated using digital twin technology. In this embodiment, the real data can include vibration data, motor current, torque, etc., which can reflect the health status of the key parts. Among them, the statistical period of each piece of real data is a fixed value, such as being collected every 3 seconds. The twin data is virtual data corresponding to the real data such as vibration data, motor current, and torque.

[0064] The data processing module 120 is used to respectively determine the initial reference degree and the missing core degree of each piece of real data, and based on the missing core degree, correct the initial reference degree of each piece of real data to obtain the effective reference degree of each piece of real data.

[0065] In this embodiment, the initial reference degree can characterize the preliminary referenceability of each piece of real data in the subsequent monitoring and maintenance process of the key parts of the industrial robot. The missing core degree can characterize the influence degree of data loss in each piece of real data on the accuracy of representing the state of the subsequent key parts.

[0066] The data calculation module 130 is used to determine the abnormal delay degree of each piece of twin data, and based on the abnormal delay degree of each piece of twin data and the effective reference degree of the corresponding piece of real data, calculate the effective reference degree of each piece of twin data.

[0067] It can be understood that the abnormal delay degree can characterize the degree of delay in the twin data generation link due to abnormal conditions. The effective reference degree of the twin data can characterize the referenceability of the twin data in the subsequent monitoring and maintenance process of the key parts of the industrial robot.

[0068] The machine learning module 140 is used to extract the data features of each piece of core data, set the weight value of the corresponding data feature according to the effective reference degree of the real data and the effective reference degree of the twin data in each piece of core data, and input the data features and weight values of each piece of core data into a pre-constructed machine learning model to obtain the state monitoring result output by the machine learning model.

[0069] In this embodiment, the state monitoring result can characterize the operating condition of the key parts of the industrial robot and can provide an effective data basis for the predictive monitoring and maintenance of the key parts.

[0070] The monitoring and maintenance module 150 is used to perform predictive monitoring and maintenance on the key parts of the industrial robot according to the status monitoring results.

[0071] In the solution provided in this embodiment, since the core data on which the monitoring and maintenance link of the key parts of the industrial robot is based includes not only real data but also twin data generated through digital twin technology, the data source is more diverse. Subsequently, by calculating the effective reference degrees of the real data and the twin data, more accurate data support can be provided for the generation of the status monitoring results. Combining with the machine learning model, the status monitoring results of the key parts of the industrial robot can be obtained efficiently and accurately, and then more precise and reliable predictive monitoring and maintenance can be carried out on the key parts of the industrial robot.

[0072] In one embodiment, the data processing module determines the initial reference degree of each item of real data, specifically including:

[0073] First step, determine the theoretical quantity, actual quantity, time series sequence of the real data, and the time series sequence of the twin data in each item of core data.

[0074] Second step, divide the actual quantity by the theoretical quantity to obtain the quantity ratio.

[0075] Third step, calculate the dynamic time warping distance between the time series sequence of the real data and the time series sequence of the twin data in each item of core data.

[0076] Fourth step, add 1 to the dynamic time warping distance and take the reciprocal to obtain the distance correlation value.

[0077] Fifth step, multiply the quantity ratio by the distance correlation value to calculate the initial reference degree of each item of real data.

[0078] In this embodiment, it is assumed that a total of M items of core data of the key parts of the industrial robot are obtained, and there is real data and corresponding twin data in each item of core data. Since the twin data is generated based on the real data, the real data may be missing during the acquisition process due to reasons such as sensor failures and connection failures. And since the twin data integrates data completion technology, the twin data generated based on the real data is relatively complete.

[0079] It can be understood that the less complete a certain item of real data of the industrial robot is, the lower its referenceability in the subsequent monitoring and maintenance process of the key parts of the industrial robot. Since the twin data is generated based on the real data, the similarity between the same item of real data and the corresponding twin data is relatively high. Therefore, through the integrity of the real data itself and the similarity of the corresponding twin data, the initial reference degree of each item of real data can be determined.

[0080] Specifically, the initial reference degree of the m-th item of real data can be expressed as follows:

[0081]

[0082] where θ m represents the initial reference degree of the m-th item of real data, represents the theoretical quantity of the m-th item of real data, represents the actual quantity of the m-th item of real data, DTW() represents the dynamic time warping distance calculation function, and L m represents the time series sequence of the m-th item of real data, represents the time series sequence of the m-th item of twin data, and +1 is to avoid the denominator being zero.

[0083] It can be understood that the function DTW() mainly realizes the calculation of the DTW (Dynamic Time Warping) distance between two time series sequences. The core idea is to find an optimal warping path through the dynamic programming algorithm to non-linearly align the two time series sequences, so that the two time series sequences can "bend" and "stretch" on the time axis, thereby more accurately measuring the similarity between the two time series sequences, that is, calculating the distance between the two time series sequences.

[0084] In practical applications, due to the key parts of industrial robots, such as joint parts, the data fluctuates greatly under high-speed movement, heavy load conditions, and high-precision machining tasks. If there is data loss or the degree of data loss is too large, the state representation of the key parts of industrial robots will be inaccurate during the subsequent monitoring and maintenance of the key parts of industrial robots.

[0085] Therefore, for the referability of each item of real data, in addition to considering the integrity of the real data itself and the similarity with the corresponding twin data, it is also necessary to consider the position of the missing data. Different processing tasks corresponding to the position of the missing data have different degrees of influence on the state of the key parts of industrial robots. Therefore, the missing core degree of each item of real data can be determined by calculating the missing position of each item of real data.

[0086] In one embodiment, the data processing module determines the missing core degree of each item of real data, specifically including:

[0087] First, determine the fluctuation instability of each element data in the time series sequence of each item of real data.

[0088] In a specific implementation, determining the fluctuation instability of each element data in the time series sequence of each item of real data specifically includes:

[0089] First step, determine the data value, slope value of each element data in the time series of each piece of real data, as well as the data value and slope value of each surrounding data corresponding to each element data.

[0090] In this embodiment, all the element data on the time series of each piece of real data can be classified according to the fluctuation instability degree, and the element data with unstable fluctuations can be classified into the same category. Specifically, the fluctuation instability degree of each element data can be determined by obtaining the fluctuation difference between each element data on the time series of each piece of real data and its two or more surrounding data before and after it. Among them, the number of surrounding data of a single element data can take an empirical value of 10, and the implementer can set it according to the specific actual situation without specific limitation. The surrounding data are the data adjacent to the element data on the time series.

[0091] Specifically, in this embodiment, the total amount of surrounding data of the p-th element data on the time series of the m-th piece of real data is denoted as I, and then the slope value C of the p-th element data on the time series of the m-th piece of real data is calculated m,p , and the slope value C of the i-th surrounding data corresponding to the p-th element data on the time series of the m-th piece of real data is calculated m,pi , the data value of the p-th element data on the time series of the m-th piece of real data is denoted as D m,p , and the data value of the i-th surrounding data corresponding to the p-th element data on the time series of the m-th piece of real data is denoted as D m,pi .

[0092] In practical applications, the slope value of each element data and the slope value of each surrounding data can be obtained by using any slope determination scheme. For example, the time series of the m-th piece of real data can be first fitted, and the function expression of the fitted curve can be determined, and then the slope value can be obtained by taking the derivative according to the function expression and the data value of the element data or the surrounding data. Or, on the basis of knowing the data values of each element data in the time series of the m-th piece of real data, the slope value can be estimated by using the central difference method, the forward difference method or the backward difference method. The specific slope value determination scheme can be reasonably selected according to actual needs and will not be specifically limited here.

[0093] Second step, calculate the absolute value of the difference between the slope value of each element data and the slope value of each surrounding data to obtain the slope difference value of each surrounding data under each element data.

[0094] Third step, calculate the absolute value of the difference between the data value of each element data and the data value of each surrounding data to obtain the numerical difference value of each surrounding data under each element data.

[0095] Step 4: Multiply the slope difference value corresponding to each surrounding data by the numerical difference value to obtain the difference product value corresponding to each surrounding data.

[0096] Step 5: Calculate the average of the difference product values corresponding to all surrounding data under each element data to obtain the fluctuation instability of each element data in the time series of each piece of real data.

[0097] When the slope value and data value of the p-th element data on the time series of the m-th piece of real data differ significantly from those of the surrounding data, it indicates that the fluctuation of the p-th element data is relatively abnormal. In this embodiment, the fluctuation instability of the p-th element data on the time series of the m-th piece of real data can be expressed as follows:

[0098]

[0099] where α m,p represents the fluctuation instability of the p-th element data on the time series of the m-th piece of real data, norm() represents the normalization function, I represents the total amount of surrounding data of the p-th element data on the time series of the m-th piece of real data, C m,p represents the slope value of the p-th element data on the time series of the m-th piece of real data, C m,pi represents the slope value of the i-th surrounding data corresponding to the p-th element data on the time series of the m-th piece of real data, D m,p represents the data value of the p-th element data on the time series of the m-th piece of real data, D m,pi represents the data value of the i-th surrounding data corresponding to the p-th element data on the time series of the m-th piece of real data.

[0100] Then, based on the fluctuation instability, perform clustering on all data elements in the time series of each piece of real data to generate multiple clusters.

[0101] In this embodiment, the k-means clustering method can be used to cluster all element data on the time series of the m-th piece of real data. Figure 2 Exemplarily shows the distribution of all element data on the time series of the m-th piece of real data. During the clustering process, the distance between different element data is represented by the difference in fluctuation instability, and the number of clustering clusters is determined using the silhouette coefficient method. Suppose a total of N clusters are obtained after clustering. The fluctuation instability of the real data in each cluster is similar. Different element data in each cluster correspond to different processing tasks. Figure 3 Exemplarily shows Figure 2 the clustering result of all element data on the shown time series.

[0102] Subsequently, determine the fluctuation instability and data missing amount corresponding to each cluster.

[0103] In this embodiment, the mean value of the fluctuation instability of all element data in each cluster is used as the fluctuation instability corresponding to each cluster. The larger the value of the fluctuation instability corresponding to each cluster, the higher the fluctuation instability of the real data in the cluster. If there is data missing in the time series of the real data in the cluster, the greater the impact on the accuracy of the state representation of the key parts of the subsequent industrial robot.

[0104] In practical applications, a sliding window-based method can be used to calculate all the missing values in the time series of each item of real data, and then calculate the number of missing values within the corresponding range of the real data in each cluster in the time series, that is, the data missing amount.

[0105] Finally, based on the fluctuation instability and data missing amount of each cluster corresponding to each item of real data, the missing core degree of each item of real data is calculated.

[0106] In a specific implementation, based on the fluctuation instability and data missing amount of each cluster corresponding to each item of real data, the missing core degree of each item of real data is calculated, specifically including:

[0107] In the first step, the fluctuation instabilities of all clusters corresponding to each item of real data are summed to obtain the cluster fluctuation sum value.

[0108] In the second step, the fluctuation instability of each cluster is divided by the cluster fluctuation sum value to obtain the missing value influence degree.

[0109] In the third step, the data missing amount of each cluster is multiplied by the missing value influence degree to obtain the cluster missing importance degree of each cluster.

[0110] In the fourth step, the mean values of the cluster missing importance degrees of all clusters corresponding to each item of real data are averaged and normalized to calculate the missing core degree of each item of real data.

[0111] In this embodiment, the missing core degree of the m-th item of real data can be expressed as follows:

[0112]

[0113] Among them, β m represents the missing core degree of the m-th item of real data, norm() represents the normalization function, N represents the number of clusters obtained by clustering, E n represents the data missing amount of the n-th cluster, D n represents the fluctuation instability corresponding to the n-th cluster, D j represents the fluctuation instability corresponding to the j-th cluster, represents the cluster fluctuation sum value of N clusters, Indicates the cluster missing importance of the nth cluster.

[0114] In one embodiment, the data processing module corrects the initial reference degree of each piece of real data based on the degree of missing core to obtain the effective reference degree of each piece of real data, specifically including:

[0115] Multiply the degree of missing core of each piece of real data by the initial reference value to obtain the effective reference degree of each piece of real data.

[0116] In this embodiment, the effective reference degree of the mth piece of real data can be expressed as follows:

[0117]

[0118] Where Indicates the effective reference degree of the mth piece of real data, θ m Indicates the initial reference degree of the mth piece of real data, β m Indicates the degree of missing core of the mth piece of real data.

[0119] In practical applications, it is usually considered that the effective reference degree of each piece of real data is the same as that of the corresponding piece of twin data. However, since twin data is generated based on real data, when there are situations such as network latency and excessive data volume, the generation of twin data will be relatively slow, but the status monitoring of the key parts of the industrial robot is real-time. At this time, if the real data of the key parts of the industrial robot is abnormal and the twin data has not been generated in time or is incompletely generated, the effective reference degree of this piece of twin data will be relatively low.

[0120] If the effective reference degree of the twin data is not reset at this time, it will lead to a relatively low effective reference degree of this piece of core data, and further lead to inaccurate status monitoring results of the key parts of the industrial robot.

[0121] In practical applications, the acquisition time of each element data in each piece of real data, that is, the timestamp information, can be obtained from the background management system, and at the same time, the generation time of the corresponding twin data of this piece of real data can be obtained. There is a fixed waiting time from the generation of real data to the generation of twin data. When the waiting time ends, the status monitoring system of the key parts of the industrial robot starts to run. If there is a delay, the difference between the generation time of the twin data and the acquisition time of the real data will be greater, and at the same time, the data volume of the twin data will also become smaller. Therefore, through the difference between the generation time of the twin data and the acquisition time of the real data and the difference in the data volume between the two, the abnormal delay degree of each piece of twin data can be determined.

[0122] In one embodiment, the data calculation module determines the abnormal delay degree of each piece of twin data, specifically including:

[0123] First step, obtain the acquisition time of the last element data in the time series of each piece of real data, the generation time of the first element data in the time series of each piece of twin data, the theoretical data volume of each piece of real data, and the data generation volume of each piece of twin data.

[0124] In this embodiment, the acquisition time of the last element data in the time series of the m-th piece of real data can be denoted as T m1 , and the generation time of the first element data in the time series of the m-th piece of twin data can be denoted as After the waiting time ends, the data generation volume of the m-th piece of twin data can be denoted as The theoretical data volume of the m-th piece of real data can be denoted as

[0125] Second step, calculate the absolute value of the difference between the acquisition time of the last element data in the time series of each piece of real data and the generation time of the first element data in the time series of the corresponding piece of twin data, and obtain the time difference value of each piece of twin data.

[0126] Third step, calculate the absolute value of the difference between the theoretical data volume of each piece of real data and the data generation volume of the corresponding piece of twin data, and obtain the data volume difference value of each piece of twin data.

[0127] Fourth step, based on the time difference value and the data volume difference value of each piece of twin data, calculate the abnormal delay degree of each piece of twin data.

[0128] In one embodiment, based on the time difference value and the data volume difference value of each piece of twin data, calculating the abnormal delay degree of each piece of twin data specifically includes:

[0129] First, divide the time difference value of each piece of twin data by the preset time difference threshold to obtain the time difference ratio.

[0130] Then, divide the data volume difference value of each piece of twin data by the preset data volume difference threshold to obtain the data volume difference ratio.

[0131] Finally, multiply the time difference ratio by the data volume difference ratio to calculate the abnormal delay degree of each piece of twin data.

[0132] In this embodiment, the abnormal delay degree of the m-th piece of twin data can be expressed as follows:

[0133]

[0134] Among them, γ m represents the abnormal delay degree of the m-th piece of twin data, T m1Denotes the acquisition time of the last element data in the time series of the m-th item of real data, Denotes the generation time of the first element data in the time series of the m-th item of twin data, Denotes the data generation amount of the m-th item of twin data, Denotes the theoretical data amount of the m-th item of real data, T denotes a preset time difference threshold, and A denotes a preset data amount difference threshold.

[0135] In one embodiment, the data calculation module calculates the effective reference degree of each item of twin data based on the abnormal delay degree of each item of twin data and the effective reference degree of the corresponding item of real data, including:

[0136] First, based on the abnormal delay degree of each item of twin data, determine the first intermediate parameter value of each item of twin data.

[0137] Then, multiply the first intermediate parameter value by a preset adjustment coefficient to obtain the second intermediate parameter value of each item of twin data.

[0138] Finally, subtract the effective reference degree of each item of real data from the second intermediate parameter value of the corresponding item of twin data to calculate the effective reference degree of each item of twin data.

[0139] It can be understood that when the abnormal delay degree of the twin data is relatively large, the effective reference degree of the twin data will also be relatively large. Specifically, when the abnormal delay degree of the twin data is greater than 0, it indicates that the twin data has an abnormal delay. At this time, the effective reference degree of the twin data needs to be adjusted downward based on the effective reference degree of the real data. In this embodiment, the effective reference degree of the m-th item of twin data can be expressed as follows:

[0140]

[0141] Among them, ρ m Denotes the effective reference degree of the m-th item of twin data, Denotes the effective reference degree of the m-th item of real data, γ m Denotes the abnormal delay degree of the m-th item of twin data, and F denotes a preset adjustment coefficient, which is a fixed value.

[0142] In practical applications, the effective reference degrees of the real data and the twin data in each core data can be calculated through the above analysis. Then, the data features of each core data can be extracted. Specifically, in this embodiment, the MRMR (Minimum Redundancy Maximum Relevance) method is used to extract the data features of each core data. The MRMR method is a feature selection algorithm, and its core idea is to find a feature subset that has the greatest relevance to the target variable and the least redundancy among features. Among them, relevance refers to the degree of association between a feature and the target variable. The higher the relevance, the more valuable the feature is for predicting the target variable. Redundancy refers to the similarity degree between features. Features with high redundancy contain a lot of repetitive information. Retaining multiple such features has limited improvement on the model performance and may even increase the computational complexity.

[0143] After obtaining the data features of each core data, the weight value of the data feature of each core data can be set according to the effective reference degrees of the real data and the twin data in each core data. Subsequently, the data features and weight values of each core data are input into a pre-constructed machine learning model, and then the state monitoring results of the key parts of the industrial robot are output.

[0144] In this embodiment, the machine learning model can adopt the PNN (Product-based Neural Network) architecture. The PNN architecture uses a multi-layer neural network, which can explicitly perform cross operations on data features and effectively capture the second-order and higher-order interaction relationships between data features. Compared with traditional neural networks, it can better handle complex relationships in sparse data. In practical applications, the ReLU function can be used as the activation function.

[0145] In some embodiments, the process of setting the weight value of the data feature of each core data according to the effective reference degrees of the real data and the twin data in each core data can be specifically implemented through the following process:

[0146] First, calculate the mean value of the effective reference degrees of the real data and the twin data in each core data to obtain the comprehensive reference degree of each core data.

[0147] Then, sum up the comprehensive reference degrees of all core data to obtain the comprehensive reference sum value.

[0148] Subsequently, divide the comprehensive reference degree of each core data by the comprehensive reference sum value to obtain the comprehensive reference ratio.

[0149] Finally, use the comprehensive reference ratio corresponding to each core data as the weight value of the data feature of each core data.

[0150] In one embodiment, the monitoring and maintenance module specifically includes:

[0151] A data comparison unit, configured to compare each state prediction value in the state monitoring result with its corresponding state reference value to obtain a comparison result.

[0152] A monitoring and maintenance unit, configured to generate a monitoring and maintenance plan for predictive monitoring and maintenance of key parts of the industrial robot when at least one state prediction value in the comparison result deviates from the corresponding state reference value.

[0153] In this embodiment, taking the joint part as an example of the key part, the state monitoring result may specifically include various state prediction values such as joint parameters, motor parameters, and mechanical structure parameters. Among them, the joint parameters specifically include parameter information such as joint angle, joint speed, joint acceleration, and joint torque that can characterize the joint motion state. The motor parameters specifically include parameter information such as operating current, operating voltage, operating temperature, and operating speed that can characterize the motor working condition state. The mechanical structure parameters specifically include parameter information such as mechanical vibration, mechanical noise, mechanical component clearance, and wear degree that can characterize the mechanical structure health condition.

[0154] In practical applications, the state reference value corresponding to each state prediction value in the state monitoring result can be set, and then each state prediction value is respectively compared with its corresponding state reference value. If any one or more state prediction values deviate from the state reference value, it indicates that there are health hazards in the key parts of the industrial robot and timely maintenance is required.

[0155] In some embodiments, the allowable deviation amount between each state prediction value and the state reference value can be set. When the difference between any state prediction value and the state reference value exceeds the preset allowable deviation amount, it can be determined that the state prediction value deviates from the state reference value.

[0156] In this embodiment, a corresponding monitoring and maintenance plan can be established in advance under the condition that each state prediction value deviates from the state reference value, and a monitoring and maintenance decision table is generated. Subsequently, the corresponding monitoring and maintenance plan can be extracted from the monitoring and maintenance decision table according to the comparison result.

[0157] In summary, in this embodiment, by analyzing the missing situation of the real data of the key parts of the industrial robot and the similarity with the corresponding twin data, the initial reference degree of each real data is determined, and then the initial reference degree of each real data is corrected by analyzing the missing position of the real data. Subsequently, the effective reference degree of the twin data is determined by analyzing the delay situation of the twin data corresponding to each real data, so as to provide accurate and reliable data basis for subsequent state monitoring of the key parts of the industrial robot using machine learning models.

[0158] It should be noted that the above order of the embodiments of the present invention is only for description and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0159] Each embodiment in this specification is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other. Each embodiment focuses on the differences from other embodiments.

Claims

1. A predictive monitoring and maintenance system for industrial robots based on digital twins, characterized in that, The system includes: A data acquisition module for acquiring multiple core data of key parts of an industrial robot; wherein, each piece of the core data includes real data and twin data; A data processing module for respectively determining the initial reference degree and the missing core degree of each piece of real data, and correcting the initial reference degree of each piece of real data based on the missing core degree to obtain the effective reference degree of each piece of real data; A data calculation module for determining the abnormal delay degree of each piece of twin data, and calculating the effective reference degree of each piece of twin data based on the abnormal delay degree of each piece of twin data and the effective reference degree of the corresponding piece of real data; A machine learning module for extracting data features of each piece of core data, setting weight values for the data features of the corresponding item according to the effective reference degree of the real data and the effective reference degree of the twin data in each piece of core data, and inputting the data features and weight values of each piece of core data into a pre-constructed machine learning model to obtain the status monitoring result output by the machine learning model; A monitoring and maintenance module for performing predictive monitoring and maintenance on the key parts of the industrial robot according to the status monitoring result.

2. The predictive monitoring and maintenance system for an industrial robot based on digital twin according to claim 1, wherein The data processing module determines the initial reference degree of each piece of real data, including: Determining the theoretical quantity, actual quantity, time series of real data, and time series of twin data in each piece of core data; Dividing the actual quantity by the theoretical quantity to obtain a quantity ratio; Calculating the dynamic time warping distance between the time series of real data and the time series of twin data in each piece of core data; Adding 1 to the dynamic time warping distance and taking the reciprocal to obtain a distance correlation value; Multiplying the quantity ratio by the distance correlation value to calculate the initial reference degree of each piece of real data.

3. The predictive monitoring and maintenance system for an industrial robot based on digital twin according to claim 1, characterized in that, The data processing module determines the missing core degree of each piece of real data, including: Determining the fluctuation instability of each element data in the time series of each piece of real data; Based on the fluctuation instability, performing clustering processing on all data elements in the time series of each piece of real data to generate multiple clusters; Determining the fluctuation instability and data missing amount corresponding to each cluster; Calculating the missing core degree of each piece of real data based on the fluctuation instability and data missing amount of each cluster corresponding to each piece of real data.

4. The predictive monitoring and maintenance system for an industrial robot based on digital twin according to claim 3, wherein, Determining the fluctuation instability of each element data in the time series of each piece of real data, including: Determining the data value, slope value of each element data in the time series of each piece of real data, and the data value and slope value of each surrounding data corresponding to each element data; Calculating the absolute value of the difference between the slope value of each element data and the slope value of each surrounding data to obtain the slope difference value of each surrounding data under each element data; Calculating the absolute value of the difference between the data value of each element data and the data value of each surrounding data to obtain the numerical difference value of each surrounding data under each element data; Multiplying the slope difference value corresponding to each surrounding data by the numerical difference value to obtain the difference product value corresponding to each surrounding data; Calculate the mean of the difference product values corresponding to all surrounding data under each element data to obtain the fluctuation instability of each element data in the time series of each item of real data.

5. The predictive monitoring and maintenance system for an industrial robot based on digital twin according to claim 3, characterized in that, Based on the fluctuation instability and data missing amount of each cluster corresponding to each item of real data, calculate the missing core degree of each item of real data, including: Sum the fluctuation instability of all clusters corresponding to each item of real data to obtain the cluster fluctuation sum value; Divide the fluctuation instability of each cluster by the cluster fluctuation sum value to obtain the missing value influence degree; Multiply the data missing amount of each cluster by the missing value influence degree to obtain the cluster missing importance of each cluster; Calculate the mean of the cluster missing importance of all clusters corresponding to each item of real data and perform normalization processing to calculate the missing core degree of each item of real data.

6. The predictive monitoring and maintenance system for an industrial robot based on digital twin according to claim 1, characterized in that, The data processing module corrects the initial reference degree of each item of real data based on the missing core degree to obtain the effective reference degree of each item of real data, including: Multiply the missing core degree of each item of real data by the initial reference value to obtain the effective reference degree of each item of real data.

7. The predictive monitoring and maintenance system for an industrial robot based on digital twin according to claim 1, wherein, The data calculation module determines the abnormal delay degree of each item of twin data, including: Obtain the acquisition time of the last element data in the time series of each item of real data, the generation time of the first element data in the time series of each item of twin data, the theoretical data volume of each item of real data, and the data generation volume of each item of twin data; Calculate the absolute value of the difference between the acquisition time of the last element data in the time series of each item of real data and the generation time of the first element data in the time series of the corresponding item of twin data to obtain the time difference value of each item of twin data; Calculate the absolute value of the difference between the theoretical data volume of each item of real data and the data generation volume of the corresponding item of twin data to obtain the data volume difference value of each item of twin data; Based on the time difference value and the data volume difference value of each item of twin data, calculate the abnormal delay degree of each item of twin data.

8. The predictive monitoring and maintenance system for an industrial robot based on digital twin according to claim 7, wherein, Based on the time difference value and the data volume difference value of each item of twin data, calculate the abnormal delay degree of each item of twin data, including: Divide the time difference value of each item of twin data by the preset time difference threshold to obtain the time difference ratio; Divide the data volume difference value of each item of twin data by the preset data volume difference threshold to obtain the data volume difference ratio; Multiply the time difference ratio by the data volume difference ratio to calculate the abnormal delay degree of each item of twin data.

9. The predictive monitoring and maintenance system for an industrial robot based on digital twin according to claim 1, wherein, The data calculation module calculates the effective reference degree of each item of twin data based on the abnormal delay degree of each item of twin data and the effective reference degree of the corresponding item of real data, including: Based on the abnormal delay degree of each item of twin data, determine the first intermediate parameter value of each item of twin data; Multiply the first intermediate parameter value by the preset adjustment coefficient to obtain the second intermediate parameter value of each item of twin data; Subtract the second intermediate parameter value of each item of twin data from the effective reference degree of the corresponding item of real data to calculate the effective reference degree of each item of twin data.

10. The predictive monitoring and maintenance system for an industrial robot based on digital twin according to claim 1, wherein, The monitoring and maintenance module includes: A data comparison unit for comparing each state prediction value in the state monitoring result with its corresponding state reference value to obtain a comparison result; A monitoring and maintenance unit for generating a monitoring and maintenance plan for predictive monitoring and maintenance of key parts of the industrial robot when at least one state prediction value in the comparison result deviates from the corresponding state reference value.

Images