Data management method based on digital twin intelligent factory

Abstract

The invention relates to the field of electric data processing, in particular to a data management method based on a digital twin smart factory, and the method comprises the steps: obtaining the related data of the digital twin smart factory, carrying out the preprocessing, clustering, obtaining a clustering result, calculating the Gaussian density of each data point in each clustering cluster, and calculating the Gaussian density of each data point in each clustering cluster; obtaining the high-density optimization degree of each data point according to the Gaussian density of all data points in each cluster, and determining a high-density center point; counting the number of added real-time data points, classifying the data points into each cluster according to Euclidean distance, dynamically updating the average real-time distance of the data points in each cluster, analyzing and adjusting necessity according to the real-time distance, and judging whether to iterate a high-density center point or not so as to measure the uniformity of data distribution and complete digital twinning intelligent factory data management. According to the method, the clustering center is dynamically optimized, so that the timeliness, the accuracy and the early warning reliability of digital twin factory data classification are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of electrical data processing. In particular, it relates to a data management method based on a digital twin intelligent factory. Background Art

[0002] In the wave of industrial digital transformation, digital twin technology, as an emerging technology, brings new ideas and methods for the development of intelligent factories, helps to achieve a higher level of intelligence, and realizes the deep integration of virtual and real by constructing a virtual model to accurately map various elements such as equipment, production lines, and process flows in the physical factory. Digital twin processes factory data in real time, enabling managers to comprehensively control the operating status in the virtual space and providing accurate basis for production decisions, which is the key for intelligent factories to move towards a higher level of intelligence.

[0003] For intelligent factories, data is like the lifeblood, carrying key information such as equipment operation, production progress, and quality control. Digital twin integrates sensor, equipment status, process, and business system data to form an efficient data network, and constructs a hierarchical early warning mechanism based on real-time data categories to accurately monitor anomalies in equipment, quality, energy consumption, etc., and give early warnings in a timely manner to ensure stable and efficient production. However, due to the variety of data types, the data classification link faces technical bottlenecks, affecting the overall data management efficiency.

[0004] Currently, the K-means algorithm is often used for data classification. However, digital twin requires data to be updated at the minute level, and the iterative convergence speed of K-means cannot meet the requirements. When dealing with large-scale and highly dynamic data, the algorithm needs to iterate multiple times to determine the clustering center, resulting in data accumulation and classification lag in real-time scenarios, and unable to extract key features in a timely manner, affecting the response and effectiveness of the hierarchical early warning, which becomes an important bottleneck restricting the application of digital twin in intelligent factory data management. Summary of the Invention

[0005] To solve the problem that when the K-means algorithm is applied to data classification, the iterative convergence speed is difficult to match the minute-level data update requirement of digital twin, resulting in data classification lag, untimely extraction of key features, and affecting the effectiveness of the hierarchical early warning mechanism, the present invention provides solutions in the following aspects.

[0006] A data management method for a digital twin intelligent factory includes: obtaining relevant data of the digital twin intelligent factory and performing preprocessing, where the relevant data includes: equipment temperature, vibration, rotation speed, and energy consumption; clustering based on the preprocessed relevant data to obtain a clustering result, calculating the Gaussian density of each data point in each clustering cluster, obtaining the high-density preference degree of each data point in each clustering cluster based on the Gaussian density of all data points in each clustering cluster, and obtaining the data point with the highest high-density preference degree in each clustering cluster as the high-density center point; counting the number of real-time data points added and classifying them into each clustering cluster according to the Euclidean distance, dynamically updating the average real-time distance of the data points in each clustering cluster, analyzing the necessity of adjustment based on the real-time distance, and judging whether to iterate the high-density center point according to the adjustment necessity to measure the uniformity of data distribution, completing the iteration of the high-density center point to obtain a classification result, and completing the data management of the digital twin intelligent factory according to the classification result.

[0007] By obtaining and preprocessing data such as equipment temperature, vibration, rotation speed, and energy consumption, using clustering analysis to determine the Gaussian density and high-density center point of each data point; by dynamically updating the average real-time distance of the data points to analyze the necessity of adjustment, and iterating the high-density center point accordingly to complete data classification and management. It can classify data in a timely and efficient manner, extract key features, improve the accuracy and reliability of data management, and support intelligent decision-making.

[0008] Preferably, the calculation method of the Gaussian density includes: Taking any clustering cluster in the clustering result as the target clustering cluster, calculating the square of the distance between each data point in the target clustering cluster and other data points, and calculating the sum of the squares of the standard deviations of each item of data in the target clustering cluster; Taking the ratio between the square of the distance and the sum of the squares of the standard deviations, performing exponential decay using the negative exponential function, and obtaining the average value of the exponential decay terms corresponding to all data points in the target clustering cluster, to obtain the Gaussian density of each data point in the target clustering cluster.

[0009] By calculating the ratio of the square of the distance between each data point in the clustering cluster and the sum of the squares of the standard deviations of other data points, and performing exponential decay processing using the negative exponential function, and taking the average value of the sum, the Gaussian density of each data point in the target clustering cluster can be accurately evaluated. It enhances the quantization accuracy of the local density of the data points, improves the accuracy and reliability of clustering analysis, and helps to more accurately identify the data distribution characteristics and high-density regions.

[0010] Preferably, the calculation method of the high-density preference degree includes: Taking any clustering cluster as the target clustering cluster, the average value of the distances between all data points in the target clustering cluster and the center is used as the neighborhood radius. The sum of the Gaussian densities of the data points belonging to the target clustering cluster within the neighborhood radius of each data point in the target clustering cluster is statistically calculated, and a negative exponential function is used for mapping. Subtracting the mapped result from 1 gives the overall density of the data within the neighborhood radius. Calculate the difference between the Gaussian density of each data point and the average Gaussian density of all data points in the target clustering cluster, divide the difference by the standard deviation of the Gaussian densities of all data points in the target clustering cluster for normalization processing, and then use an exponential function for mapping to obtain the density difference coefficient. Multiply the overall density by the density difference coefficient to obtain the high-density preference degree of each data point in the target clustering cluster.

[0011] By calculating the sum of Gaussian densities within the neighborhood radius of data points and mapping to obtain the overall density, and at the same time calculating the density difference coefficient, and combining the two to obtain the high-density preference degree of each data point, the relative importance and density characteristics of data points in the clustering cluster can be accurately evaluated, the key data points in the high-density area can be effectively identified, and the accuracy of clustering analysis and the efficiency of data management can be improved.

[0012] Preferably, the average real-time distance includes: Taking any clustering cluster as the target clustering cluster, calculate the range of each dimension data corresponding to all data points in the target clustering cluster, divide the average value of all dimension ranges by the number of real-time data points, and obtain the average real-time distance of the target clustering cluster.

[0013] By evaluating the uniformity and overall characteristics of the distribution of data points within the clustering cluster, a quantitative index is provided. Specifically, the average real-time distance reflects the distribution range of data points within the clustering cluster in each feature dimension. Through normalization processing, it enables fair comparison of the data distributions of clustering clusters of different scales. It helps to detect changes in data distribution in a timely manner, thereby optimizing the clustering results and improving the accuracy and adaptability of intelligent factory data management.

[0014] Preferably, the calculation method of the necessity for adjustment includes: Taking any clustering cluster as the target clustering cluster, calculate the ratio of the average real-time distance of the target clustering cluster to the real-time distance after each real-time data is added to the target clustering cluster, and then take the absolute difference between 1 and the ratio as the deviation degree of the real-time data point. Calculate the average value of the deviation degrees of all real-time data points in each clustering cluster to obtain the necessity for adjustment of the target clustering cluster.

[0015] By quantitatively evaluating the uniformity of data point distribution in clusters, it is possible to keenly capture changes in data distribution caused by the addition of real-time data. When the necessity of adjustment exceeds the preset threshold, it can trigger the iterative update of high-density center points in a timely manner to ensure that the cluster center is always in the center of the high-density area, thereby effectively improving the accuracy and stability of clustering results and enhancing the adaptability and reliability of the smart factory data management system based on digital twins.

[0016] Preferably, judging whether to iterate the high-density center point according to the necessity of adjustment includes: In response to the necessity of adjustment being greater than a preset threshold, the high-density preference of the data points in each cluster is recalculated, and the high-density center point is updated to ensure that the cluster center point is always in the center of the high-density area, thereby adapting to changes in data distribution and maintaining the accuracy of the clustering results.

[0017] Preferably, the high-density center point also includes: In response to the presence of multiple data points in a cluster with similar high-density preference levels and exceeding a preset high-density threshold, the data points are marked as center point candidates, and the average distance and average density of each center candidate point and other high-density data points in the cluster are calculated, and the data points with an average distance less than the preset distance threshold and an average density greater than the preset high-density threshold are selected as the final high-density center points.

[0018] Preferably, the obtaining of relevant data of the digital twin smart factory and preprocessing includes: Check the missing values ​​of the relevant data and fill them in using interpolation methods, and eliminate data that is out of a reasonable range; standardize the data to eliminate dimensional differences, and align the data according to timestamps to complete the preprocessing of the relevant data.

[0019] The present invention has the following effects: 1. The present invention can effectively adapt to the minute-level data update requirements in digital twin smart factories by dynamically updating high-density center points and iterating according to the necessity of adjustment. Compared with the K-means algorithm, it effectively avoids the data classification lag problem caused by insufficient iterative convergence speed, thereby completing data classification more efficiently and extracting key features in time, significantly improving the effectiveness and timeliness of the hierarchical early warning mechanism.

[0020] 2. The present invention can accurately determine the center point of the cluster through the calculation of Gaussian density and high-density optimization degree, which not only improves the accuracy of data classification, but also makes data management more reliable. By continuously monitoring and analyzing the uniformity of data distribution and updating the center point when necessary, the actual distribution of data can be better reflected, the classification error caused by changes in data distribution can be reduced, and the stability and credibility of smart factory data management can be enhanced. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a method flow chart of steps S1 to S3 in the data management method based on the digital twin smart factory in an embodiment of the present invention. DETAILED DESCRIPTION

[0022] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments.

[0023] Reference Figure 1 The data management method based on the digital twin smart factory includes steps S1 to S3, which are as follows: S1: Obtain relevant data of the digital twin smart factory and perform preprocessing. The relevant data includes: equipment temperature, vibration, rotation speed and energy consumption.

[0024] It should be noted that the relevant data include but are not limited to: real-time data such as equipment temperature, vibration, rotation speed, energy consumption, etc., and the data collection cycle is set to , obtain all collected factory equipment data, and The first collection Item data is recorded as , recorded as historical data, and the total number of historical data collected is recorded as , the number of data items collected each time is recorded as .

[0025] Preprocessing includes: Check the missing values ​​of the relevant data and fill them in using interpolation methods, and eliminate data that is out of a reasonable range; standardize the data to eliminate dimensional differences, and align the data according to timestamps to complete the preprocessing of the relevant data.

[0026] Further analysis shows that in the process of smart factory data management, classifying smart factory data can significantly improve the efficiency of data management. Although the K-means algorithm is often used for data classification management, the conventional K-means algorithm needs to iterate data clustering and is easily affected by abnormal data or biased data during the iterative clustering process, resulting in inaccurate data clustering results.

[0027] S2: Perform clustering according to the preprocessed relevant data to obtain clustering results, and calculate the Gaussian density of each data point in each cluster. According to the Gaussian density of all data points in each cluster, obtain the high-density preference degree of each data point, and obtain the data point with the highest high-density preference degree in each cluster as the high-density center point.

[0028] The calculation method of Gaussian density includes: Taking any cluster in the clustering result as the target cluster, calculating the square of the distance between each data point in the target cluster and other data points, and summing the squares of the standard deviations of each item of data in the target cluster; Taking the ratio between the sum of the squares of the distances and the sum of the squares of the standard deviations, performing exponential decay using the negative exponential function, and obtaining the average value of the sum of the exponential decay terms corresponding to all data points in the target cluster, to get the Gaussian density of each data point in the target cluster.

[0029] Specifically, the Gaussian density satisfies the following relationship: ; In the formula, represents the Gaussian density of the th data point in the th cluster, represents the number of data points in the th cluster, represents the square of the Euclidean distance between the th data point and the th data point, represents the standard deviation of the th item of data in the th cluster, represents the exponential function with the natural number as the base.

[0030] That is to say, the Euclidean distance measures the straight-line distance between two data points in space, and the square operation amplifies the difference in distances, making the contribution of farther data points to the density significantly decrease. It reflects the similarity between two data points. The smaller the distance, the higher the similarity and the greater the contribution to the density; the standard deviation measures the degree of dispersion of data in each dimension. By summing the squares of the standard deviations of each dimension and taking the square root, a scalar is obtained that comprehensively reflects the degree of dispersion of data within the cluster. is used to normalize the Euclidean distance, making the distance metrics between different clusters comparable, and avoiding the deviation of distance metrics caused by larger dimensions or higher degrees of dispersion of data in some dimensions.

[0031] Further analysis shows that, based on the Gaussian density of data points in each cluster, the density preference degree of data points is obtained, so that the data points with higher Gaussian density themselves and higher Gaussian density around them than the average level of the Gaussian density of all data points in the cluster have a higher high-density preference degree; the data point with the highest high-density preference degree in each cluster is recorded as the high-density center point.

[0032] It should be noted that when selecting high-density central points, to ensure that the selected central points have both a relatively high local density and are located in the data aggregation area, the present invention calculates the high-density preference degree by comprehensively considering the Gaussian density of the data points themselves and the density distribution of their neighboring data points. Specifically, the high-density preference degree of a data point not only depends on its own Gaussian density value but is also affected by the density distribution of its surrounding neighboring data points, so as to more accurately identify the central points that truly represent the data distribution characteristics.

[0033] The calculation method of the high-density preference degree includes: Taking any clustering cluster as the target clustering cluster, taking the average value of the distances between all data points in the target clustering cluster and the center as the neighborhood radius, counting the sum of the Gaussian densities of the data points belonging to the target clustering cluster within the neighborhood radius of each data point in the target clustering cluster, and using the negative exponential function for mapping, and taking 1 minus the mapped result as the overall density degree of the data within the neighborhood radius; Calculating the difference between the Gaussian density of each data point and the average Gaussian density of all data points in the target clustering cluster, dividing the difference by the standard deviation of the Gaussian densities of all data points in the target clustering cluster for normalization processing, and then using the exponential function for mapping to obtain the density difference coefficient; Multiplying the overall density degree by the density difference coefficient as the high-density preference degree of each data point in the target clustering cluster.

[0034] Specifically, the high-density preference degree satisfies the following relational expression: ; In the formula, represents the high-density preference degree of the th data point in the th clustering cluster, represents the number of data points within the neighborhood radius of the th data point in the th clustering cluster, represents the Gaussian density of the th neighboring data point of the th data point in the th clustering cluster, represents the Gaussian density of the th data point in the th clustering cluster, represents the average Gaussian density of the data points in the th clustering cluster, represents the standard deviation of the Gaussian densities of all data points in the th clustering cluster, represents the exponential function with the natural number as the base.

[0035] That is to say, It is used to represent the overall density characteristics of data points in the neighborhood, reflecting the density of data points in the neighborhood. The more neighborhood data points and the greater their Gaussian density, it indicates that the data points are in a data point cluster with a higher Gaussian density, and it is considered that the higher the preference degree of the data points for high density. Using the negative exponential function, the neighborhood density is transformed into within the range of , but further explanation is that the complement of the mapping result is selected, so that the denser the data in the neighborhood, the closer the value is to 1; the sparser the data in the neighborhood, the closer the value is to 0; represents the density level of the th data point in the th clustering cluster. The higher the value, the higher the density level of the data point in the clustering cluster and the higher the preference degree for high density.

[0036] The high-density center point further includes: In response to the situation that the preference degrees for high density of multiple data points in the clustering cluster are similar and exceed the preset high-density threshold, the data points are marked as center point candidates, and the average distance and average density between each center candidate point and other high-density data points in the clustering cluster are calculated. The data points with an average distance less than the preset distance threshold and an average density greater than the preset high-density threshold are selected as the final high-density center points.

[0037] Exemplarily, the preset high-density threshold is 0.8, and the preset distance threshold is 2.5 times the standard deviation of the data distribution. Implementers can adjust according to specific situations.

[0038] Further analysis shows that in the process of classifying time-series data, due to the continuous addition of time-series data, the distribution of data points in each clustering cluster changes, thus changing the aggregation center of the data points. Such changes need to be considered, and the position of the high-density center point is adjusted so that the high-density center point can always meet the distribution of data points in the intelligent factory.

[0039] S3: Count the number of real-time data points added and classify them into each clustering cluster according to the Euclidean distance, dynamically update the average real-time distance of the data points in each clustering cluster, analyze the necessity for adjustment according to the real-time distance, judge whether to iterate the high-density center point according to the necessity for adjustment, so as to measure the degree of uniformity of data distribution, complete the iteration of the high-density center point, obtain the classification result, and complete the intelligent factory data management of the digital twin according to the classification result.

[0040] When the number of real-time data points included in all clustering clusters is greater than 2, the necessity for adjustment is obtained according to the real-time distance of the real-time data points in all clustering clusters, as follows: The average real-time distance includes: Taking any clustering cluster as the target clustering cluster, calculate the range of the data corresponding to each dimension for all data points in the target clustering cluster, and divide the average of all dimension ranges by the number of real-time data points to obtain the average real-time distance of the target clustering cluster.

[0041] Specifically, the average real-time distance satisfies the following relational expression: ; In the formula, represents the average real-time distance of the th clustering cluster, represents the number of real-time data points included in the th clustering cluster, represents the number of data items of each data point in the clustering cluster, represents the range of the th item of data of the th clustering cluster.

[0042] That is to say, reflects the distribution range of the data points within the clustering cluster in each feature dimension, comprehensively considering the range of each feature dimension and the number of data points, and is used to measure the overall distribution of the data points within the clustering cluster. By calculating the average real-time distance of each clustering cluster, it comprehensively reflects the distribution of the data points within the clustering cluster in each feature dimension. It helps to evaluate the distribution uniformity and overall characteristics of the data points within the clustering cluster, and provides a basis for subsequent clustering analysis and center point adjustment.

[0043] The calculation method of the necessity for adjustment includes: Taking any clustering cluster as the target clustering cluster, calculate the ratio of the average real-time distance of the target clustering cluster to the real-time distance after each real-time data is added to the target clustering cluster, and then take the absolute difference between 1 and the ratio as the deviation degree of the real-time data point. Calculate the average of the deviation degrees of all real-time data points in each clustering cluster to obtain the necessity for adjustment of the target clustering cluster.

[0044] Specifically, the necessity for adjustment satisfies the following relational expression: ; In the formula, represents the necessity for adjustment, represents the total number of clustering clusters, represents the number of real-time data points included in the th clustering cluster, represents the average real-time distance of the th clustering cluster, represents the real-time distance of the th real-time data point of the th clustering cluster.

[0045] That is to say, represents the average value of the real-time distances in the th clustering cluster. According to the range of each item of data in the clustering cluster, the average level of the real-time distances of the real-time data points therein is obtained. represents the degree of deviation of the real-time distance of the real-time data point from the average distance of the clustering cluster. The larger the deviation value, the greater the difference between the real-time distance of the real-time data point and the average distance, and the more special the position of the data point in the clustering cluster, which indicates that the distribution of the real-time data points in the clustering cluster is more uneven and more adjustment is needed. By calculating the average value of the deviation values of all data points, the overall uniformity of the distribution of the data points within the clustering cluster can be obtained. If the average deviation value is large, it indicates that the distribution of the data points within the clustering cluster is uneven and there may be outliers or boundary points; if the average deviation value is small, it indicates that the data points are more evenly distributed.

[0046] It should be noted that in any clustering cluster, the more evenly the newly added real-time data points are distributed in the clustering cluster, the smaller the impact on the overall Gaussian density of all data points. Therefore, the necessity of adjustment is defined by the degree of uniformity of the real-time data points in the clustering cluster.

[0047] Judge whether to iterate the high-density center point according to the necessity of adjustment, including: When the necessity of adjustment is greater than the preset threshold, recalculate the high-density preference degree of the data points in each clustering cluster and update the high-density center point to ensure that the clustering cluster center point is always at the center of the high-density area, so as to adapt to the change of data distribution and maintain the accuracy of the clustering result.

[0048] Exemplarily, the preset threshold is , and the implementer can adjust it according to the actual implementation situation to complete the iteration of the high-density center point.

[0049] Classify and manage the intelligent factory data according to the data classification result, strengthen the data classification management ability of digital twin real-time data, and make preparations for subsequent intelligent decision-making.

[0050] It should be pointed out that for those of ordinary skill in the art, without departing from the concept of the present invention, several deformations and improvements can still be made, and these all belong to the protection scope of the present invention. Therefore, the protection scope of this invention patent shall be subject to the appended claims.

Claims

1. A data management method for a digital twin intelligent factory, characterized in that, Including: Obtain the relevant data of the digital twin intelligent factory and perform preprocessing. Among them, the relevant data includes: equipment temperature, vibration, rotation speed, and energy consumption. Cluster according to the preprocessed relevant data, obtain the clustering result, calculate the Gaussian density of each data point in each cluster, obtain the high-density preference degree of each data point according to the Gaussian density of all data points in each cluster, and obtain the data point with the highest high-density preference degree in each cluster as the high-density center point. Count the number of real-time data points added and classify them into each cluster according to the Euclidean distance, dynamically update the average real-time distance of the data points in each cluster, analyze the necessity of adjustment according to the real-time distance, and judge whether to iterate the high-density center point according to the necessity of adjustment to measure the uniformity of data distribution, complete the iteration of the high-density center point, obtain the classification result, and complete the data management of the digital twin intelligent factory according to the classification result.

2. The data management method based on the digital twin intelligent factory according to claim 1, wherein The calculation method of the Gaussian density includes: Taking any cluster in the clustering result as the target cluster, calculate the square of the distance between each data point in the target cluster and other data points, and calculate the sum of the squares of the standard deviations of each item of data in the target cluster. Taking the ratio between the square of the distance and the sum of the squares of the standard deviations, performing exponential decay using the negative exponential function, and obtaining the average value of the exponential decay terms corresponding to all data points in the target cluster, to obtain the Gaussian density of each data point in the target cluster.

3. The data management method based on the digital twin intelligent factory according to claim 1, wherein The calculation method of the high-density preference degree includes: Taking any cluster as the target cluster, using the average value of the distances between all data points in the target cluster and the center as the neighborhood radius, counting the sum of the Gaussian densities of the data points belonging to the target cluster within the neighborhood radius of each data point in the target cluster, and performing mapping using the negative exponential function, and taking 1 minus the mapped result as the overall density of the data within the neighborhood radius; calculating the difference between the Gaussian density of each data point and the average Gaussian density of all data points in the target cluster, dividing the difference by the standard deviation of the Gaussian densities of all data points in the target cluster for normalization processing, and then performing mapping using the exponential function to obtain the density difference coefficient; multiplying the overall density by the density difference coefficient as the high-density preference degree of each data point in the target cluster.

4. The data management method based on a digital twin intelligent factory according to claim 1, characterized in that The average real-time distance includes: Taking any cluster as the target cluster, calculating the range of each dimension data corresponding to all data points in the target cluster, and dividing the average value of all dimension ranges by the number of real-time data points to obtain the average real-time distance of the target cluster.

5. The data management method based on the digital twin intelligent factory according to claim 1, characterized in that, The calculation method of the necessity of adjustment includes: Taking any cluster as the target cluster, calculating the ratio of the average real-time distance of the target cluster to the real-time distance after each real-time data is added to the target cluster, and then taking the absolute difference between 1 and the ratio as the deviation degree of the real-time data point, calculating the average value of the deviation degrees of all real-time data points in each cluster to obtain the necessity of adjustment of the target cluster.

6. The data management method based on the digital twin intelligent factory according to claim 1, characterized in that Judging whether to iterate the high-density center point according to the necessity of adjustment includes: In response to the necessity of adjustment being greater than a preset threshold, the high-density preference of the data points in each cluster is recalculated, and the high-density center point is updated to ensure that the cluster center point is always in the center of the high-density area, thereby adapting to changes in data distribution and maintaining the accuracy of the clustering results.

7. The data management method based on a digital twin intelligent factory according to claim 1, characterized in that The high-density center points also include: In response to the presence of multiple data points in a cluster with similar high-density preference levels and exceeding a preset high-density threshold, the data points are marked as center point candidates, and the average distance and average density of each center candidate point and other high-density data points in the cluster are calculated, and the data points with an average distance less than the preset distance threshold and an average density greater than the preset high-density threshold are selected as the final high-density center points.

8. The data management method based on a digital twin intelligent factory according to claim 1, characterized in that The obtaining of relevant data of the digital twin smart factory and preprocessing thereof include: Check the missing values ​​of the relevant data and fill them in using interpolation methods, and eliminate data that is out of a reasonable range; standardize the data to eliminate dimensional differences, and align the data according to timestamps to complete the preprocessing of the relevant data.

Images